✨ v0.7.5-rc2 (#3976)

* ✨ v0.7.5-rc2

* docs: update README

* refactor(settings): Update rememberForkOption default value

* a11y: proper screen reader announcements for content blocks

* Update version to 0.7.423 in package-lock.json and packages/data-provider/package.json

* chore: rename rememberForkOption -> rememberDefaultFork to apply new default value

* fix: headlessui menu stealing focus from Settings Dialog when pressing Enter

.env.example

@@ -65,6 +65,7 @@ PROXY=
 # ANYSCALE_API_KEY=
 # APIPIE_API_KEY=
 # COHERE_API_KEY=
+# DEEPSEEK_API_KEY=
 # DATABRICKS_API_KEY=
 # FIREWORKS_API_KEY=
 # GROQ_API_KEY=
@@ -74,6 +75,7 @@ PROXY=
 # PERPLEXITY_API_KEY=
 # SHUTTLEAI_API_KEY=
 # TOGETHERAI_API_KEY=
+# UNIFY_API_KEY=
 
 #============#
 # Anthropic  #
@@ -87,7 +89,6 @@ ANTHROPIC_API_KEY=user_provided
 # Azure      #
 #============#
 
-
 # Note: these variables are DEPRECATED
 # Use the `librechat.yaml` configuration for `azureOpenAI` instead
 # You may also continue to use them if you opt out of using the `librechat.yaml` configuration
@@ -110,6 +111,26 @@ ANTHROPIC_API_KEY=user_provided
 BINGAI_TOKEN=user_provided
 # BINGAI_HOST=https://cn.bing.com
 
+#=================#
+#   AWS Bedrock   #
+#=================#
+
+# BEDROCK_AWS_DEFAULT_REGION=us-east-1 # A default region must be provided
+# BEDROCK_AWS_ACCESS_KEY_ID=someAccessKey
+# BEDROCK_AWS_SECRET_ACCESS_KEY=someSecretAccessKey
+
+# Note: This example list is not meant to be exhaustive. If omitted, all known, supported model IDs will be included for you.
+# BEDROCK_AWS_MODELS=anthropic.claude-3-5-sonnet-20240620-v1:0,meta.llama3-1-8b-instruct-v1:0
+
+# See all Bedrock model IDs here: https://docs.aws.amazon.com/bedrock/latest/userguide/model-ids.html#model-ids-arns
+
+# Notes on specific models:
+# The following models are not support due to not supporting streaming:
+# ai21.j2-mid-v1
+
+# The following models are not support due to not supporting conversation history:
+# ai21.j2-ultra-v1, cohere.command-text-v14, cohere.command-light-text-v14
+
 #============#
 # Google     #
 #============#
@@ -117,7 +138,7 @@ BINGAI_TOKEN=user_provided
 GOOGLE_KEY=user_provided
 # GOOGLE_REVERSE_PROXY=
 
-# Gemini API
+# Gemini API (AI Studio)
 # GOOGLE_MODELS=gemini-1.5-flash-latest,gemini-1.0-pro,gemini-1.0-pro-001,gemini-1.0-pro-latest,gemini-1.0-pro-vision-latest,gemini-1.5-pro-latest,gemini-pro,gemini-pro-vision
 
 # Vertex AI
@@ -125,26 +146,30 @@ GOOGLE_KEY=user_provided
 
 # GOOGLE_TITLE_MODEL=gemini-pro
 
-# Google Gemini Safety Settings
-# NOTE (Vertex AI): You do not have access to the BLOCK_NONE setting by default.
-# To use this restricted HarmBlockThreshold setting, you will need to either:
+# Google Safety Settings
+# NOTE: These settings apply to both Vertex AI and Gemini API (AI Studio)
 #
-# (a) Get access through an allowlist via your Google account team
-# (b) Switch your account type to monthly invoiced billing following this instruction:
-#     https://cloud.google.com/billing/docs/how-to/invoiced-billing
+# For Vertex AI:
+# To use the BLOCK_NONE setting, you need either:
+# (a) Access through an allowlist via your Google account team, or
+# (b) Switch to monthly invoiced billing: https://cloud.google.com/billing/docs/how-to/invoiced-billing
+#
+# For Gemini API (AI Studio):
+# BLOCK_NONE is available by default, no special account requirements.
+#
+# Available options: BLOCK_NONE, BLOCK_ONLY_HIGH, BLOCK_MEDIUM_AND_ABOVE, BLOCK_LOW_AND_ABOVE
 #
 # GOOGLE_SAFETY_SEXUALLY_EXPLICIT=BLOCK_ONLY_HIGH
 # GOOGLE_SAFETY_HATE_SPEECH=BLOCK_ONLY_HIGH
 # GOOGLE_SAFETY_HARASSMENT=BLOCK_ONLY_HIGH
 # GOOGLE_SAFETY_DANGEROUS_CONTENT=BLOCK_ONLY_HIGH
 
-
 #============#
 # OpenAI     #
 #============#
 
 OPENAI_API_KEY=user_provided
-# OPENAI_MODELS=gpt-4o,gpt-3.5-turbo-0125,gpt-3.5-turbo-0301,gpt-3.5-turbo,gpt-4,gpt-4-0613,gpt-4-vision-preview,gpt-3.5-turbo-0613,gpt-3.5-turbo-16k-0613,gpt-4-0125-preview,gpt-4-turbo-preview,gpt-4-1106-preview,gpt-3.5-turbo-1106,gpt-3.5-turbo-instruct,gpt-3.5-turbo-instruct-0914,gpt-3.5-turbo-16k
+# OPENAI_MODELS=gpt-4o,chatgpt-4o-latest,gpt-4o-mini,gpt-3.5-turbo-0125,gpt-3.5-turbo-0301,gpt-3.5-turbo,gpt-4,gpt-4-0613,gpt-4-vision-preview,gpt-3.5-turbo-0613,gpt-3.5-turbo-16k-0613,gpt-4-0125-preview,gpt-4-turbo-preview,gpt-4-1106-preview,gpt-3.5-turbo-1106,gpt-3.5-turbo-instruct,gpt-3.5-turbo-instruct-0914,gpt-3.5-turbo-16k
 
 DEBUG_OPENAI=false
 
@@ -166,7 +191,7 @@ DEBUG_OPENAI=false
 
 ASSISTANTS_API_KEY=user_provided
 # ASSISTANTS_BASE_URL=
-# ASSISTANTS_MODELS=gpt-4o,gpt-3.5-turbo-0125,gpt-3.5-turbo-16k-0613,gpt-3.5-turbo-16k,gpt-3.5-turbo,gpt-4,gpt-4-0314,gpt-4-32k-0314,gpt-4-0613,gpt-3.5-turbo-0613,gpt-3.5-turbo-1106,gpt-4-0125-preview,gpt-4-turbo-preview,gpt-4-1106-preview
+# ASSISTANTS_MODELS=gpt-4o,gpt-4o-mini,gpt-3.5-turbo-0125,gpt-3.5-turbo-16k-0613,gpt-3.5-turbo-16k,gpt-3.5-turbo,gpt-4,gpt-4-0314,gpt-4-32k-0314,gpt-4-0613,gpt-3.5-turbo-0613,gpt-3.5-turbo-1106,gpt-4-0125-preview,gpt-4-turbo-preview,gpt-4-1106-preview
 
 #==========================#
 #   Azure Assistants API   #
@@ -188,7 +213,7 @@ ASSISTANTS_API_KEY=user_provided
 # Plugins    #
 #============#
 
-# PLUGIN_MODELS=gpt-4o,gpt-4,gpt-4-turbo-preview,gpt-4-0125-preview,gpt-4-1106-preview,gpt-4-0613,gpt-3.5-turbo,gpt-3.5-turbo-0125,gpt-3.5-turbo-1106,gpt-3.5-turbo-0613
+# PLUGIN_MODELS=gpt-4o,gpt-4o-mini,gpt-4,gpt-4-turbo-preview,gpt-4-0125-preview,gpt-4-1106-preview,gpt-4-0613,gpt-3.5-turbo,gpt-3.5-turbo-0125,gpt-3.5-turbo-1106,gpt-3.5-turbo-0613
 
 DEBUG_PLUGINS=true
 
@@ -261,7 +286,6 @@ MEILI_NO_ANALYTICS=true
 MEILI_HOST=http://0.0.0.0:7700
 MEILI_MASTER_KEY=DrhYf7zENyR6AlUCKmnz0eYASOQdl6zxH7s7MKFSfFCt
 
-
 #==================================================#
 #          Speech to Text & Text to Speech         #
 #==================================================#
@@ -269,6 +293,16 @@ MEILI_MASTER_KEY=DrhYf7zENyR6AlUCKmnz0eYASOQdl6zxH7s7MKFSfFCt
 STT_API_KEY=
 TTS_API_KEY=
 
+#==================================================#
+#                        RAG                       #
+#==================================================#
+# More info: https://www.librechat.ai/docs/configuration/rag_api
+
+# RAG_OPENAI_BASEURL=
+# RAG_OPENAI_API_KEY=
+# EMBEDDINGS_PROVIDER=openai
+# EMBEDDINGS_MODEL=text-embedding-3-small
+
 #===================================================#
 #                    User System                    #
 #===================================================#
@@ -374,6 +408,8 @@ LDAP_BIND_CREDENTIALS=
 LDAP_USER_SEARCH_BASE=
 LDAP_SEARCH_FILTER=mail={{username}}
 LDAP_CA_CERT_PATH=
+# LDAP_TLS_REJECT_UNAUTHORIZED=
+# LDAP_LOGIN_USES_USERNAME=true
 # LDAP_ID=
 # LDAP_USERNAME=
 # LDAP_FULL_NAME=
@@ -411,6 +447,18 @@ FIREBASE_APP_ID=
 ALLOW_SHARED_LINKS=true
 ALLOW_SHARED_LINKS_PUBLIC=true
 
+#==============================#
+# Static File Cache Control    #
+#==============================#
+
+# Leave commented out to use defaults: 1 day (86400 seconds) for s-maxage and 2 days (172800 seconds) for max-age
+# NODE_ENV must be set to production for these to take effect
+# STATIC_CACHE_MAX_AGE=172800
+# STATIC_CACHE_S_MAX_AGE=86400
+
+# If you have another service in front of your LibreChat doing compression, disable express based compression here
+# DISABLE_COMPRESSION=true
+
 #===================================================#
 #                        UI                         #
 #===================================================#

.eslintrc.js

@@ -12,6 +12,7 @@ module.exports = {
     'plugin:react-hooks/recommended',
     'plugin:jest/recommended',
     'prettier',
+    'plugin:jsx-a11y/recommended',
   ],
   ignorePatterns: [
     'client/dist/**/*',
@@ -32,7 +33,7 @@ module.exports = {
       jsx: true,
     },
   },
-  plugins: ['react', 'react-hooks', '@typescript-eslint', 'import'],
+  plugins: ['react', 'react-hooks', '@typescript-eslint', 'import', 'jsx-a11y'],
   rules: {
     'react/react-in-jsx-scope': 'off',
     '@typescript-eslint/ban-ts-comment': ['error', { 'ts-ignore': 'allow' }],
@@ -65,6 +66,7 @@ module.exports = {
     'no-restricted-syntax': 'off',
     'react/prop-types': ['off'],
     'react/display-name': ['off'],
+    'no-nested-ternary': 'error',
     'no-unused-vars': ['error', { varsIgnorePattern: '^_' }],
     quotes: ['error', 'single'],
   },
@@ -118,6 +120,8 @@ module.exports = {
       ],
       rules: {
         '@typescript-eslint/no-explicit-any': 'error',
+        '@typescript-eslint/no-unnecessary-condition': 'warn',
+        '@typescript-eslint/strict-boolean-expressions': 'warn',
       },
     },
     {

.github/workflows/a11y.yml

@@ -0,0 +1,26 @@
+name: Lint for accessibility issues
+
+on:
+  pull_request:
+    paths:
+      - 'client/src/**'
+  workflow_dispatch:
+    inputs:
+      run_workflow:
+        description: 'Set to true to run this workflow'
+        required: true
+        default: 'false'
+
+jobs:
+  axe-linter:
+    runs-on: ubuntu-latest
+    if: >
+      (github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name == 'danny-avila/LibreChat') ||
+      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_workflow == 'true')
+    
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dequelabs/axe-linter-action@v1
+        with:
+          api_key: ${{ secrets.AXE_LINTER_API_KEY }}
+          github_token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/deploy-dev.yml

@@ -0,0 +1,41 @@
+name: Update Test Server
+
+on:
+  workflow_run:
+    workflows: ["Docker Dev Images Build"]
+    types:
+      - completed
+  workflow_dispatch:
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    if: |
+      github.repository == 'danny-avila/LibreChat' &&
+      (github.event_name == 'workflow_dispatch' || github.event.workflow_run.conclusion == 'success')
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Install SSH Key
+      uses: shimataro/ssh-key-action@v2
+      with:
+        key: ${{ secrets.DO_SSH_PRIVATE_KEY }}
+        known_hosts: ${{ secrets.DO_KNOWN_HOSTS }}
+
+    - name: Run update script on DigitalOcean Droplet
+      env:
+        DO_HOST: ${{ secrets.DO_HOST }}
+        DO_USER: ${{ secrets.DO_USER }}
+      run: |
+        ssh -o StrictHostKeyChecking=no ${DO_USER}@${DO_HOST} << EOF
+        sudo -i -u danny bash << EEOF
+        cd ~/LibreChat && \
+        git fetch origin main && \
+        npm run update:deployed && \
+        git checkout do-deploy && \
+        git rebase main && \
+        npm run start:deployed && \
+        echo "Update completed. Application should be running now."
+        EEOF
+        EOF

.github/workflows/frontend-review.yml

@@ -53,4 +53,4 @@ jobs:
 
       - name: Run unit tests
         run: npm run test:ci --verbose
-        working-directory: client
+        working-directory: client

.github/workflows/helmcharts.yml

@@ -0,0 +1,35 @@
+name: Build Helm Charts on Tag
+
+# The workflow is triggered when a tag is pushed
+on:
+  push:
+    tags:
+      - "*"
+
+jobs:
+  release:
+    permissions:
+      contents: write
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Configure Git
+        run: |
+          git config user.name "$GITHUB_ACTOR"
+          git config user.email "$GITHUB_ACTOR@users.noreply.github.com"
+
+      - name: Install Helm
+        uses: azure/setup-helm@v4
+        env:
+          GITHUB_TOKEN: "${{ secrets.GITHUB_TOKEN }}"       
+
+      - name: Run chart-releaser
+        uses: helm/chart-releaser-action@v1.6.0
+        with:
+          charts_dir: helmchart
+        env:
+          CR_TOKEN: "${{ secrets.GITHUB_TOKEN }}"

.vscode/launch.json

@@ -0,0 +1,16 @@
+{
+  "version": "0.2.0",
+  "configurations": [
+    {
+      "type": "node",
+      "request": "launch",
+      "name": "Launch LibreChat (debug)",
+      "skipFiles": ["<node_internals>/**"],
+      "program": "${workspaceFolder}/api/server/index.js",
+      "env": {
+        "NODE_ENV": "production"
+      },
+      "console": "integratedTerminal"
+    }
+  ]
+}

Dockerfile

@@ -1,4 +1,4 @@
-# v0.7.3
+# v0.7.5-rc2
 
 # Base node image
 FROM node:20-alpine AS node

Dockerfile.multi

@@ -1,43 +1,44 @@
-# v0.7.3
+# Dockerfile.multi
+# v0.7.5-rc2
 
-# Build API, Client and Data Provider
+# Base for all builds
 FROM node:20-alpine AS base
+WORKDIR /app
+RUN apk --no-cache add curl
+RUN npm config set fetch-retry-maxtimeout 600000 && \
+    npm config set fetch-retries 5 && \
+    npm config set fetch-retry-mintimeout 15000
+COPY package*.json ./
+COPY packages/data-provider/package*.json ./packages/data-provider/
+COPY client/package*.json ./client/
+COPY api/package*.json ./api/
+RUN npm ci
 
 # Build data-provider
 FROM base AS data-provider-build
 WORKDIR /app/packages/data-provider
-COPY ./packages/data-provider ./
-RUN npm install; npm cache clean --force
+COPY packages/data-provider ./
 RUN npm run build
 RUN npm prune --production
 
-# React client build
+# Client build
 FROM base AS client-build
 WORKDIR /app/client
-COPY ./client/package*.json ./
-# Copy data-provider to client's node_modules
-COPY --from=data-provider-build /app/packages/data-provider/ /app/client/node_modules/librechat-data-provider/
-RUN npm install; npm cache clean --force
-COPY ./client/ ./
+COPY client ./
+COPY --from=data-provider-build /app/packages/data-provider/dist /app/packages/data-provider/dist
 ENV NODE_OPTIONS="--max-old-space-size=2048"
 RUN npm run build
+RUN npm prune --production
 
-# Node API setup
+# API setup (including client dist)
 FROM base AS api-build
+WORKDIR /app
+COPY api ./api
+COPY config ./config
+COPY --from=data-provider-build /app/packages/data-provider/dist ./packages/data-provider/dist
+COPY --from=client-build /app/client/dist ./client/dist
 WORKDIR /app/api
-COPY api/package*.json ./
-COPY api/ ./
-# Copy helper scripts
-COPY config/ ./
-# Copy data-provider to API's node_modules
-COPY --from=data-provider-build /app/packages/data-provider/ /app/api/node_modules/librechat-data-provider/
-RUN npm install --include prod; npm cache clean --force
-COPY --from=client-build /app/client/dist /app/client/dist
+RUN npm prune --production
 EXPOSE 3080
 ENV HOST=0.0.0.0
 CMD ["node", "server/index.js"]
-
-# Nginx setup
-FROM nginx:1.21.1-alpine AS prod-stage
-COPY ./client/nginx.conf /etc/nginx/conf.d/default.conf
-CMD ["nginx", "-g", "daemon off;"]

README.md

@@ -42,15 +42,17 @@
 
 - 🖥️ UI matching ChatGPT, including Dark mode, Streaming, and latest updates
 - 🤖 AI model selection:
-  - OpenAI, Azure OpenAI, BingAI, ChatGPT, Google Vertex AI, Anthropic (Claude), Plugins, Assistants API (including Azure Assistants)
+  - Anthropic (Claude), AWS Bedrock, OpenAI, Azure OpenAI, BingAI, ChatGPT, Google Vertex AI, Plugins, Assistants API (including Azure Assistants)
 - ✅ Compatible across both **[Remote & Local AI services](https://www.librechat.ai/docs/configuration/librechat_yaml/ai_endpoints):**
   - groq, Ollama, Cohere, Mistral AI, Apple MLX, koboldcpp, OpenRouter, together.ai, Perplexity, ShuttleAI, and more
+- 🪄 Generative UI with **[Code Artifacts](https://youtu.be/GfTj7O4gmd0?si=WJbdnemZpJzBrJo3)**
+   - Create React, HTML code, and Mermaid diagrams right in chat
 - 💾 Create, Save, & Share Custom Presets
 - 🔀 Switch between AI Endpoints and Presets, mid-chat
 - 🔄 Edit, Resubmit, and Continue Messages with Conversation branching
 - 🌿 Fork Messages & Conversations for Advanced Context control
 - 💬 Multimodal Chat:
-    - Upload and analyze images with Claude 3, GPT-4 (including `gpt-4o`), and Gemini Vision 📸
+    - Upload and analyze images with Claude 3, GPT-4 (including `gpt-4o` and `gpt-4o-mini`), and Gemini Vision 📸
     - Chat with Files using Custom Endpoints, OpenAI, Azure, Anthropic, & Google. 🗃️
     - Advanced Agents with Files, Code Interpreter, Tools, and API Actions 🔦
       - Available through the [OpenAI Assistants API](https://platform.openai.com/docs/assistants/overview) 🌤️
@@ -81,7 +83,7 @@ LibreChat brings together the future of assistant AIs with the revolutionary tec
 
 With LibreChat, you no longer need to opt for ChatGPT Plus and can instead use free or pay-per-call APIs. We welcome contributions, cloning, and forking to enhance the capabilities of this advanced chatbot platform.
 
-[![Watch the video](https://img.youtube.com/vi/bSVHEbVPNl4/maxresdefault.jpg)](https://www.youtube.com/watch?v=bSVHEbVPNl4)
+[![Watch the video](https://raw.githubusercontent.com/LibreChat-AI/librechat.ai/main/public/images/changelog/v0.7.4.png)](https://www.youtube.com/watch?v=cvosUxogdpI)
 Click on the thumbnail to open the video☝️
 
 ---

api/app/clients/AnthropicClient.js

@@ -2,20 +2,24 @@ const Anthropic = require('@anthropic-ai/sdk');
 const { HttpsProxyAgent } = require('https-proxy-agent');
 const { encoding_for_model: encodingForModel, get_encoding: getEncoding } = require('tiktoken');
 const {
-  getResponseSender,
+  Constants,
   EModelEndpoint,
+  anthropicSettings,
+  getResponseSender,
   validateVisionModel,
 } = require('librechat-data-provider');
 const { encodeAndFormat } = require('~/server/services/Files/images/encode');
 const {
   truncateText,
   formatMessage,
+  addCacheControl,
   titleFunctionPrompt,
   parseParamFromPrompt,
   createContextHandlers,
 } = require('./prompts');
-const spendTokens = require('~/models/spendTokens');
-const { getModelMaxTokens } = require('~/utils');
+const { spendTokens, spendStructuredTokens } = require('~/models/spendTokens');
+const { getModelMaxTokens, matchModelName } = require('~/utils');
+const { sleep } = require('~/server/utils');
 const BaseClient = require('./BaseClient');
 const { logger } = require('~/config');
 
@@ -29,6 +33,9 @@ function delayBeforeRetry(attempts, baseDelay = 1000) {
   return new Promise((resolve) => setTimeout(resolve, baseDelay * attempts));
 }
 
+const tokenEventTypes = new Set(['message_start', 'message_delta']);
+const { legacy } = anthropicSettings;
+
 class AnthropicClient extends BaseClient {
   constructor(apiKey, options = {}) {
     super(apiKey, options);
@@ -39,6 +46,24 @@ class AnthropicClient extends BaseClient {
       ? options.contextStrategy.toLowerCase()
       : 'discard';
     this.setOptions(options);
+    /** @type {string | undefined} */
+    this.systemMessage;
+    /** @type {AnthropicMessageStartEvent| undefined} */
+    this.message_start;
+    /** @type {AnthropicMessageDeltaEvent| undefined} */
+    this.message_delta;
+    /** Whether the model is part of the Claude 3 Family
+     * @type {boolean} */
+    this.isClaude3;
+    /** Whether to use Messages API or Completions API
+     * @type {boolean} */
+    this.useMessages;
+    /** Whether or not the model is limited to the legacy amount of output tokens
+     * @type {boolean} */
+    this.isLegacyOutput;
+    /** Whether or not the model supports Prompt Caching
+     * @type {boolean} */
+    this.supportsCacheControl;
   }
 
   setOptions(options) {
@@ -58,18 +83,28 @@ class AnthropicClient extends BaseClient {
       this.options = options;
     }
 
-    const modelOptions = this.options.modelOptions || {};
-    this.modelOptions = {
-      ...modelOptions,
-      // set some good defaults (check for undefined in some cases because they may be 0)
-      model: modelOptions.model || 'claude-1',
-      temperature: typeof modelOptions.temperature === 'undefined' ? 1 : modelOptions.temperature, // 0 - 1, 1 is default
-      topP: typeof modelOptions.topP === 'undefined' ? 0.7 : modelOptions.topP, // 0 - 1, default: 0.7
-      topK: typeof modelOptions.topK === 'undefined' ? 40 : modelOptions.topK, // 1-40, default: 40
-      stop: modelOptions.stop, // no stop method for now
-    };
+    this.modelOptions = Object.assign(
+      {
+        model: anthropicSettings.model.default,
+      },
+      this.modelOptions,
+      this.options.modelOptions,
+    );
+
+    const modelMatch = matchModelName(this.modelOptions.model, EModelEndpoint.anthropic);
+    this.isClaude3 = modelMatch.startsWith('claude-3');
+    this.isLegacyOutput = !modelMatch.startsWith('claude-3-5-sonnet');
+    this.supportsCacheControl =
+      this.options.promptCache && this.checkPromptCacheSupport(modelMatch);
+
+    if (
+      this.isLegacyOutput &&
+      this.modelOptions.maxOutputTokens &&
+      this.modelOptions.maxOutputTokens > legacy.maxOutputTokens.default
+    ) {
+      this.modelOptions.maxOutputTokens = legacy.maxOutputTokens.default;
+    }
 
-    this.isClaude3 = this.modelOptions.model.includes('claude-3');
     this.useMessages = this.isClaude3 || !!this.options.attachments;
 
     this.defaultVisionModel = this.options.visionModel ?? 'claude-3-sonnet-20240229';
@@ -119,10 +154,11 @@ class AnthropicClient extends BaseClient {
 
   /**
    * Get the initialized Anthropic client.
+   * @param {Partial<Anthropic.ClientOptions>} requestOptions - The options for the client.
    * @returns {Anthropic} The Anthropic client instance.
    */
-  getClient() {
-    /** @type {Anthropic.default.RequestOptions} */
+  getClient(requestOptions) {
+    /** @type {Anthropic.ClientOptions} */
     const options = {
       fetch: this.fetch,
       apiKey: this.apiKey,
@@ -136,13 +172,74 @@ class AnthropicClient extends BaseClient {
       options.baseURL = this.options.reverseProxyUrl;
     }
 
+    if (
+      this.supportsCacheControl &&
+      requestOptions?.model &&
+      requestOptions.model.includes('claude-3-5-sonnet')
+    ) {
+      options.defaultHeaders = {
+        'anthropic-beta': 'max-tokens-3-5-sonnet-2024-07-15,prompt-caching-2024-07-31',
+      };
+    } else if (this.supportsCacheControl) {
+      options.defaultHeaders = {
+        'anthropic-beta': 'prompt-caching-2024-07-31',
+      };
+    }
+
     return new Anthropic(options);
   }
 
-  getTokenCountForResponse(response) {
+  /**
+   * Get stream usage as returned by this client's API response.
+   * @returns {AnthropicStreamUsage} The stream usage object.
+   */
+  getStreamUsage() {
+    const inputUsage = this.message_start?.message?.usage ?? {};
+    const outputUsage = this.message_delta?.usage ?? {};
+    return Object.assign({}, inputUsage, outputUsage);
+  }
+
+  /**
+   * Calculates the correct token count for the current message based on the token count map and API usage.
+   * Edge case: If the calculation results in a negative value, it returns the original estimate.
+   * If revisiting a conversation with a chat history entirely composed of token estimates,
+   * the cumulative token count going forward should become more accurate as the conversation progresses.
+   * @param {Object} params - The parameters for the calculation.
+   * @param {Record<string, number>} params.tokenCountMap - A map of message IDs to their token counts.
+   * @param {string} params.currentMessageId - The ID of the current message to calculate.
+   * @param {AnthropicStreamUsage} params.usage - The usage object returned by the API.
+   * @returns {number} The correct token count for the current message.
+   */
+  calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage }) {
+    const originalEstimate = tokenCountMap[currentMessageId] || 0;
+
+    if (!usage || typeof usage.input_tokens !== 'number') {
+      return originalEstimate;
+    }
+
+    tokenCountMap[currentMessageId] = 0;
+    const totalTokensFromMap = Object.values(tokenCountMap).reduce((sum, count) => {
+      const numCount = Number(count);
+      return sum + (isNaN(numCount) ? 0 : numCount);
+    }, 0);
+    const totalInputTokens =
+      (usage.input_tokens ?? 0) +
+      (usage.cache_creation_input_tokens ?? 0) +
+      (usage.cache_read_input_tokens ?? 0);
+
+    const currentMessageTokens = totalInputTokens - totalTokensFromMap;
+    return currentMessageTokens > 0 ? currentMessageTokens : originalEstimate;
+  }
+
+  /**
+   * Get Token Count for LibreChat Message
+   * @param {TMessage} responseMessage
+   * @returns {number}
+   */
+  getTokenCountForResponse(responseMessage) {
     return this.getTokenCountForMessage({
       role: 'assistant',
-      content: response.text,
+      content: responseMessage.text,
     });
   }
 
@@ -195,7 +292,38 @@ class AnthropicClient extends BaseClient {
     return files;
   }
 
-  async recordTokenUsage({ promptTokens, completionTokens, model, context = 'message' }) {
+  /**
+   * @param {object} params
+   * @param {number} params.promptTokens
+   * @param {number} params.completionTokens
+   * @param {AnthropicStreamUsage} [params.usage]
+   * @param {string} [params.model]
+   * @param {string} [params.context='message']
+   * @returns {Promise<void>}
+   */
+  async recordTokenUsage({ promptTokens, completionTokens, usage, model, context = 'message' }) {
+    if (usage != null && usage?.input_tokens != null) {
+      const input = usage.input_tokens ?? 0;
+      const write = usage.cache_creation_input_tokens ?? 0;
+      const read = usage.cache_read_input_tokens ?? 0;
+
+      await spendStructuredTokens(
+        {
+          context,
+          user: this.user,
+          conversationId: this.conversationId,
+          model: model ?? this.modelOptions.model,
+          endpointTokenConfig: this.options.endpointTokenConfig,
+        },
+        {
+          promptTokens: { input, write, read },
+          completionTokens,
+        },
+      );
+
+      return;
+    }
+
     await spendTokens(
       {
         context,
@@ -364,7 +492,10 @@ class AnthropicClient extends BaseClient {
       identityPrefix = `${identityPrefix}\nYou are ${this.options.modelLabel}`;
     }
 
-    let promptPrefix = (this.options.promptPrefix || '').trim();
+    let promptPrefix = (this.options.promptPrefix ?? '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+    }
     if (promptPrefix) {
       // If the prompt prefix doesn't end with the end token, add it.
       if (!promptPrefix.endsWith(`${this.endToken}`)) {
@@ -543,6 +674,18 @@ class AnthropicClient extends BaseClient {
       : await client.completions.create(options);
   }
 
+  /**
+   * @param {string} modelName
+   * @returns {boolean}
+   */
+  checkPromptCacheSupport(modelName) {
+    const modelMatch = matchModelName(modelName, EModelEndpoint.anthropic);
+    if (modelMatch === 'claude-3-5-sonnet' || modelMatch === 'claude-3-haiku') {
+      return true;
+    }
+    return false;
+  }
+
   async sendCompletion(payload, { onProgress, abortController }) {
     if (!abortController) {
       abortController = new AbortController();
@@ -556,8 +699,6 @@ class AnthropicClient extends BaseClient {
     }
 
     logger.debug('modelOptions', { modelOptions });
-
-    const client = this.getClient();
     const metadata = {
       user_id: this.user,
     };
@@ -585,16 +726,28 @@ class AnthropicClient extends BaseClient {
 
     if (this.useMessages) {
       requestOptions.messages = payload;
-      requestOptions.max_tokens = maxOutputTokens || 1500;
+      requestOptions.max_tokens = maxOutputTokens || legacy.maxOutputTokens.default;
     } else {
       requestOptions.prompt = payload;
       requestOptions.max_tokens_to_sample = maxOutputTokens || 1500;
     }
 
-    if (this.systemMessage) {
+    if (this.systemMessage && this.supportsCacheControl === true) {
+      requestOptions.system = [
+        {
+          type: 'text',
+          text: this.systemMessage,
+          cache_control: { type: 'ephemeral' },
+        },
+      ];
+    } else if (this.systemMessage) {
       requestOptions.system = this.systemMessage;
     }
 
+    if (this.supportsCacheControl === true && this.useMessages) {
+      requestOptions.messages = addCacheControl(requestOptions.messages);
+    }
+
     logger.debug('[AnthropicClient]', { ...requestOptions });
 
     const handleChunk = (currentChunk) => {
@@ -605,12 +758,14 @@ class AnthropicClient extends BaseClient {
     };
 
     const maxRetries = 3;
+    const streamRate = this.options.streamRate ?? Constants.DEFAULT_STREAM_RATE;
     async function processResponse() {
       let attempts = 0;
 
       while (attempts < maxRetries) {
         let response;
         try {
+          const client = this.getClient(requestOptions);
           response = await this.createResponse(client, requestOptions);
 
           signal.addEventListener('abort', () => {
@@ -622,11 +777,18 @@ class AnthropicClient extends BaseClient {
 
           for await (const completion of response) {
             // Handle each completion as before
+            const type = completion?.type ?? '';
+            if (tokenEventTypes.has(type)) {
+              logger.debug(`[AnthropicClient] ${type}`, completion);
+              this[type] = completion;
+            }
             if (completion?.delta?.text) {
               handleChunk(completion.delta.text);
             } else if (completion.completion) {
               handleChunk(completion.completion);
             }
+
+            await sleep(streamRate);
           }
 
           // Successful processing, exit loop
@@ -661,8 +823,10 @@ class AnthropicClient extends BaseClient {
   getSaveOptions() {
     return {
       maxContextTokens: this.options.maxContextTokens,
+      artifacts: this.options.artifacts,
       promptPrefix: this.options.promptPrefix,
       modelLabel: this.options.modelLabel,
+      promptCache: this.options.promptCache,
       resendFiles: this.options.resendFiles,
       iconURL: this.options.iconURL,
       greeting: this.options.greeting,
@@ -708,6 +872,8 @@ class AnthropicClient extends BaseClient {
    */
   async titleConvo({ text, responseText = '' }) {
     let title = 'New Chat';
+    this.message_delta = undefined;
+    this.message_start = undefined;
     const convo = `<initial_message>
   ${truncateText(text)}
   </initial_message>
@@ -737,7 +903,11 @@ class AnthropicClient extends BaseClient {
       };
 
       try {
-        const response = await this.createResponse(this.getClient(), requestOptions, true);
+        const response = await this.createResponse(
+          this.getClient(requestOptions),
+          requestOptions,
+          true,
+        );
         let promptTokens = response?.usage?.input_tokens;
         let completionTokens = response?.usage?.output_tokens;
         if (!promptTokens) {

api/app/clients/BaseClient.js

@@ -1,10 +1,19 @@
 const crypto = require('crypto');
 const fetch = require('node-fetch');
-const { supportsBalanceCheck, Constants } = require('librechat-data-provider');
+const {
+  supportsBalanceCheck,
+  isAgentsEndpoint,
+  paramEndpoints,
+  ErrorTypes,
+  Constants,
+  CacheKeys,
+  Time,
+} = require('librechat-data-provider');
 const { getMessages, saveMessage, updateMessage, saveConvo } = require('~/models');
 const { addSpaceIfNeeded, isEnabled } = require('~/server/utils');
 const checkBalance = require('~/models/checkBalance');
 const { getFiles } = require('~/models/File');
+const { getLogStores } = require('~/cache');
 const TextStream = require('./TextStream');
 const { logger } = require('~/config');
 
@@ -27,6 +36,12 @@ class BaseClient {
     this.userMessagePromise;
     /** @type {ClientDatabaseSavePromise} */
     this.responsePromise;
+    /** @type {string} */
+    this.user;
+    /** @type {string} */
+    this.conversationId;
+    /** @type {string} */
+    this.responseMessageId;
   }
 
   setOptions() {
@@ -53,10 +68,33 @@ class BaseClient {
     throw new Error('Subclasses attempted to call summarizeMessages without implementing it');
   }
 
-  async getTokenCountForResponse(response) {
-    logger.debug('`[BaseClient] recordTokenUsage` not implemented.', response);
+  /**
+   * @returns {string}
+   */
+  getResponseModel() {
+    if (isAgentsEndpoint(this.options.endpoint) && this.options.agent && this.options.agent.id) {
+      return this.options.agent.id;
+    }
+
+    return this.modelOptions.model;
   }
 
+  /**
+   * Abstract method to get the token count for a message. Subclasses must implement this method.
+   * @param {TMessage} responseMessage
+   * @returns {number}
+   */
+  getTokenCountForResponse(responseMessage) {
+    logger.debug('`[BaseClient] recordTokenUsage` not implemented.', responseMessage);
+  }
+
+  /**
+   * Abstract method to record token usage. Subclasses must implement this method.
+   * If a correction to the token usage is needed, the method should return an object with the corrected token counts.
+   * @param {number} promptTokens
+   * @param {number} completionTokens
+   * @returns {Promise<void>}
+   */
   async recordTokenUsage({ promptTokens, completionTokens }) {
     logger.debug('`[BaseClient] recordTokenUsage` not implemented.', {
       promptTokens,
@@ -142,6 +180,8 @@ class BaseClient {
       this.currentMessages[this.currentMessages.length - 1].messageId = head;
     }
 
+    this.responseMessageId = responseMessageId;
+
     return {
       ...opts,
       user,
@@ -190,6 +230,7 @@ class BaseClient {
         userMessage,
         conversationId,
         responseMessageId,
+        sender: this.sender,
       });
     }
 
@@ -328,7 +369,12 @@ class BaseClient {
     };
   }
 
-  async handleContextStrategy({ instructions, orderedMessages, formattedMessages }) {
+  async handleContextStrategy({
+    instructions,
+    orderedMessages,
+    formattedMessages,
+    buildTokenMap = true,
+  }) {
     let _instructions;
     let tokenCount;
 
@@ -370,9 +416,10 @@ class BaseClient {
 
     const latestMessage = orderedWithInstructions[orderedWithInstructions.length - 1];
     if (payload.length === 0 && !shouldSummarize && latestMessage) {
-      throw new Error(
-        `Prompt token count of ${latestMessage.tokenCount} exceeds max token count of ${this.maxContextTokens}.`,
-      );
+      const info = `${latestMessage.tokenCount} / ${this.maxContextTokens}`;
+      const errorMessage = `{ "type": "${ErrorTypes.INPUT_LENGTH}", "info": "${info}" }`;
+      logger.warn(`Prompt token count exceeds max token count (${info}).`);
+      throw new Error(errorMessage);
     }
 
     if (usePrevSummary) {
@@ -397,19 +444,23 @@ class BaseClient {
       maxContextTokens: this.maxContextTokens,
     });
 
-    let tokenCountMap = orderedWithInstructions.reduce((map, message, index) => {
-      const { messageId } = message;
-      if (!messageId) {
+    /** @type {Record<string, number> | undefined} */
+    let tokenCountMap;
+    if (buildTokenMap) {
+      tokenCountMap = orderedWithInstructions.reduce((map, message, index) => {
+        const { messageId } = message;
+        if (!messageId) {
+          return map;
+        }
+
+        if (shouldSummarize && index === summaryIndex && !usePrevSummary) {
+          map.summaryMessage = { ...summaryMessage, messageId, tokenCount: summaryTokenCount };
+        }
+
+        map[messageId] = orderedWithInstructions[index].tokenCount;
         return map;
-      }
-
-      if (shouldSummarize && index === summaryIndex && !usePrevSummary) {
-        map.summaryMessage = { ...summaryMessage, messageId, tokenCount: summaryTokenCount };
-      }
-
-      map[messageId] = orderedWithInstructions[index].tokenCount;
-      return map;
-    }, {});
+      }, {});
+    }
 
     const promptTokens = this.maxContextTokens - remainingContextTokens;
 
@@ -511,6 +562,7 @@ class BaseClient {
       });
     }
 
+    /** @type {string|string[]|undefined} */
     const completion = await this.sendCompletion(payload, opts);
     this.abortController.requestCompleted = true;
 
@@ -520,30 +572,128 @@ class BaseClient {
       parentMessageId: userMessage.messageId,
       isCreatedByUser: false,
       isEdited,
-      model: this.modelOptions.model,
+      model: this.getResponseModel(),
       sender: this.sender,
-      text: addSpaceIfNeeded(generation) + completion,
       promptTokens,
       iconURL: this.options.iconURL,
       endpoint: this.options.endpoint,
       ...(this.metadata ?? {}),
     };
 
+    if (typeof completion === 'string') {
+      responseMessage.text = addSpaceIfNeeded(generation) + completion;
+    } else if (Array.isArray(completion) && paramEndpoints.has(this.options.endpoint)) {
+      responseMessage.text = '';
+      responseMessage.content = completion;
+    } else if (Array.isArray(completion)) {
+      responseMessage.text = addSpaceIfNeeded(generation) + completion.join('');
+    }
+
     if (
       tokenCountMap &&
       this.recordTokenUsage &&
       this.getTokenCountForResponse &&
       this.getTokenCount
     ) {
-      responseMessage.tokenCount = this.getTokenCountForResponse(responseMessage);
-      const completionTokens = this.getTokenCount(completion);
-      await this.recordTokenUsage({ promptTokens, completionTokens });
+      let completionTokens;
+
+      /**
+       * Metadata about input/output costs for the current message. The client
+       * should provide a function to get the current stream usage metadata; if not,
+       * use the legacy token estimations.
+       * @type {StreamUsage | null} */
+      const usage = this.getStreamUsage != null ? this.getStreamUsage() : null;
+
+      if (usage != null && Number(usage.output_tokens) > 0) {
+        responseMessage.tokenCount = usage.output_tokens;
+        completionTokens = responseMessage.tokenCount;
+        await this.updateUserMessageTokenCount({ usage, tokenCountMap, userMessage, opts });
+      } else {
+        responseMessage.tokenCount = this.getTokenCountForResponse(responseMessage);
+        completionTokens = this.getTokenCount(completion);
+      }
+
+      await this.recordTokenUsage({ promptTokens, completionTokens, usage });
     }
+
+    if (this.userMessagePromise) {
+      await this.userMessagePromise;
+    }
+
     this.responsePromise = this.saveMessageToDatabase(responseMessage, saveOptions, user);
+    const messageCache = getLogStores(CacheKeys.MESSAGES);
+    messageCache.set(
+      responseMessageId,
+      {
+        text: responseMessage.text,
+        complete: true,
+      },
+      Time.FIVE_MINUTES,
+    );
     delete responseMessage.tokenCount;
     return responseMessage;
   }
 
+  /**
+   * Stream usage should only be used for user message token count re-calculation if:
+   * - The stream usage is available, with input tokens greater than 0,
+   * - the client provides a function to calculate the current token count,
+   * - files are being resent with every message (default behavior; or if `false`, with no attachments),
+   * - the `promptPrefix` (custom instructions) is not set.
+   *
+   * In these cases, the legacy token estimations would be more accurate.
+   *
+   * TODO: included system messages in the `orderedMessages` accounting, potentially as a
+   * separate message in the UI. ChatGPT does this through "hidden" system messages.
+   * @param {object} params
+   * @param {StreamUsage} params.usage
+   * @param {Record<string, number>} params.tokenCountMap
+   * @param {TMessage} params.userMessage
+   * @param {object} params.opts
+   */
+  async updateUserMessageTokenCount({ usage, tokenCountMap, userMessage, opts }) {
+    /** @type {boolean} */
+    const shouldUpdateCount =
+      this.calculateCurrentTokenCount != null &&
+      Number(usage.input_tokens) > 0 &&
+      (this.options.resendFiles ||
+        (!this.options.resendFiles && !this.options.attachments?.length)) &&
+      !this.options.promptPrefix;
+
+    if (!shouldUpdateCount) {
+      return;
+    }
+
+    const userMessageTokenCount = this.calculateCurrentTokenCount({
+      currentMessageId: userMessage.messageId,
+      tokenCountMap,
+      usage,
+    });
+
+    if (userMessageTokenCount === userMessage.tokenCount) {
+      return;
+    }
+
+    userMessage.tokenCount = userMessageTokenCount;
+    /*
+      Note: `AskController` saves the user message, so we update the count of its `userMessage` reference
+    */
+    if (typeof opts?.getReqData === 'function') {
+      opts.getReqData({
+        userMessage,
+      });
+    }
+    /*
+      Note: we update the user message to be sure it gets the calculated token count;
+      though `AskController` saves the user message, EditController does not
+    */
+    await this.userMessagePromise;
+    await this.updateMessageInDatabase({
+      messageId: userMessage.messageId,
+      tokenCount: userMessageTokenCount,
+    });
+  }
+
   async loadHistory(conversationId, parentMessageId = null) {
     logger.debug('[BaseClient] Loading history:', { conversationId, parentMessageId });
 
@@ -598,28 +748,45 @@ class BaseClient {
    * @param {string | null} user
    */
   async saveMessageToDatabase(message, endpointOptions, user = null) {
-    const savedMessage = await saveMessage({
-      ...message,
-      endpoint: this.options.endpoint,
-      unfinished: false,
-      user,
-    });
+    if (this.user && user !== this.user) {
+      throw new Error('User mismatch.');
+    }
+
+    const savedMessage = await saveMessage(
+      this.options.req,
+      {
+        ...message,
+        endpoint: this.options.endpoint,
+        unfinished: false,
+        user,
+      },
+      { context: 'api/app/clients/BaseClient.js - saveMessageToDatabase #saveMessage' },
+    );
 
     if (this.skipSaveConvo) {
       return { message: savedMessage };
     }
-    const conversation = await saveConvo(user, {
-      conversationId: message.conversationId,
-      endpoint: this.options.endpoint,
-      endpointType: this.options.endpointType,
-      ...endpointOptions,
-    });
+
+    const conversation = await saveConvo(
+      this.options.req,
+      {
+        conversationId: message.conversationId,
+        endpoint: this.options.endpoint,
+        endpointType: this.options.endpointType,
+        ...endpointOptions,
+      },
+      { context: 'api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo' },
+    );
 
     return { message: savedMessage, conversation };
   }
 
+  /**
+   * Update a message in the database.
+   * @param {Partial<TMessage>} message
+   */
   async updateMessageInDatabase(message) {
-    await updateMessage(message);
+    await updateMessage(this.options.req, message);
   }
 
   /**
@@ -741,8 +908,12 @@ class BaseClient {
 
           processValue(nestedValue);
         }
-      } else {
+      } else if (typeof value === 'string') {
         numTokens += this.getTokenCount(value);
+      } else if (typeof value === 'number') {
+        numTokens += this.getTokenCount(value.toString());
+      } else if (typeof value === 'boolean') {
+        numTokens += this.getTokenCount(value.toString());
       }
     };
 

api/app/clients/GoogleClient.js

@@ -13,10 +13,12 @@ const {
   endpointSettings,
   EModelEndpoint,
   VisionModes,
+  Constants,
   AuthKeys,
 } = require('librechat-data-provider');
 const { encodeAndFormat } = require('~/server/services/Files/images');
 const { getModelMaxTokens } = require('~/utils');
+const { sleep } = require('~/server/utils');
 const { logger } = require('~/config');
 const {
   formatMessage,
@@ -118,19 +120,7 @@ class GoogleClient extends BaseClient {
       .filter((ex) => ex)
       .filter((obj) => obj.input.content !== '' && obj.output.content !== '');
 
-    const modelOptions = this.options.modelOptions || {};
-    this.modelOptions = {
-      ...modelOptions,
-      // set some good defaults (check for undefined in some cases because they may be 0)
-      model: modelOptions.model || settings.model.default,
-      temperature:
-        typeof modelOptions.temperature === 'undefined'
-          ? settings.temperature.default
-          : modelOptions.temperature,
-      topP: typeof modelOptions.topP === 'undefined' ? settings.topP.default : modelOptions.topP,
-      topK: typeof modelOptions.topK === 'undefined' ? settings.topK.default : modelOptions.topK,
-      // stop: modelOptions.stop // no stop method for now
-    };
+    this.modelOptions = this.options.modelOptions || {};
 
     this.options.attachments?.then((attachments) => this.checkVisionRequest(attachments));
 
@@ -400,8 +390,13 @@ class GoogleClient extends BaseClient {
       parameters: this.modelOptions,
     };
 
-    if (this.options.promptPrefix) {
-      payload.instances[0].context = this.options.promptPrefix;
+    let promptPrefix = (this.options.promptPrefix ?? '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+    }
+
+    if (promptPrefix) {
+      payload.instances[0].context = promptPrefix;
     }
 
     if (this.options.examples.length > 0) {
@@ -455,7 +450,10 @@ class GoogleClient extends BaseClient {
       identityPrefix = `${identityPrefix}\nYou are ${this.options.modelLabel}`;
     }
 
-    let promptPrefix = (this.options.promptPrefix || '').trim();
+    let promptPrefix = (this.options.promptPrefix ?? '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+    }
     if (promptPrefix) {
       // If the prompt prefix doesn't end with the end token, add it.
       if (!promptPrefix.endsWith(`${this.endToken}`)) {
@@ -620,8 +618,9 @@ class GoogleClient extends BaseClient {
   }
 
   async getCompletion(_payload, options = {}) {
-    const { onProgress, abortController } = options;
     const { parameters, instances } = _payload;
+    const { onProgress, abortController } = options;
+    const streamRate = this.options.streamRate ?? Constants.DEFAULT_STREAM_RATE;
     const { messages: _messages, context, examples: _examples } = instances?.[0] ?? {};
 
     let examples;
@@ -674,24 +673,27 @@ class GoogleClient extends BaseClient {
 
     const modelName = clientOptions.modelName ?? clientOptions.model ?? '';
     if (modelName?.includes('1.5') && !this.project_id) {
-      /** @type {GenerativeModel} */
       const client = model;
       const requestOptions = {
         contents: _payload,
       };
 
+      let promptPrefix = (this.options.promptPrefix ?? '').trim();
+      if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+        promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+      }
+
       if (this.options?.promptPrefix?.length) {
         requestOptions.systemInstruction = {
           parts: [
             {
-              text: this.options.promptPrefix,
+              text: promptPrefix,
             },
           ],
         };
       }
 
-      const safetySettings = _payload.safetySettings;
-      requestOptions.safetySettings = safetySettings;
+      requestOptions.safetySettings = _payload.safetySettings;
 
       const delay = modelName.includes('flash') ? 8 : 14;
       const result = await client.generateContentStream(requestOptions);
@@ -701,21 +703,28 @@ class GoogleClient extends BaseClient {
           delay,
         });
         reply += chunkText;
+        await sleep(streamRate);
       }
       return reply;
     }
 
-    const safetySettings = _payload.safetySettings;
     const stream = await model.stream(messages, {
       signal: abortController.signal,
       timeout: 7000,
-      safetySettings: safetySettings,
+      safetySettings: _payload.safetySettings,
     });
 
-    let delay = this.isGenerativeModel ? 12 : 8;
-    if (modelName.includes('flash')) {
-      delay = 5;
+    let delay = this.options.streamRate || 8;
+
+    if (!this.options.streamRate) {
+      if (this.isGenerativeModel) {
+        delay = 12;
+      }
+      if (modelName.includes('flash')) {
+        delay = 5;
+      }
     }
+
     for await (const chunk of stream) {
       const chunkText = chunk?.content ?? chunk;
       await this.generateTextStream(chunkText, onProgress, {
@@ -771,11 +780,16 @@ class GoogleClient extends BaseClient {
         contents: _payload,
       };
 
+      let promptPrefix = (this.options.promptPrefix ?? '').trim();
+      if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+        promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+      }
+
       if (this.options?.promptPrefix?.length) {
         requestOptions.systemInstruction = {
           parts: [
             {
-              text: this.options.promptPrefix,
+              text: promptPrefix,
             },
           ],
         };
@@ -800,7 +814,7 @@ class GoogleClient extends BaseClient {
       });
 
       reply = titleResponse.content;
-
+      // TODO: RECORD TOKEN USAGE
       return reply;
     }
   }
@@ -846,6 +860,7 @@ class GoogleClient extends BaseClient {
 
   getSaveOptions() {
     return {
+      artifacts: this.options.artifacts,
       promptPrefix: this.options.promptPrefix,
       modelLabel: this.options.modelLabel,
       iconURL: this.options.iconURL,
@@ -860,38 +875,36 @@ class GoogleClient extends BaseClient {
   }
 
   async sendCompletion(payload, opts = {}) {
-    const modelName = payload.parameters?.model;
-
-    if (modelName && modelName.toLowerCase().includes('gemini')) {
-      const safetySettings = [
-        {
-          category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
-          threshold:
-            process.env.GOOGLE_SAFETY_SEXUALLY_EXPLICIT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
-        },
-        {
-          category: 'HARM_CATEGORY_HATE_SPEECH',
-          threshold: process.env.GOOGLE_SAFETY_HATE_SPEECH || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
-        },
-        {
-          category: 'HARM_CATEGORY_HARASSMENT',
-          threshold: process.env.GOOGLE_SAFETY_HARASSMENT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
-        },
-        {
-          category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
-          threshold:
-            process.env.GOOGLE_SAFETY_DANGEROUS_CONTENT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
-        },
-      ];
-
-      payload.safetySettings = safetySettings;
-    }
+    payload.safetySettings = this.getSafetySettings();
 
     let reply = '';
     reply = await this.getCompletion(payload, opts);
     return reply.trim();
   }
 
+  getSafetySettings() {
+    return [
+      {
+        category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
+        threshold:
+          process.env.GOOGLE_SAFETY_SEXUALLY_EXPLICIT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
+      },
+      {
+        category: 'HARM_CATEGORY_HATE_SPEECH',
+        threshold: process.env.GOOGLE_SAFETY_HATE_SPEECH || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
+      },
+      {
+        category: 'HARM_CATEGORY_HARASSMENT',
+        threshold: process.env.GOOGLE_SAFETY_HARASSMENT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
+      },
+      {
+        category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
+        threshold:
+          process.env.GOOGLE_SAFETY_DANGEROUS_CONTENT || 'HARM_BLOCK_THRESHOLD_UNSPECIFIED',
+      },
+    ];
+  }
+
   /* TO-DO: Handle tokens with Google tokenization NOTE: these are required */
   static getTokenizer(encoding, isModelName = false, extendSpecialTokens = {}) {
     if (tokenizersCache[encoding]) {

api/app/clients/OllamaClient.js

@@ -1,7 +1,9 @@
 const { z } = require('zod');
 const axios = require('axios');
 const { Ollama } = require('ollama');
+const { Constants } = require('librechat-data-provider');
 const { deriveBaseURL } = require('~/utils');
+const { sleep } = require('~/server/utils');
 const { logger } = require('~/config');
 
 const ollamaPayloadSchema = z.object({
@@ -40,6 +42,7 @@ const getValidBase64 = (imageUrl) => {
 class OllamaClient {
   constructor(options = {}) {
     const host = deriveBaseURL(options.baseURL ?? 'http://localhost:11434');
+    this.streamRate = options.streamRate ?? Constants.DEFAULT_STREAM_RATE;
     /** @type {Ollama} */
     this.client = new Ollama({ host });
   }
@@ -136,6 +139,8 @@ class OllamaClient {
           stream.controller.abort();
           break;
         }
+
+        await sleep(this.streamRate);
       }
     }
     // TODO: regular completion

api/app/clients/OpenAIClient.js

@@ -6,6 +6,7 @@ const {
   ImageDetail,
   EModelEndpoint,
   resolveHeaders,
+  openAISettings,
   ImageDetailCost,
   CohereConstants,
   getResponseSender,
@@ -27,9 +28,9 @@ const {
   createContextHandlers,
 } = require('./prompts');
 const { encodeAndFormat } = require('~/server/services/Files/images/encode');
+const { spendTokens } = require('~/models/spendTokens');
 const { isEnabled, sleep } = require('~/server/utils');
 const { handleOpenAIErrors } = require('./tools/util');
-const spendTokens = require('~/models/spendTokens');
 const { createLLM, RunManager } = require('./llm');
 const ChatGPTClient = require('./ChatGPTClient');
 const { summaryBuffer } = require('./memory');
@@ -85,26 +86,13 @@ class OpenAIClient extends BaseClient {
       this.apiKey = this.options.openaiApiKey;
     }
 
-    const modelOptions = this.options.modelOptions || {};
-
-    if (!this.modelOptions) {
-      this.modelOptions = {
-        ...modelOptions,
-        model: modelOptions.model || 'gpt-3.5-turbo',
-        temperature:
-          typeof modelOptions.temperature === 'undefined' ? 0.8 : modelOptions.temperature,
-        top_p: typeof modelOptions.top_p === 'undefined' ? 1 : modelOptions.top_p,
-        presence_penalty:
-          typeof modelOptions.presence_penalty === 'undefined' ? 1 : modelOptions.presence_penalty,
-        stop: modelOptions.stop,
-      };
-    } else {
-      // Update the modelOptions if it already exists
-      this.modelOptions = {
-        ...this.modelOptions,
-        ...modelOptions,
-      };
-    }
+    this.modelOptions = Object.assign(
+      {
+        model: openAISettings.model.default,
+      },
+      this.modelOptions,
+      this.options.modelOptions,
+    );
 
     this.defaultVisionModel = this.options.visionModel ?? 'gpt-4-vision-preview';
     if (typeof this.options.attachments?.then === 'function') {
@@ -413,6 +401,7 @@ class OpenAIClient extends BaseClient {
 
   getSaveOptions() {
     return {
+      artifacts: this.options.artifacts,
       maxContextTokens: this.options.maxContextTokens,
       chatGptLabel: this.options.chatGptLabel,
       promptPrefix: this.options.promptPrefix,
@@ -475,6 +464,9 @@ class OpenAIClient extends BaseClient {
     let promptTokens;
 
     promptPrefix = (promptPrefix || this.options.promptPrefix || '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      promptPrefix = `${promptPrefix ?? ''}\n${this.options.artifactsPrompt}`.trim();
+    }
 
     if (this.options.attachments) {
       const attachments = await this.options.attachments;
@@ -827,7 +819,7 @@ class OpenAIClient extends BaseClient {
 
       const instructionsPayload = [
         {
-          role: this.options.titleMessageRole ?? 'system',
+          role: this.options.titleMessageRole ?? (this.isOllama ? 'user' : 'system'),
           content: `Please generate ${titleInstruction}
 
 ${convo}
@@ -1031,7 +1023,7 @@ ${convo}
   async chatCompletion({ payload, onProgress, abortController = null }) {
     let error = null;
     const errorCallback = (err) => (error = err);
-    let intermediateReply = '';
+    const intermediateReply = [];
     try {
       if (!abortController) {
         abortController = new AbortController();
@@ -1182,8 +1174,10 @@ ${convo}
         });
       }
 
+      const streamRate = this.options.streamRate ?? Constants.DEFAULT_STREAM_RATE;
+
       if (this.message_file_map && this.isOllama) {
-        const ollamaClient = new OllamaClient({ baseURL });
+        const ollamaClient = new OllamaClient({ baseURL, streamRate });
         return await ollamaClient.chatCompletion({
           payload: modelOptions,
           onProgress,
@@ -1192,7 +1186,15 @@ ${convo}
       }
 
       let UnexpectedRoleError = false;
+      /** @type {Promise<void>} */
+      let streamPromise;
+      /** @type {(value: void | PromiseLike<void>) => void} */
+      let streamResolve;
+
       if (modelOptions.stream) {
+        streamPromise = new Promise((resolve) => {
+          streamResolve = resolve;
+        });
         const stream = await openai.beta.chat.completions
           .stream({
             ...modelOptions,
@@ -1204,39 +1206,41 @@ ${convo}
           .on('error', (err) => {
             handleOpenAIErrors(err, errorCallback, 'stream');
           })
-          .on('finalChatCompletion', (finalChatCompletion) => {
+          .on('finalChatCompletion', async (finalChatCompletion) => {
             const finalMessage = finalChatCompletion?.choices?.[0]?.message;
-            if (finalMessage && finalMessage?.role !== 'assistant') {
+            if (!finalMessage) {
+              return;
+            }
+            await streamPromise;
+            if (finalMessage?.role !== 'assistant') {
               finalChatCompletion.choices[0].message.role = 'assistant';
             }
 
-            if (finalMessage && !finalMessage?.content?.trim()) {
-              finalChatCompletion.choices[0].message.content = intermediateReply;
+            if (typeof finalMessage.content !== 'string' || finalMessage.content.trim() === '') {
+              finalChatCompletion.choices[0].message.content = intermediateReply.join('');
             }
           })
           .on('finalMessage', (message) => {
             if (message?.role !== 'assistant') {
-              stream.messages.push({ role: 'assistant', content: intermediateReply });
+              stream.messages.push({ role: 'assistant', content: intermediateReply.join('') });
               UnexpectedRoleError = true;
             }
           });
 
-        const azureDelay = this.modelOptions.model?.includes('gpt-4') ? 30 : 17;
-
         for await (const chunk of stream) {
           const token = chunk.choices[0]?.delta?.content || '';
-          intermediateReply += token;
+          intermediateReply.push(token);
           onProgress(token);
           if (abortController.signal.aborted) {
             stream.controller.abort();
             break;
           }
 
-          if (this.azure) {
-            await sleep(azureDelay);
-          }
+          await sleep(streamRate);
         }
 
+        streamResolve();
+
         if (!UnexpectedRoleError) {
           chatCompletion = await stream.finalChatCompletion().catch((err) => {
             handleOpenAIErrors(err, errorCallback, 'finalChatCompletion');
@@ -1264,19 +1268,29 @@ ${convo}
         throw new Error('Chat completion failed');
       }
 
-      const { message, finish_reason } = chatCompletion.choices[0];
-      if (chatCompletion) {
-        this.metadata = { finish_reason };
+      const { choices } = chatCompletion;
+      if (!Array.isArray(choices) || choices.length === 0) {
+        logger.warn('[OpenAIClient] Chat completion response has no choices');
+        return intermediateReply.join('');
       }
 
+      const { message, finish_reason } = choices[0] ?? {};
+      this.metadata = { finish_reason };
+
       logger.debug('[OpenAIClient] chatCompletion response', chatCompletion);
 
-      if (!message?.content?.trim() && intermediateReply.length) {
+      if (!message) {
+        logger.warn('[OpenAIClient] Message is undefined in chatCompletion response');
+        return intermediateReply.join('');
+      }
+
+      if (typeof message.content !== 'string' || message.content.trim() === '') {
+        const reply = intermediateReply.join('');
         logger.debug(
           '[OpenAIClient] chatCompletion: using intermediateReply due to empty message.content',
-          { intermediateReply },
+          { intermediateReply: reply },
         );
-        return intermediateReply;
+        return reply;
       }
 
       return message.content;
@@ -1285,7 +1299,7 @@ ${convo}
         err?.message?.includes('abort') ||
         (err instanceof OpenAI.APIError && err?.message?.includes('abort'))
       ) {
-        return intermediateReply;
+        return intermediateReply.join('');
       }
       if (
         err?.message?.includes(
@@ -1300,10 +1314,10 @@ ${convo}
         (err instanceof OpenAI.OpenAIError && err?.message?.includes('missing finish_reason'))
       ) {
         logger.error('[OpenAIClient] Known OpenAI error:', err);
-        return intermediateReply;
+        return intermediateReply.join('');
       } else if (err instanceof OpenAI.APIError) {
-        if (intermediateReply) {
-          return intermediateReply;
+        if (intermediateReply.length > 0) {
+          return intermediateReply.join('');
         } else {
           throw err;
         }

api/app/clients/PluginsClient.js

@@ -1,5 +1,6 @@
 const OpenAIClient = require('./OpenAIClient');
 const { CallbackManager } = require('langchain/callbacks');
+const { CacheKeys, Time } = require('librechat-data-provider');
 const { BufferMemory, ChatMessageHistory } = require('langchain/memory');
 const { initializeCustomAgent, initializeFunctionsAgent } = require('./agents');
 const { addImages, buildErrorInput, buildPromptPrefix } = require('./output_parsers');
@@ -11,6 +12,7 @@ const { SelfReflectionTool } = require('./tools');
 const { isEnabled } = require('~/server/utils');
 const { extractBaseURL } = require('~/utils');
 const { loadTools } = require('./tools/util');
+const { getLogStores } = require('~/cache');
 const { logger } = require('~/config');
 
 class PluginsClient extends OpenAIClient {
@@ -40,6 +42,7 @@ class PluginsClient extends OpenAIClient {
 
   getSaveOptions() {
     return {
+      artifacts: this.options.artifacts,
       chatGptLabel: this.options.chatGptLabel,
       promptPrefix: this.options.promptPrefix,
       tools: this.options.tools,
@@ -143,16 +146,22 @@ class PluginsClient extends OpenAIClient {
 
     // initialize agent
     const initializer = this.functionsAgent ? initializeFunctionsAgent : initializeCustomAgent;
+
+    let customInstructions = (this.options.promptPrefix ?? '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      customInstructions = `${customInstructions ?? ''}\n${this.options.artifactsPrompt}`.trim();
+    }
+
     this.executor = await initializer({
       model,
       signal,
       pastMessages,
       tools: this.tools,
+      customInstructions,
       verbose: this.options.debug,
       returnIntermediateSteps: true,
       customName: this.options.chatGptLabel,
       currentDateString: this.currentDateString,
-      customInstructions: this.options.promptPrefix,
       callbackManager: CallbackManager.fromHandlers({
         async handleAgentAction(action, runId) {
           handleAction(action, runId, onAgentAction);
@@ -220,6 +229,13 @@ class PluginsClient extends OpenAIClient {
     }
   }
 
+  /**
+   *
+   * @param {TMessage} responseMessage
+   * @param {Partial<TMessage>} saveOptions
+   * @param {string} user
+   * @returns
+   */
   async handleResponseMessage(responseMessage, saveOptions, user) {
     const { output, errorMessage, ...result } = this.result;
     logger.debug('[PluginsClient][handleResponseMessage] Output:', {
@@ -239,6 +255,15 @@ class PluginsClient extends OpenAIClient {
     }
 
     this.responsePromise = this.saveMessageToDatabase(responseMessage, saveOptions, user);
+    const messageCache = getLogStores(CacheKeys.MESSAGES);
+    messageCache.set(
+      responseMessage.messageId,
+      {
+        text: responseMessage.text,
+        complete: true,
+      },
+      Time.FIVE_MINUTES,
+    );
     delete responseMessage.tokenCount;
     return { ...responseMessage, ...result };
   }

api/app/clients/llm/RunManager.js

@@ -1,5 +1,5 @@
 const { createStartHandler } = require('~/app/clients/callbacks');
-const spendTokens = require('~/models/spendTokens');
+const { spendTokens } = require('~/models/spendTokens');
 const { logger } = require('~/config');
 
 class RunManager {

api/app/clients/llm/createLLM.js

@@ -8,7 +8,7 @@ const { isEnabled } = require('~/server/utils');
  * @param {Object} options - The options for creating the LLM.
  * @param {ModelOptions} options.modelOptions - The options specific to the model, including modelName, temperature, presence_penalty, frequency_penalty, and other model-related settings.
  * @param {ConfigOptions} options.configOptions - Configuration options for the API requests, including proxy settings and custom headers.
- * @param {Callbacks} options.callbacks - Callback functions for managing the lifecycle of the LLM, including token buffers, context, and initial message count.
+ * @param {Callbacks} [options.callbacks] - Callback functions for managing the lifecycle of the LLM, including token buffers, context, and initial message count.
  * @param {boolean} [options.streaming=false] - Determines if the LLM should operate in streaming mode.
  * @param {string} options.openAIApiKey - The API key for OpenAI, used for authentication.
  * @param {AzureOptions} [options.azure={}] - Optional Azure-specific configurations. If provided, Azure configurations take precedence over OpenAI configurations.

api/app/clients/output_parsers/addImages.js

@@ -60,10 +60,10 @@ function addImages(intermediateSteps, responseMessage) {
     if (!observation || !observation.includes('![')) {
       return;
     }
-    const observedImagePath = observation.match(/!\[.*\]\([^)]*\)/g);
+    const observedImagePath = observation.match(/!\[[^(]*\]\([^)]*\)/g);
     if (observedImagePath && !responseMessage.text.includes(observedImagePath[0])) {
-      responseMessage.text += '\n' + observation;
-      logger.debug('[addImages] added image from intermediateSteps:', observation);
+      responseMessage.text += '\n' + observedImagePath[0];
+      logger.debug('[addImages] added image from intermediateSteps:', observedImagePath[0]);
     }
   });
 }

api/app/clients/output_parsers/addImages.spec.js

@@ -81,4 +81,62 @@ describe('addImages', () => {
     addImages(intermediateSteps, responseMessage);
     expect(responseMessage.text).toBe(`${originalText}\n${imageMarkdown}`);
   });
+
+  it('should extract only image markdowns when there is text between them', () => {
+    const markdownWithTextBetweenImages = `
+      ![image1](/images/image1.png)
+      Some text between images that should not be included.
+      ![image2](/images/image2.png)
+      More text that should be ignored.
+      ![image3](/images/image3.png)
+    `;
+    intermediateSteps.push({ observation: markdownWithTextBetweenImages });
+    addImages(intermediateSteps, responseMessage);
+    expect(responseMessage.text).toBe('\n![image1](/images/image1.png)');
+  });
+
+  it('should only return the first image when multiple images are present', () => {
+    const markdownWithMultipleImages = `
+      ![image1](/images/image1.png)
+      ![image2](/images/image2.png)
+      ![image3](/images/image3.png)
+    `;
+    intermediateSteps.push({ observation: markdownWithMultipleImages });
+    addImages(intermediateSteps, responseMessage);
+    expect(responseMessage.text).toBe('\n![image1](/images/image1.png)');
+  });
+
+  it('should not include any text or metadata surrounding the image markdown', () => {
+    const markdownWithMetadata = `
+      Title: Test Document
+      Author: John Doe
+      ![image1](/images/image1.png)
+      Some content after the image.
+      Vector values: [0.1, 0.2, 0.3]
+    `;
+    intermediateSteps.push({ observation: markdownWithMetadata });
+    addImages(intermediateSteps, responseMessage);
+    expect(responseMessage.text).toBe('\n![image1](/images/image1.png)');
+  });
+
+  it('should handle complex markdown with multiple images and only return the first one', () => {
+    const complexMarkdown = `
+      # Document Title
+      
+      ## Section 1
+      Here's some text with an embedded image:
+      ![image1](/images/image1.png)
+      
+      ## Section 2
+      More text here...
+      ![image2](/images/image2.png)
+      
+      ### Subsection
+      Even more content
+      ![image3](/images/image3.png)
+    `;
+    intermediateSteps.push({ observation: complexMarkdown });
+    addImages(intermediateSteps, responseMessage);
+    expect(responseMessage.text).toBe('\n![image1](/images/image1.png)');
+  });
 });

api/app/clients/prompts/addCacheControl.js

@@ -0,0 +1,43 @@
+/**
+ * Anthropic API: Adds cache control to the appropriate user messages in the payload.
+ * @param {Array<AnthropicMessage>} messages - The array of message objects.
+ * @returns {Array<AnthropicMessage>} - The updated array of message objects with cache control added.
+ */
+function addCacheControl(messages) {
+  if (!Array.isArray(messages) || messages.length < 2) {
+    return messages;
+  }
+
+  const updatedMessages = [...messages];
+  let userMessagesModified = 0;
+
+  for (let i = updatedMessages.length - 1; i >= 0 && userMessagesModified < 2; i--) {
+    const message = updatedMessages[i];
+    if (message.role !== 'user') {
+      continue;
+    }
+
+    if (typeof message.content === 'string') {
+      message.content = [
+        {
+          type: 'text',
+          text: message.content,
+          cache_control: { type: 'ephemeral' },
+        },
+      ];
+      userMessagesModified++;
+    } else if (Array.isArray(message.content)) {
+      for (let j = message.content.length - 1; j >= 0; j--) {
+        if (message.content[j].type === 'text') {
+          message.content[j].cache_control = { type: 'ephemeral' };
+          userMessagesModified++;
+          break;
+        }
+      }
+    }
+  }
+
+  return updatedMessages;
+}
+
+module.exports = addCacheControl;

api/app/clients/prompts/addCacheControl.spec.js

@@ -0,0 +1,227 @@
+const addCacheControl = require('./addCacheControl');
+
+describe('addCacheControl', () => {
+  test('should add cache control to the last two user messages with array content', () => {
+    const messages = [
+      { role: 'user', content: [{ type: 'text', text: 'Hello' }] },
+      { role: 'assistant', content: [{ type: 'text', text: 'Hi there' }] },
+      { role: 'user', content: [{ type: 'text', text: 'How are you?' }] },
+      { role: 'assistant', content: [{ type: 'text', text: 'I\'m doing well, thanks!' }] },
+      { role: 'user', content: [{ type: 'text', text: 'Great!' }] },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).not.toHaveProperty('cache_control');
+    expect(result[2].content[0].cache_control).toEqual({ type: 'ephemeral' });
+    expect(result[4].content[0].cache_control).toEqual({ type: 'ephemeral' });
+  });
+
+  test('should add cache control to the last two user messages with string content', () => {
+    const messages = [
+      { role: 'user', content: 'Hello' },
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'user', content: 'How are you?' },
+      { role: 'assistant', content: 'I\'m doing well, thanks!' },
+      { role: 'user', content: 'Great!' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content).toBe('Hello');
+    expect(result[2].content[0]).toEqual({
+      type: 'text',
+      text: 'How are you?',
+      cache_control: { type: 'ephemeral' },
+    });
+    expect(result[4].content[0]).toEqual({
+      type: 'text',
+      text: 'Great!',
+      cache_control: { type: 'ephemeral' },
+    });
+  });
+
+  test('should handle mixed string and array content', () => {
+    const messages = [
+      { role: 'user', content: 'Hello' },
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'user', content: [{ type: 'text', text: 'How are you?' }] },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).toEqual({
+      type: 'text',
+      text: 'Hello',
+      cache_control: { type: 'ephemeral' },
+    });
+    expect(result[2].content[0].cache_control).toEqual({ type: 'ephemeral' });
+  });
+
+  test('should handle less than two user messages', () => {
+    const messages = [
+      { role: 'user', content: 'Hello' },
+      { role: 'assistant', content: 'Hi there' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).toEqual({
+      type: 'text',
+      text: 'Hello',
+      cache_control: { type: 'ephemeral' },
+    });
+    expect(result[1].content).toBe('Hi there');
+  });
+
+  test('should return original array if no user messages', () => {
+    const messages = [
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'assistant', content: 'How can I help?' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result).toEqual(messages);
+  });
+
+  test('should handle empty array', () => {
+    const messages = [];
+    const result = addCacheControl(messages);
+    expect(result).toEqual([]);
+  });
+
+  test('should handle non-array input', () => {
+    const messages = 'not an array';
+    const result = addCacheControl(messages);
+    expect(result).toBe('not an array');
+  });
+
+  test('should not modify assistant messages', () => {
+    const messages = [
+      { role: 'user', content: 'Hello' },
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'user', content: 'How are you?' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[1].content).toBe('Hi there');
+  });
+
+  test('should handle multiple content items in user messages', () => {
+    const messages = [
+      {
+        role: 'user',
+        content: [
+          { type: 'text', text: 'Hello' },
+          { type: 'image', url: 'http://example.com/image.jpg' },
+          { type: 'text', text: 'This is an image' },
+        ],
+      },
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'user', content: 'How are you?' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).not.toHaveProperty('cache_control');
+    expect(result[0].content[1]).not.toHaveProperty('cache_control');
+    expect(result[0].content[2].cache_control).toEqual({ type: 'ephemeral' });
+    expect(result[2].content[0]).toEqual({
+      type: 'text',
+      text: 'How are you?',
+      cache_control: { type: 'ephemeral' },
+    });
+  });
+
+  test('should handle an array with mixed content types', () => {
+    const messages = [
+      { role: 'user', content: 'Hello' },
+      { role: 'assistant', content: 'Hi there' },
+      { role: 'user', content: [{ type: 'text', text: 'How are you?' }] },
+      { role: 'assistant', content: 'I\'m doing well, thanks!' },
+      { role: 'user', content: 'Great!' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content).toEqual('Hello');
+    expect(result[2].content[0]).toEqual({
+      type: 'text',
+      text: 'How are you?',
+      cache_control: { type: 'ephemeral' },
+    });
+    expect(result[4].content).toEqual([
+      {
+        type: 'text',
+        text: 'Great!',
+        cache_control: { type: 'ephemeral' },
+      },
+    ]);
+    expect(result[1].content).toBe('Hi there');
+    expect(result[3].content).toBe('I\'m doing well, thanks!');
+  });
+
+  test('should handle edge case with multiple content types', () => {
+    const messages = [
+      {
+        role: 'user',
+        content: [
+          {
+            type: 'image',
+            source: { type: 'base64', media_type: 'image/png', data: 'some_base64_string' },
+          },
+          {
+            type: 'image',
+            source: { type: 'base64', media_type: 'image/png', data: 'another_base64_string' },
+          },
+          { type: 'text', text: 'what do all these images have in common' },
+        ],
+      },
+      { role: 'assistant', content: 'I see multiple images.' },
+      { role: 'user', content: 'Correct!' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).not.toHaveProperty('cache_control');
+    expect(result[0].content[1]).not.toHaveProperty('cache_control');
+    expect(result[0].content[2].cache_control).toEqual({ type: 'ephemeral' });
+    expect(result[2].content[0]).toEqual({
+      type: 'text',
+      text: 'Correct!',
+      cache_control: { type: 'ephemeral' },
+    });
+  });
+
+  test('should handle user message with no text block', () => {
+    const messages = [
+      {
+        role: 'user',
+        content: [
+          {
+            type: 'image',
+            source: { type: 'base64', media_type: 'image/png', data: 'some_base64_string' },
+          },
+          {
+            type: 'image',
+            source: { type: 'base64', media_type: 'image/png', data: 'another_base64_string' },
+          },
+        ],
+      },
+      { role: 'assistant', content: 'I see two images.' },
+      { role: 'user', content: 'Correct!' },
+    ];
+
+    const result = addCacheControl(messages);
+
+    expect(result[0].content[0]).not.toHaveProperty('cache_control');
+    expect(result[0].content[1]).not.toHaveProperty('cache_control');
+    expect(result[2].content[0]).toEqual({
+      type: 'text',
+      text: 'Correct!',
+      cache_control: { type: 'ephemeral' },
+    });
+  });
+});

api/app/clients/prompts/artifacts.js

@@ -0,0 +1,527 @@
+const dedent = require('dedent');
+const { EModelEndpoint, ArtifactModes } = require('librechat-data-provider');
+const { generateShadcnPrompt } = require('~/app/clients/prompts/shadcn-docs/generate');
+const { components } = require('~/app/clients/prompts/shadcn-docs/components');
+
+// eslint-disable-next-line no-unused-vars
+const artifactsPromptV1 = dedent`The assistant can create and reference artifacts during conversations.
+  
+Artifacts are for substantial, self-contained content that users might modify or reuse, displayed in a separate UI window for clarity.
+
+# Good artifacts are...
+- Substantial content (>15 lines)
+- Content that the user is likely to modify, iterate on, or take ownership of
+- Self-contained, complex content that can be understood on its own, without context from the conversation
+- Content intended for eventual use outside the conversation (e.g., reports, emails, presentations)
+- Content likely to be referenced or reused multiple times
+
+# Don't use artifacts for...
+- Simple, informational, or short content, such as brief code snippets, mathematical equations, or small examples
+- Primarily explanatory, instructional, or illustrative content, such as examples provided to clarify a concept
+- Suggestions, commentary, or feedback on existing artifacts
+- Conversational or explanatory content that doesn't represent a standalone piece of work
+- Content that is dependent on the current conversational context to be useful
+- Content that is unlikely to be modified or iterated upon by the user
+- Request from users that appears to be a one-off question
+
+# Usage notes
+- One artifact per message unless specifically requested
+- Prefer in-line content (don't use artifacts) when possible. Unnecessary use of artifacts can be jarring for users.
+- If a user asks the assistant to "draw an SVG" or "make a website," the assistant does not need to explain that it doesn't have these capabilities. Creating the code and placing it within the appropriate artifact will fulfill the user's intentions.
+- If asked to generate an image, the assistant can offer an SVG instead. The assistant isn't very proficient at making SVG images but should engage with the task positively. Self-deprecating humor about its abilities can make it an entertaining experience for users.
+- The assistant errs on the side of simplicity and avoids overusing artifacts for content that can be effectively presented within the conversation.
+- Always provide complete, specific, and fully functional content without any placeholders, ellipses, or 'remains the same' comments.
+
+<artifact_instructions>
+  When collaborating with the user on creating content that falls into compatible categories, the assistant should follow these steps:
+
+  1. Create the artifact using the following format:
+
+     :::artifact{identifier="unique-identifier" type="mime-type" title="Artifact Title"}
+     \`\`\`
+     Your artifact content here
+     \`\`\`
+     :::
+
+  2. Assign an identifier to the \`identifier\` attribute. For updates, reuse the prior identifier. For new artifacts, the identifier should be descriptive and relevant to the content, using kebab-case (e.g., "example-code-snippet"). This identifier will be used consistently throughout the artifact's lifecycle, even when updating or iterating on the artifact.
+  3. Include a \`title\` attribute to provide a brief title or description of the content.
+  4. Add a \`type\` attribute to specify the type of content the artifact represents. Assign one of the following values to the \`type\` attribute:
+    - HTML: "text/html"
+      - The user interface can render single file HTML pages placed within the artifact tags. HTML, JS, and CSS should be in a single file when using the \`text/html\` type.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - The only place external scripts can be imported from is https://cdnjs.cloudflare.com
+    - Mermaid Diagrams: "application/vnd.mermaid"
+      - The user interface will render Mermaid diagrams placed within the artifact tags.
+    - React Components: "application/vnd.react"
+      - Use this for displaying either: React elements, e.g. \`<strong>Hello World!</strong>\`, React pure functional components, e.g. \`() => <strong>Hello World!</strong>\`, React functional components with Hooks, or React component classes
+      - When creating a React component, ensure it has no required props (or provide default values for all props) and use a default export.
+      - Use Tailwind classes for styling. DO NOT USE ARBITRARY VALUES (e.g. \`h-[600px]\`).
+      - Base React is available to be imported. To use hooks, first import it at the top of the artifact, e.g. \`import { useState } from "react"\`
+      - The lucide-react@0.263.1 library is available to be imported. e.g. \`import { Camera } from "lucide-react"\` & \`<Camera color="red" size={48} />\`
+      - The recharts charting library is available to be imported, e.g. \`import { LineChart, XAxis, ... } from "recharts"\` & \`<LineChart ...><XAxis dataKey="name"> ...\`
+      - The assistant can use prebuilt components from the \`shadcn/ui\` library after it is imported: \`import { Alert, AlertDescription, AlertTitle, AlertDialog, AlertDialogAction } from '/components/ui/alert';\`. If using components from the shadcn/ui library, the assistant mentions this to the user and offers to help them install the components if necessary.
+      - Components MUST be imported from \`/components/ui/name\` and NOT from \`/components/name\` or \`@/components/ui/name\`.
+      - NO OTHER LIBRARIES (e.g. zod, hookform) ARE INSTALLED OR ABLE TO BE IMPORTED.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - If you are unable to follow the above requirements for any reason, don't use artifacts and use regular code blocks instead, which will not attempt to render the component.
+  5. Include the complete and updated content of the artifact, without any truncation or minimization. Don't use "// rest of the code remains the same...".
+  6. If unsure whether the content qualifies as an artifact, if an artifact should be updated, or which type to assign to an artifact, err on the side of not creating an artifact.
+  7. Always use triple backticks (\`\`\`) to enclose the content within the artifact, regardless of the content type.
+</artifact_instructions>
+
+Here are some examples of correct usage of artifacts:
+
+<examples>
+  <example_docstring>
+    This example demonstrates how to create a Mermaid artifact for a simple flow chart.
+  </example_docstring>
+
+  <example>
+    <user_query>Can you create a simple flow chart showing the process of making tea using Mermaid?</user_query>
+
+    <assistant_response>
+      Sure! Here's a simple flow chart depicting the process of making tea using Mermaid syntax:
+
+      :::artifact{identifier="tea-making-flowchart" type="application/vnd.mermaid" title="Flow chart: Making Tea"}
+      \`\`\`mermaid
+      graph TD
+          A[Start] --> B{Water boiled?}
+          B -->|Yes| C[Add tea leaves to cup]
+          B -->|No| D[Boil water]
+          D --> B
+          C --> E[Pour boiling water into cup]
+          E --> F[Steep tea for desired time]
+          F --> G[Remove tea leaves]
+          G --> H[Add milk or sugar, if desired]
+          H --> I[Enjoy your tea!]
+          I --> J[End]
+      \`\`\`
+      :::
+
+      This flow chart uses Mermaid syntax to visualize the steps involved in making a cup of tea. Here's a brief explanation of the process:
+
+      1. Start
+      2. Check if water is boiled
+      3. If not boiled, boil the water
+      4. Once water is boiled, add tea leaves to the cup
+      5. Pour boiling water into the cup
+      6. Steep the tea for the desired time
+      7. Remove the tea leaves
+      8. Optionally add milk or sugar
+      9. Enjoy your tea!
+      10. End
+
+      This chart provides a clear visual representation of the tea-making process. You can easily modify or expand this chart if you want to add more details or steps to the process. Let me know if you'd like any changes or have any questions!
+    </assistant_response>
+  </example>
+</examples>`;
+const artifactsPrompt = dedent`The assistant can create and reference artifacts during conversations.
+  
+Artifacts are for substantial, self-contained content that users might modify or reuse, displayed in a separate UI window for clarity.
+
+# Good artifacts are...
+- Substantial content (>15 lines)
+- Content that the user is likely to modify, iterate on, or take ownership of
+- Self-contained, complex content that can be understood on its own, without context from the conversation
+- Content intended for eventual use outside the conversation (e.g., reports, emails, presentations)
+- Content likely to be referenced or reused multiple times
+
+# Don't use artifacts for...
+- Simple, informational, or short content, such as brief code snippets, mathematical equations, or small examples
+- Primarily explanatory, instructional, or illustrative content, such as examples provided to clarify a concept
+- Suggestions, commentary, or feedback on existing artifacts
+- Conversational or explanatory content that doesn't represent a standalone piece of work
+- Content that is dependent on the current conversational context to be useful
+- Content that is unlikely to be modified or iterated upon by the user
+- Request from users that appears to be a one-off question
+
+# Usage notes
+- One artifact per message unless specifically requested
+- Prefer in-line content (don't use artifacts) when possible. Unnecessary use of artifacts can be jarring for users.
+- If a user asks the assistant to "draw an SVG" or "make a website," the assistant does not need to explain that it doesn't have these capabilities. Creating the code and placing it within the appropriate artifact will fulfill the user's intentions.
+- If asked to generate an image, the assistant can offer an SVG instead. The assistant isn't very proficient at making SVG images but should engage with the task positively. Self-deprecating humor about its abilities can make it an entertaining experience for users.
+- The assistant errs on the side of simplicity and avoids overusing artifacts for content that can be effectively presented within the conversation.
+- Always provide complete, specific, and fully functional content for artifacts without any snippets, placeholders, ellipses, or 'remains the same' comments.
+- If an artifact is not necessary or requested, the assistant should not mention artifacts at all, and respond to the user accordingly.
+
+<artifact_instructions>
+  When collaborating with the user on creating content that falls into compatible categories, the assistant should follow these steps:
+
+  1. Create the artifact using the following format:
+
+     :::artifact{identifier="unique-identifier" type="mime-type" title="Artifact Title"}
+     \`\`\`
+     Your artifact content here
+     \`\`\`
+     :::
+
+  2. Assign an identifier to the \`identifier\` attribute. For updates, reuse the prior identifier. For new artifacts, the identifier should be descriptive and relevant to the content, using kebab-case (e.g., "example-code-snippet"). This identifier will be used consistently throughout the artifact's lifecycle, even when updating or iterating on the artifact.
+  3. Include a \`title\` attribute to provide a brief title or description of the content.
+  4. Add a \`type\` attribute to specify the type of content the artifact represents. Assign one of the following values to the \`type\` attribute:
+    - HTML: "text/html"
+      - The user interface can render single file HTML pages placed within the artifact tags. HTML, JS, and CSS should be in a single file when using the \`text/html\` type.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - The only place external scripts can be imported from is https://cdnjs.cloudflare.com
+    - SVG: "image/svg+xml"
+      - The user interface will render the Scalable Vector Graphics (SVG) image within the artifact tags.
+      - The assistant should specify the viewbox of the SVG rather than defining a width/height
+    - Mermaid Diagrams: "application/vnd.mermaid"
+      - The user interface will render Mermaid diagrams placed within the artifact tags.
+    - React Components: "application/vnd.react"
+      - Use this for displaying either: React elements, e.g. \`<strong>Hello World!</strong>\`, React pure functional components, e.g. \`() => <strong>Hello World!</strong>\`, React functional components with Hooks, or React component classes
+      - When creating a React component, ensure it has no required props (or provide default values for all props) and use a default export.
+      - Use Tailwind classes for styling. DO NOT USE ARBITRARY VALUES (e.g. \`h-[600px]\`).
+      - Base React is available to be imported. To use hooks, first import it at the top of the artifact, e.g. \`import { useState } from "react"\`
+      - The lucide-react@0.394.0 library is available to be imported. e.g. \`import { Camera } from "lucide-react"\` & \`<Camera color="red" size={48} />\`
+      - The recharts charting library is available to be imported, e.g. \`import { LineChart, XAxis, ... } from "recharts"\` & \`<LineChart ...><XAxis dataKey="name"> ...\`
+      - The three.js library is available to be imported, e.g. \`import * as THREE from "three";\`
+      - The date-fns library is available to be imported, e.g. \`import { compareAsc, format } from "date-fns";\`
+      - The react-day-picker library is available to be imported, e.g. \`import { DayPicker } from "react-day-picker";\`
+      - The assistant can use prebuilt components from the \`shadcn/ui\` library after it is imported: \`import { Alert, AlertDescription, AlertTitle, AlertDialog, AlertDialogAction } from '/components/ui/alert';\`. If using components from the shadcn/ui library, the assistant mentions this to the user and offers to help them install the components if necessary.
+      - Components MUST be imported from \`/components/ui/name\` and NOT from \`/components/name\` or \`@/components/ui/name\`.
+      - NO OTHER LIBRARIES (e.g. zod, hookform) ARE INSTALLED OR ABLE TO BE IMPORTED.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - When iterating on code, ensure that the code is complete and functional without any snippets, placeholders, or ellipses.
+      - If you are unable to follow the above requirements for any reason, don't use artifacts and use regular code blocks instead, which will not attempt to render the component.
+  5. Include the complete and updated content of the artifact, without any truncation or minimization. Don't use "// rest of the code remains the same...".
+  6. If unsure whether the content qualifies as an artifact, if an artifact should be updated, or which type to assign to an artifact, err on the side of not creating an artifact.
+  7. Always use triple backticks (\`\`\`) to enclose the content within the artifact, regardless of the content type.
+</artifact_instructions>
+
+Here are some examples of correct usage of artifacts:
+
+<examples>
+  <example_docstring>
+    This example demonstrates how to create a Mermaid artifact for a simple flow chart.
+  </example_docstring>
+
+  <example>
+    <user_query>Can you create a simple flow chart showing the process of making tea using Mermaid?</user_query>
+
+    <assistant_response>
+      Sure! Here's a simple flow chart depicting the process of making tea using Mermaid syntax:
+
+      :::artifact{identifier="tea-making-flowchart" type="application/vnd.mermaid" title="Flow chart: Making Tea"}
+      \`\`\`mermaid
+      graph TD
+          A[Start] --> B{Water boiled?}
+          B -->|Yes| C[Add tea leaves to cup]
+          B -->|No| D[Boil water]
+          D --> B
+          C --> E[Pour boiling water into cup]
+          E --> F[Steep tea for desired time]
+          F --> G[Remove tea leaves]
+          G --> H[Add milk or sugar, if desired]
+          H --> I[Enjoy your tea!]
+          I --> J[End]
+      \`\`\`
+      :::
+
+      This flow chart uses Mermaid syntax to visualize the steps involved in making a cup of tea. Here's a brief explanation of the process:
+
+      1. Start
+      2. Check if water is boiled
+      3. If not boiled, boil the water
+      4. Once water is boiled, add tea leaves to the cup
+      5. Pour boiling water into the cup
+      6. Steep the tea for the desired time
+      7. Remove the tea leaves
+      8. Optionally add milk or sugar
+      9. Enjoy your tea!
+      10. End
+
+      This chart provides a clear visual representation of the tea-making process. You can easily modify or expand this chart if you want to add more details or steps to the process. Let me know if you'd like any changes or have any questions!
+    </assistant_response>
+  </example>
+
+  <example>
+    <user_query>Create a simple React counter component</user_query>
+    <assistant_response>
+      Here's a simple React counter component:
+
+      :::artifact{identifier="react-counter" type="application/vnd.react" title="React Counter"}
+      \`\`\`
+      import { useState } from 'react';
+
+      export default function Counter() {
+        const [count, setCount] = useState(0);
+        return (
+          <div className="p-4">
+            <p className="mb-2">Count: {count}</p>
+            <button className="bg-blue-500 text-white px-4 py-2 rounded" onClick={() => setCount(count + 1)}>
+              Increment
+            </button>
+          </div>
+        );
+      }
+      \`\`\`
+      :::
+
+      This component creates a simple counter with an increment button.
+    </assistant_response>
+  </example>
+
+  <example>
+    <user_query>Create a basic HTML structure for a blog post</user_query>
+    <assistant_response>
+      Here's a basic HTML structure for a blog post:
+
+      :::artifact{identifier="blog-post-html" type="text/html" title="Blog Post HTML"}
+      \`\`\`
+      <!DOCTYPE html>
+      <html lang="en">
+      <head>
+        <meta charset="UTF-8">
+        <meta name="viewport" content="width=device-width, initial-scale=1.0">
+        <title>My Blog Post</title>
+        <style>
+          body { font-family: Arial, sans-serif; line-height: 1.6; max-width: 800px; margin: 0 auto; padding: 20px; }
+          h1 { color: #333; }
+          p { margin-bottom: 15px; }
+        </style>
+      </head>
+      <body>
+        <header>
+          <h1>My First Blog Post</h1>
+        </header>
+        <main>
+          <article>
+            <p>This is the content of my blog post. It's short and sweet!</p>
+          </article>
+        </main>
+        <footer>
+          <p>&copy; 2023 My Blog</p>
+        </footer>
+      </body>
+      </html>
+      \`\`\`
+      :::
+
+      This HTML structure provides a simple layout for a blog post.
+    </assistant_response>
+  </example>
+</examples>`;
+
+const artifactsOpenAIPrompt = dedent`The assistant can create and reference artifacts during conversations.
+  
+Artifacts are for substantial, self-contained content that users might modify or reuse, displayed in a separate UI window for clarity.
+
+# Good artifacts are...
+- Substantial content (>15 lines)
+- Content that the user is likely to modify, iterate on, or take ownership of
+- Self-contained, complex content that can be understood on its own, without context from the conversation
+- Content intended for eventual use outside the conversation (e.g., reports, emails, presentations)
+- Content likely to be referenced or reused multiple times
+
+# Don't use artifacts for...
+- Simple, informational, or short content, such as brief code snippets, mathematical equations, or small examples
+- Primarily explanatory, instructional, or illustrative content, such as examples provided to clarify a concept
+- Suggestions, commentary, or feedback on existing artifacts
+- Conversational or explanatory content that doesn't represent a standalone piece of work
+- Content that is dependent on the current conversational context to be useful
+- Content that is unlikely to be modified or iterated upon by the user
+- Request from users that appears to be a one-off question
+
+# Usage notes
+- One artifact per message unless specifically requested
+- Prefer in-line content (don't use artifacts) when possible. Unnecessary use of artifacts can be jarring for users.
+- If a user asks the assistant to "draw an SVG" or "make a website," the assistant does not need to explain that it doesn't have these capabilities. Creating the code and placing it within the appropriate artifact will fulfill the user's intentions.
+- If asked to generate an image, the assistant can offer an SVG instead. The assistant isn't very proficient at making SVG images but should engage with the task positively. Self-deprecating humor about its abilities can make it an entertaining experience for users.
+- The assistant errs on the side of simplicity and avoids overusing artifacts for content that can be effectively presented within the conversation.
+- Always provide complete, specific, and fully functional content for artifacts without any snippets, placeholders, ellipses, or 'remains the same' comments.
+- If an artifact is not necessary or requested, the assistant should not mention artifacts at all, and respond to the user accordingly.
+
+## Artifact Instructions
+  When collaborating with the user on creating content that falls into compatible categories, the assistant should follow these steps:
+
+  1. Create the artifact using the following remark-directive markdown format:
+
+      :::artifact{identifier="unique-identifier" type="mime-type" title="Artifact Title"}
+      \`\`\`
+      Your artifact content here
+      \`\`\`
+      :::
+
+  a. Example of correct format:
+
+      :::artifact{identifier="example-artifact" type="text/plain" title="Example Artifact"}
+      \`\`\`
+      This is the content of the artifact.
+      It can span multiple lines.
+      \`\`\`
+      :::
+
+  b. Common mistakes to avoid:
+   - Don't split the opening ::: line
+   - Don't add extra backticks outside the artifact structure
+   - Don't omit the closing :::
+
+  2. Assign an identifier to the \`identifier\` attribute. For updates, reuse the prior identifier. For new artifacts, the identifier should be descriptive and relevant to the content, using kebab-case (e.g., "example-code-snippet"). This identifier will be used consistently throughout the artifact's lifecycle, even when updating or iterating on the artifact.
+  3. Include a \`title\` attribute to provide a brief title or description of the content.
+  4. Add a \`type\` attribute to specify the type of content the artifact represents. Assign one of the following values to the \`type\` attribute:
+    - HTML: "text/html"
+      - The user interface can render single file HTML pages placed within the artifact tags. HTML, JS, and CSS should be in a single file when using the \`text/html\` type.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - The only place external scripts can be imported from is https://cdnjs.cloudflare.com
+    - SVG: "image/svg+xml"
+      - The user interface will render the Scalable Vector Graphics (SVG) image within the artifact tags.
+      - The assistant should specify the viewbox of the SVG rather than defining a width/height
+    - Mermaid Diagrams: "application/vnd.mermaid"
+      - The user interface will render Mermaid diagrams placed within the artifact tags.
+    - React Components: "application/vnd.react"
+      - Use this for displaying either: React elements, e.g. \`<strong>Hello World!</strong>\`, React pure functional components, e.g. \`() => <strong>Hello World!</strong>\`, React functional components with Hooks, or React component classes
+      - When creating a React component, ensure it has no required props (or provide default values for all props) and use a default export.
+      - Use Tailwind classes for styling. DO NOT USE ARBITRARY VALUES (e.g. \`h-[600px]\`).
+      - Base React is available to be imported. To use hooks, first import it at the top of the artifact, e.g. \`import { useState } from "react"\`
+      - The lucide-react@0.394.0 library is available to be imported. e.g. \`import { Camera } from "lucide-react"\` & \`<Camera color="red" size={48} />\`
+      - The recharts charting library is available to be imported, e.g. \`import { LineChart, XAxis, ... } from "recharts"\` & \`<LineChart ...><XAxis dataKey="name"> ...\`
+      - The three.js library is available to be imported, e.g. \`import * as THREE from "three";\`
+      - The date-fns library is available to be imported, e.g. \`import { compareAsc, format } from "date-fns";\`
+      - The react-day-picker library is available to be imported, e.g. \`import { DayPicker } from "react-day-picker";\`
+      - The assistant can use prebuilt components from the \`shadcn/ui\` library after it is imported: \`import { Alert, AlertDescription, AlertTitle, AlertDialog, AlertDialogAction } from '/components/ui/alert';\`. If using components from the shadcn/ui library, the assistant mentions this to the user and offers to help them install the components if necessary.
+      - Components MUST be imported from \`/components/ui/name\` and NOT from \`/components/name\` or \`@/components/ui/name\`.
+      - NO OTHER LIBRARIES (e.g. zod, hookform) ARE INSTALLED OR ABLE TO BE IMPORTED.
+      - Images from the web are not allowed, but you can use placeholder images by specifying the width and height like so \`<img src="/api/placeholder/400/320" alt="placeholder" />\`
+      - When iterating on code, ensure that the code is complete and functional without any snippets, placeholders, or ellipses.
+      - If you are unable to follow the above requirements for any reason, don't use artifacts and use regular code blocks instead, which will not attempt to render the component.
+  5. Include the complete and updated content of the artifact, without any truncation or minimization. Don't use "// rest of the code remains the same...".
+  6. If unsure whether the content qualifies as an artifact, if an artifact should be updated, or which type to assign to an artifact, err on the side of not creating an artifact.
+  7. NEVER use triple backticks to enclose the artifact, ONLY the content within the artifact.
+
+Here are some examples of correct usage of artifacts:
+
+## Examples
+
+### Example 1
+
+    This example demonstrates how to create a Mermaid artifact for a simple flow chart.
+
+    User: Can you create a simple flow chart showing the process of making tea using Mermaid?
+
+    Assistant: Sure! Here's a simple flow chart depicting the process of making tea using Mermaid syntax:
+
+      :::artifact{identifier="tea-making-flowchart" type="application/vnd.mermaid" title="Flow chart: Making Tea"}
+      \`\`\`mermaid
+      graph TD
+          A[Start] --> B{Water boiled?}
+          B -->|Yes| C[Add tea leaves to cup]
+          B -->|No| D[Boil water]
+          D --> B
+          C --> E[Pour boiling water into cup]
+          E --> F[Steep tea for desired time]
+          F --> G[Remove tea leaves]
+          G --> H[Add milk or sugar, if desired]
+          H --> I[Enjoy your tea!]
+          I --> J[End]
+      \`\`\`
+      :::
+
+      This flow chart uses Mermaid syntax to visualize the steps involved in making a cup of tea. Here's a brief explanation of the process:
+
+      1. Start
+      2. Check if water is boiled
+      3. If not boiled, boil the water
+      4. Once water is boiled, add tea leaves to the cup
+      5. Pour boiling water into the cup
+      6. Steep the tea for the desired time
+      7. Remove the tea leaves
+      8. Optionally add milk or sugar
+      9. Enjoy your tea!
+      10. End
+
+      This chart provides a clear visual representation of the tea-making process. You can easily modify or expand this chart if you want to add more details or steps to the process. Let me know if you'd like any changes or have any questions!
+
+---
+
+### Example 2
+
+    User: Create a simple React counter component
+    
+    Assistant: Here's a simple React counter component:
+
+      :::artifact{identifier="react-counter" type="application/vnd.react" title="React Counter"}
+      \`\`\`
+      import { useState } from 'react';
+
+      export default function Counter() {
+        const [count, setCount] = useState(0);
+        return (
+          <div className="p-4">
+            <p className="mb-2">Count: {count}</p>
+            <button className="bg-blue-500 text-white px-4 py-2 rounded" onClick={() => setCount(count + 1)}>
+              Increment
+            </button>
+          </div>
+        );
+      }
+      \`\`\`
+      :::
+
+      This component creates a simple counter with an increment button.
+
+---
+
+### Example 3
+    User: Create a basic HTML structure for a blog post
+    Assistant: Here's a basic HTML structure for a blog post:
+
+      :::artifact{identifier="blog-post-html" type="text/html" title="Blog Post HTML"}
+      \`\`\`
+      <!DOCTYPE html>
+      <html lang="en">
+      <head>
+        <meta charset="UTF-8">
+        <meta name="viewport" content="width=device-width, initial-scale=1.0">
+        <title>My Blog Post</title>
+        <style>
+          body { font-family: Arial, sans-serif; line-height: 1.6; max-width: 800px; margin: 0 auto; padding: 20px; }
+          h1 { color: #333; }
+          p { margin-bottom: 15px; }
+        </style>
+      </head>
+      <body>
+        <header>
+          <h1>My First Blog Post</h1>
+        </header>
+        <main>
+          <article>
+            <p>This is the content of my blog post. It's short and sweet!</p>
+          </article>
+        </main>
+        <footer>
+          <p>&copy; 2023 My Blog</p>
+        </footer>
+      </body>
+      </html>
+      \`\`\`
+      :::
+
+      This HTML structure provides a simple layout for a blog post.
+
+---`;
+
+/**
+ *
+ * @param {Object} params
+ * @param {EModelEndpoint | string} params.endpoint - The current endpoint
+ * @param {ArtifactModes} params.artifacts - The current artifact mode
+ * @returns
+ */
+const generateArtifactsPrompt = ({ endpoint, artifacts }) => {
+  if (artifacts === ArtifactModes.CUSTOM) {
+    return null;
+  }
+
+  let prompt = artifactsPrompt;
+  if (endpoint !== EModelEndpoint.anthropic) {
+    prompt = artifactsOpenAIPrompt;
+  }
+
+  if (artifacts === ArtifactModes.SHADCNUI) {
+    prompt += generateShadcnPrompt({ components, useXML: endpoint === EModelEndpoint.anthropic });
+  }
+
+  return prompt;
+};
+
+module.exports = generateArtifactsPrompt;

api/app/clients/prompts/formatMessages.js

@@ -1,4 +1,5 @@
-const { EModelEndpoint } = require('librechat-data-provider');
+const { ToolMessage } = require('@langchain/core/messages');
+const { EModelEndpoint, ContentTypes } = require('librechat-data-provider');
 const { HumanMessage, AIMessage, SystemMessage } = require('langchain/schema');
 
 /**
@@ -14,11 +15,11 @@ const { HumanMessage, AIMessage, SystemMessage } = require('langchain/schema');
  */
 const formatVisionMessage = ({ message, image_urls, endpoint }) => {
   if (endpoint === EModelEndpoint.anthropic) {
-    message.content = [...image_urls, { type: 'text', text: message.content }];
+    message.content = [...image_urls, { type: ContentTypes.TEXT, text: message.content }];
     return message;
   }
 
-  message.content = [{ type: 'text', text: message.content }, ...image_urls];
+  message.content = [{ type: ContentTypes.TEXT, text: message.content }, ...image_urls];
 
   return message;
 };
@@ -51,7 +52,7 @@ const formatMessage = ({ message, userName, assistantName, endpoint, langChain =
     _role = roleMapping[lc_id[2]];
   }
   const role = _role ?? (sender && sender?.toLowerCase() === 'user' ? 'user' : 'assistant');
-  const content = text ?? _content ?? '';
+  const content = _content ?? text ?? '';
   const formattedMessage = {
     role,
     content,
@@ -131,4 +132,79 @@ const formatFromLangChain = (message) => {
   };
 };
 
-module.exports = { formatMessage, formatLangChainMessages, formatFromLangChain };
+/**
+ * Formats an array of messages for LangChain, handling tool calls and creating ToolMessage instances.
+ *
+ * @param {Array<Partial<TMessage>>} payload - The array of messages to format.
+ * @returns {Array<(HumanMessage|AIMessage|SystemMessage|ToolMessage)>} - The array of formatted LangChain messages, including ToolMessages for tool calls.
+ */
+const formatAgentMessages = (payload) => {
+  const messages = [];
+
+  for (const message of payload) {
+    if (message.role !== 'assistant') {
+      messages.push(formatMessage({ message, langChain: true }));
+      continue;
+    }
+
+    let currentContent = [];
+    let lastAIMessage = null;
+
+    for (const part of message.content) {
+      if (part.type === ContentTypes.TEXT && part.tool_call_ids) {
+        // If there's pending content, add it as an AIMessage
+        if (currentContent.length > 0) {
+          messages.push(new AIMessage({ content: currentContent }));
+          currentContent = [];
+        }
+
+        // Create a new AIMessage with this text and prepare for tool calls
+        lastAIMessage = new AIMessage({
+          content: part.text || '',
+        });
+
+        messages.push(lastAIMessage);
+      } else if (part.type === ContentTypes.TOOL_CALL) {
+        if (!lastAIMessage) {
+          throw new Error('Invalid tool call structure: No preceding AIMessage with tool_call_ids');
+        }
+
+        // Note: `tool_calls` list is defined when constructed by `AIMessage` class, and outputs should be excluded from it
+        const { output, args: _args, ...tool_call } = part.tool_call;
+        // TODO: investigate; args as dictionary may need to be provider-or-tool-specific
+        let args = _args;
+        try {
+          args = JSON.parse(args);
+        } catch (e) {
+          // failed to parse, leave as is
+        }
+        tool_call.args = args;
+        lastAIMessage.tool_calls.push(tool_call);
+
+        // Add the corresponding ToolMessage
+        messages.push(
+          new ToolMessage({
+            tool_call_id: tool_call.id,
+            name: tool_call.name,
+            content: output,
+          }),
+        );
+      } else {
+        currentContent.push(part);
+      }
+    }
+
+    if (currentContent.length > 0) {
+      messages.push(new AIMessage({ content: currentContent }));
+    }
+  }
+
+  return messages;
+};
+
+module.exports = {
+  formatMessage,
+  formatFromLangChain,
+  formatAgentMessages,
+  formatLangChainMessages,
+};

api/app/clients/prompts/index.js

@@ -1,3 +1,4 @@
+const addCacheControl = require('./addCacheControl');
 const formatMessages = require('./formatMessages');
 const summaryPrompts = require('./summaryPrompts');
 const handleInputs = require('./handleInputs');
@@ -8,6 +9,7 @@ const createVisionPrompt = require('./createVisionPrompt');
 const createContextHandlers = require('./createContextHandlers');
 
 module.exports = {
+  addCacheControl,
   ...formatMessages,
   ...summaryPrompts,
   ...handleInputs,

api/app/clients/prompts/shadcn-docs/components.js

@@ -0,0 +1,495 @@
+// Essential Components
+const essentialComponents = {
+  avatar: {
+    componentName: 'Avatar',
+    importDocs: 'import { Avatar, AvatarFallback, AvatarImage } from "/components/ui/avatar"',
+    usageDocs: `
+<Avatar>
+  <AvatarImage src="https://github.com/shadcn.png" />
+  <AvatarFallback>CN</AvatarFallback>
+</Avatar>`,
+  },
+  button: {
+    componentName: 'Button',
+    importDocs: 'import { Button } from "/components/ui/button"',
+    usageDocs: `
+<Button variant="outline">Button</Button>`,
+  },
+  card: {
+    componentName: 'Card',
+    importDocs: `
+import {
+  Card,
+  CardContent,
+  CardDescription,
+  CardFooter,
+  CardHeader,
+  CardTitle,
+} from "/components/ui/card"`,
+    usageDocs: `
+<Card>
+  <CardHeader>
+    <CardTitle>Card Title</CardTitle>
+    <CardDescription>Card Description</CardDescription>
+  </CardHeader>
+  <CardContent>
+    <p>Card Content</p>
+  </CardContent>
+  <CardFooter>
+    <p>Card Footer</p>
+  </CardFooter>
+</Card>`,
+  },
+  checkbox: {
+    componentName: 'Checkbox',
+    importDocs: 'import { Checkbox } from "/components/ui/checkbox"',
+    usageDocs: '<Checkbox />',
+  },
+  input: {
+    componentName: 'Input',
+    importDocs: 'import { Input } from "/components/ui/input"',
+    usageDocs: '<Input />',
+  },
+  label: {
+    componentName: 'Label',
+    importDocs: 'import { Label } from "/components/ui/label"',
+    usageDocs: '<Label htmlFor="email">Your email address</Label>',
+  },
+  radioGroup: {
+    componentName: 'RadioGroup',
+    importDocs: `
+import { Label } from "/components/ui/label"
+import { RadioGroup, RadioGroupItem } from "/components/ui/radio-group"`,
+    usageDocs: `
+<RadioGroup defaultValue="option-one">
+  <div className="flex items-center space-x-2">
+    <RadioGroupItem value="option-one" id="option-one" />
+    <Label htmlFor="option-one">Option One</Label>
+  </div>
+  <div className="flex items-center space-x-2">
+    <RadioGroupItem value="option-two" id="option-two" />
+    <Label htmlFor="option-two">Option Two</Label>
+  </div>
+</RadioGroup>`,
+  },
+  select: {
+    componentName: 'Select',
+    importDocs: `
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "/components/ui/select"`,
+    usageDocs: `
+<Select>
+  <SelectTrigger className="w-[180px]">
+    <SelectValue placeholder="Theme" />
+  </SelectTrigger>
+  <SelectContent>
+    <SelectItem value="light">Light</SelectItem>
+    <SelectItem value="dark">Dark</SelectItem>
+    <SelectItem value="system">System</SelectItem>
+  </SelectContent>
+</Select>`,
+  },
+  textarea: {
+    componentName: 'Textarea',
+    importDocs: 'import { Textarea } from "/components/ui/textarea"',
+    usageDocs: '<Textarea />',
+  },
+};
+
+// Extra Components
+const extraComponents = {
+  accordion: {
+    componentName: 'Accordion',
+    importDocs: `
+import {
+  Accordion,
+  AccordionContent,
+  AccordionItem,
+  AccordionTrigger,
+} from "/components/ui/accordion"`,
+    usageDocs: `
+<Accordion type="single" collapsible>
+  <AccordionItem value="item-1">
+    <AccordionTrigger>Is it accessible?</AccordionTrigger>
+    <AccordionContent>
+      Yes. It adheres to the WAI-ARIA design pattern.
+    </AccordionContent>
+  </AccordionItem>
+</Accordion>`,
+  },
+  alertDialog: {
+    componentName: 'AlertDialog',
+    importDocs: `
+import {
+  AlertDialog,
+  AlertDialogAction,
+  AlertDialogCancel,
+  AlertDialogContent,
+  AlertDialogDescription,
+  AlertDialogFooter,
+  AlertDialogHeader,
+  AlertDialogTitle,
+  AlertDialogTrigger,
+} from "/components/ui/alert-dialog"`,
+    usageDocs: `
+<AlertDialog>
+  <AlertDialogTrigger>Open</AlertDialogTrigger>
+  <AlertDialogContent>
+    <AlertDialogHeader>
+      <AlertDialogTitle>Are you absolutely sure?</AlertDialogTitle>
+      <AlertDialogDescription>
+        This action cannot be undone.
+      </AlertDialogDescription>
+    </AlertDialogHeader>
+    <AlertDialogFooter>
+      <AlertDialogCancel>Cancel</AlertDialogCancel>
+      <AlertDialogAction>Continue</AlertDialogAction>
+    </AlertDialogFooter>
+  </AlertDialogContent>
+</AlertDialog>`,
+  },
+  alert: {
+    componentName: 'Alert',
+    importDocs: `
+import {
+  Alert,
+  AlertDescription,
+  AlertTitle,
+} from "/components/ui/alert"`,
+    usageDocs: `
+<Alert>
+  <AlertTitle>Heads up!</AlertTitle>
+  <AlertDescription>
+    You can add components to your app using the cli.
+  </AlertDescription>
+</Alert>`,
+  },
+  aspectRatio: {
+    componentName: 'AspectRatio',
+    importDocs: 'import { AspectRatio } from "/components/ui/aspect-ratio"',
+    usageDocs: `
+<AspectRatio ratio={16 / 9}>
+  <Image src="..." alt="Image" className="rounded-md object-cover" />
+</AspectRatio>`,
+  },
+  badge: {
+    componentName: 'Badge',
+    importDocs: 'import { Badge } from "/components/ui/badge"',
+    usageDocs: '<Badge>Badge</Badge>',
+  },
+  calendar: {
+    componentName: 'Calendar',
+    importDocs: 'import { Calendar } from "/components/ui/calendar"',
+    usageDocs: '<Calendar />',
+  },
+  carousel: {
+    componentName: 'Carousel',
+    importDocs: `
+import {
+  Carousel,
+  CarouselContent,
+  CarouselItem,
+  CarouselNext,
+  CarouselPrevious,
+} from "/components/ui/carousel"`,
+    usageDocs: `
+<Carousel>
+  <CarouselContent>
+    <CarouselItem>...</CarouselItem>
+    <CarouselItem>...</CarouselItem>
+    <CarouselItem>...</CarouselItem>
+  </CarouselContent>
+  <CarouselPrevious />
+  <CarouselNext />
+</Carousel>`,
+  },
+  collapsible: {
+    componentName: 'Collapsible',
+    importDocs: `
+import {
+  Collapsible,
+  CollapsibleContent,
+  CollapsibleTrigger,
+} from "/components/ui/collapsible"`,
+    usageDocs: `
+<Collapsible>
+  <CollapsibleTrigger>Can I use this in my project?</CollapsibleTrigger>
+  <CollapsibleContent>
+    Yes. Free to use for personal and commercial projects. No attribution required.
+  </CollapsibleContent>
+</Collapsible>`,
+  },
+  dialog: {
+    componentName: 'Dialog',
+    importDocs: `
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogHeader,
+  DialogTitle,
+  DialogTrigger,
+} from "/components/ui/dialog"`,
+    usageDocs: `
+<Dialog>
+  <DialogTrigger>Open</DialogTrigger>
+  <DialogContent>
+    <DialogHeader>
+      <DialogTitle>Are you sure absolutely sure?</DialogTitle>
+      <DialogDescription>
+        This action cannot be undone.
+      </DialogDescription>
+    </DialogHeader>
+  </DialogContent>
+</Dialog>`,
+  },
+  dropdownMenu: {
+    componentName: 'DropdownMenu',
+    importDocs: `
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuLabel,
+  DropdownMenuSeparator,
+  DropdownMenuTrigger,
+} from "/components/ui/dropdown-menu"`,
+    usageDocs: `
+<DropdownMenu>
+  <DropdownMenuTrigger>Open</DropdownMenuTrigger>
+  <DropdownMenuContent>
+    <DropdownMenuLabel>My Account</DropdownMenuLabel>
+    <DropdownMenuSeparator />
+    <DropdownMenuItem>Profile</DropdownMenuItem>
+    <DropdownMenuItem>Billing</DropdownMenuItem>
+    <DropdownMenuItem>Team</DropdownMenuItem>
+    <DropdownMenuItem>Subscription</DropdownMenuItem>
+  </DropdownMenuContent>
+</DropdownMenu>`,
+  },
+  menubar: {
+    componentName: 'Menubar',
+    importDocs: `
+import {
+  Menubar,
+  MenubarContent,
+  MenubarItem,
+  MenubarMenu,
+  MenubarSeparator,
+  MenubarShortcut,
+  MenubarTrigger,
+} from "/components/ui/menubar"`,
+    usageDocs: `
+<Menubar>
+  <MenubarMenu>
+    <MenubarTrigger>File</MenubarTrigger>
+    <MenubarContent>
+      <MenubarItem>
+        New Tab <MenubarShortcut>⌘T</MenubarShortcut>
+      </MenubarItem>
+      <MenubarItem>New Window</MenubarItem>
+      <MenubarSeparator />
+      <MenubarItem>Share</MenubarItem>
+      <MenubarSeparator />
+      <MenubarItem>Print</MenubarItem>
+    </MenubarContent>
+  </MenubarMenu>
+</Menubar>`,
+  },
+  navigationMenu: {
+    componentName: 'NavigationMenu',
+    importDocs: `
+import {
+  NavigationMenu,
+  NavigationMenuContent,
+  NavigationMenuItem,
+  NavigationMenuLink,
+  NavigationMenuList,
+  NavigationMenuTrigger,
+  navigationMenuTriggerStyle,
+} from "/components/ui/navigation-menu"`,
+    usageDocs: `
+<NavigationMenu>
+  <NavigationMenuList>
+    <NavigationMenuItem>
+      <NavigationMenuTrigger>Item One</NavigationMenuTrigger>
+      <NavigationMenuContent>
+        <NavigationMenuLink>Link</NavigationMenuLink>
+      </NavigationMenuContent>
+    </NavigationMenuItem>
+  </NavigationMenuList>
+</NavigationMenu>`,
+  },
+  popover: {
+    componentName: 'Popover',
+    importDocs: `
+import {
+  Popover,
+  PopoverContent,
+  PopoverTrigger,
+} from "/components/ui/popover"`,
+    usageDocs: `
+<Popover>
+  <PopoverTrigger>Open</PopoverTrigger>
+  <PopoverContent>Place content for the popover here.</PopoverContent>
+</Popover>`,
+  },
+  progress: {
+    componentName: 'Progress',
+    importDocs: 'import { Progress } from "/components/ui/progress"',
+    usageDocs: '<Progress value={33} />',
+  },
+  separator: {
+    componentName: 'Separator',
+    importDocs: 'import { Separator } from "/components/ui/separator"',
+    usageDocs: '<Separator />',
+  },
+  sheet: {
+    componentName: 'Sheet',
+    importDocs: `
+import {
+  Sheet,
+  SheetContent,
+  SheetDescription,
+  SheetHeader,
+  SheetTitle,
+  SheetTrigger,
+} from "/components/ui/sheet"`,
+    usageDocs: `
+<Sheet>
+  <SheetTrigger>Open</SheetTrigger>
+  <SheetContent>
+    <SheetHeader>
+      <SheetTitle>Are you sure absolutely sure?</SheetTitle>
+      <SheetDescription>
+        This action cannot be undone.
+      </SheetDescription>
+    </SheetHeader>
+  </SheetContent>
+</Sheet>`,
+  },
+  skeleton: {
+    componentName: 'Skeleton',
+    importDocs: 'import { Skeleton } from "/components/ui/skeleton"',
+    usageDocs: '<Skeleton className="w-[100px] h-[20px] rounded-full" />',
+  },
+  slider: {
+    componentName: 'Slider',
+    importDocs: 'import { Slider } from "/components/ui/slider"',
+    usageDocs: '<Slider defaultValue={[33]} max={100} step={1} />',
+  },
+  switch: {
+    componentName: 'Switch',
+    importDocs: 'import { Switch } from "/components/ui/switch"',
+    usageDocs: '<Switch />',
+  },
+  table: {
+    componentName: 'Table',
+    importDocs: `
+import {
+  Table,
+  TableBody,
+  TableCaption,
+  TableCell,
+  TableHead,
+  TableHeader,
+  TableRow,
+} from "/components/ui/table"`,
+    usageDocs: `
+<Table>
+  <TableCaption>A list of your recent invoices.</TableCaption>
+  <TableHeader>
+    <TableRow>
+      <TableHead className="w-[100px]">Invoice</TableHead>
+      <TableHead>Status</TableHead>
+      <TableHead>Method</TableHead>
+      <TableHead className="text-right">Amount</TableHead>
+    </TableRow>
+  </TableHeader>
+  <TableBody>
+    <TableRow>
+      <TableCell className="font-medium">INV001</TableCell>
+      <TableCell>Paid</TableCell>
+      <TableCell>Credit Card</TableCell>
+      <TableCell className="text-right">$250.00</TableCell>
+    </TableRow>
+  </TableBody>
+</Table>`,
+  },
+  tabs: {
+    componentName: 'Tabs',
+    importDocs: `
+import {
+  Tabs,
+  TabsContent,
+  TabsList,
+  TabsTrigger,
+} from "/components/ui/tabs"`,
+    usageDocs: `
+<Tabs defaultValue="account" className="w-[400px]">
+  <TabsList>
+    <TabsTrigger value="account">Account</TabsTrigger>
+    <TabsTrigger value="password">Password</TabsTrigger>
+  </TabsList>
+  <TabsContent value="account">Make changes to your account here.</TabsContent>
+  <TabsContent value="password">Change your password here.</TabsContent>
+</Tabs>`,
+  },
+  toast: {
+    componentName: 'Toast',
+    importDocs: `
+import { useToast } from "/components/ui/use-toast"
+import { Button } from "/components/ui/button"`,
+    usageDocs: `
+export function ToastDemo() {
+  const { toast } = useToast()
+  return (
+    <Button
+      onClick={() => {
+        toast({
+          title: "Scheduled: Catch up",
+          description: "Friday, February 10, 2023 at 5:57 PM",
+        })
+      }}
+    >
+      Show Toast
+    </Button>
+  )
+}`,
+  },
+  toggle: {
+    componentName: 'Toggle',
+    importDocs: 'import { Toggle } from "/components/ui/toggle"',
+    usageDocs: '<Toggle>Toggle</Toggle>',
+  },
+  tooltip: {
+    componentName: 'Tooltip',
+    importDocs: `
+import {
+  Tooltip,
+  TooltipContent,
+  TooltipProvider,
+  TooltipTrigger,
+} from "/components/ui/tooltip"`,
+    usageDocs: `
+<TooltipProvider>
+  <Tooltip>
+    <TooltipTrigger>Hover</TooltipTrigger>
+    <TooltipContent>
+      <p>Add to library</p>
+    </TooltipContent>
+  </Tooltip>
+</TooltipProvider>`,
+  },
+};
+
+const components = Object.assign({}, essentialComponents, extraComponents);
+
+module.exports = {
+  components,
+};

api/app/clients/prompts/shadcn-docs/generate.js

@@ -0,0 +1,50 @@
+const dedent = require('dedent');
+
+/**
+ * Generate system prompt for AI-assisted React component creation
+ * @param {Object} options - Configuration options
+ * @param {Object} options.components - Documentation for shadcn components
+ * @param {boolean} [options.useXML=false] - Whether to use XML-style formatting for component instructions
+ * @returns {string} The generated system prompt
+ */
+function generateShadcnPrompt(options) {
+  const { components, useXML = false } = options;
+
+  let systemPrompt = dedent`
+    ## Additional Artifact Instructions for React Components: "application/vnd.react"
+    
+    There are some prestyled components (primitives) available for use. Please use your best judgement to use any of these components if the app calls for one.
+
+    Here are the components that are available, along with how to import them, and how to use them:
+
+    ${Object.values(components)
+    .map((component) => {
+      if (useXML) {
+        return dedent`
+            <component>
+              <name>${component.componentName}</name>
+              <import-instructions>${component.importDocs}</import-instructions>
+              <usage-instructions>${component.usageDocs}</usage-instructions>
+            </component>
+          `;
+      } else {
+        return dedent`
+            # ${component.componentName}
+
+            ## Import Instructions
+            ${component.importDocs}
+
+            ## Usage Instructions
+            ${component.usageDocs}
+          `;
+      }
+    })
+    .join('\n\n')}
+  `;
+
+  return systemPrompt;
+}
+
+module.exports = {
+  generateShadcnPrompt,
+};

api/app/clients/specs/AnthropicClient.test.js

@@ -1,4 +1,6 @@
-const AnthropicClient = require('../AnthropicClient');
+const { anthropicSettings } = require('librechat-data-provider');
+const AnthropicClient = require('~/app/clients/AnthropicClient');
+
 const HUMAN_PROMPT = '\n\nHuman:';
 const AI_PROMPT = '\n\nAssistant:';
 
@@ -22,7 +24,7 @@ describe('AnthropicClient', () => {
     const options = {
       modelOptions: {
         model,
-        temperature: 0.7,
+        temperature: anthropicSettings.temperature.default,
       },
     };
     client = new AnthropicClient('test-api-key');
@@ -33,7 +35,42 @@ describe('AnthropicClient', () => {
     it('should set the options correctly', () => {
       expect(client.apiKey).toBe('test-api-key');
       expect(client.modelOptions.model).toBe(model);
-      expect(client.modelOptions.temperature).toBe(0.7);
+      expect(client.modelOptions.temperature).toBe(anthropicSettings.temperature.default);
+    });
+
+    it('should set legacy maxOutputTokens for non-Claude-3 models', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-2',
+          maxOutputTokens: anthropicSettings.maxOutputTokens.default,
+        },
+      });
+      expect(client.modelOptions.maxOutputTokens).toBe(
+        anthropicSettings.legacy.maxOutputTokens.default,
+      );
+    });
+    it('should not set maxOutputTokens if not provided', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-3',
+        },
+      });
+      expect(client.modelOptions.maxOutputTokens).toBeUndefined();
+    });
+
+    it('should not set legacy maxOutputTokens for Claude-3 models', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-3-opus-20240229',
+          maxOutputTokens: anthropicSettings.legacy.maxOutputTokens.default,
+        },
+      });
+      expect(client.modelOptions.maxOutputTokens).toBe(
+        anthropicSettings.legacy.maxOutputTokens.default,
+      );
     });
   });
 
@@ -136,4 +173,212 @@ describe('AnthropicClient', () => {
       expect(prompt).toContain('You are Claude-2');
     });
   });
+
+  describe('getClient', () => {
+    it('should set legacy maxOutputTokens for non-Claude-3 models', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-2',
+          maxOutputTokens: anthropicSettings.legacy.maxOutputTokens.default,
+        },
+      });
+      expect(client.modelOptions.maxOutputTokens).toBe(
+        anthropicSettings.legacy.maxOutputTokens.default,
+      );
+    });
+
+    it('should not set legacy maxOutputTokens for Claude-3 models', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-3-opus-20240229',
+          maxOutputTokens: anthropicSettings.legacy.maxOutputTokens.default,
+        },
+      });
+      expect(client.modelOptions.maxOutputTokens).toBe(
+        anthropicSettings.legacy.maxOutputTokens.default,
+      );
+    });
+
+    it('should add beta header for claude-3-5-sonnet model', () => {
+      const client = new AnthropicClient('test-api-key');
+      const modelOptions = {
+        model: 'claude-3-5-sonnet-20240307',
+      };
+      client.setOptions({ modelOptions, promptCache: true });
+      const anthropicClient = client.getClient(modelOptions);
+      expect(anthropicClient._options.defaultHeaders).toBeDefined();
+      expect(anthropicClient._options.defaultHeaders).toHaveProperty('anthropic-beta');
+      expect(anthropicClient._options.defaultHeaders['anthropic-beta']).toBe(
+        'max-tokens-3-5-sonnet-2024-07-15,prompt-caching-2024-07-31',
+      );
+    });
+
+    it('should add beta header for claude-3-haiku model', () => {
+      const client = new AnthropicClient('test-api-key');
+      const modelOptions = {
+        model: 'claude-3-haiku-2028',
+      };
+      client.setOptions({ modelOptions, promptCache: true });
+      const anthropicClient = client.getClient(modelOptions);
+      expect(anthropicClient._options.defaultHeaders).toBeDefined();
+      expect(anthropicClient._options.defaultHeaders).toHaveProperty('anthropic-beta');
+      expect(anthropicClient._options.defaultHeaders['anthropic-beta']).toBe(
+        'prompt-caching-2024-07-31',
+      );
+    });
+
+    it('should not add beta header for other models', () => {
+      const client = new AnthropicClient('test-api-key');
+      client.setOptions({
+        modelOptions: {
+          model: 'claude-2',
+        },
+      });
+      const anthropicClient = client.getClient();
+      expect(anthropicClient.defaultHeaders).not.toHaveProperty('anthropic-beta');
+    });
+  });
+
+  describe('calculateCurrentTokenCount', () => {
+    let client;
+
+    beforeEach(() => {
+      client = new AnthropicClient('test-api-key');
+    });
+
+    it('should calculate correct token count when usage is provided', () => {
+      const tokenCountMap = {
+        msg1: 10,
+        msg2: 20,
+        currentMsg: 30,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 70,
+        output_tokens: 50,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(40); // 70 - (10 + 20) = 40
+    });
+
+    it('should return original estimate if calculation results in negative value', () => {
+      const tokenCountMap = {
+        msg1: 40,
+        msg2: 50,
+        currentMsg: 30,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 80,
+        output_tokens: 50,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(30); // Original estimate
+    });
+
+    it('should handle cache creation and read input tokens', () => {
+      const tokenCountMap = {
+        msg1: 10,
+        msg2: 20,
+        currentMsg: 30,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 50,
+        cache_creation_input_tokens: 10,
+        cache_read_input_tokens: 20,
+        output_tokens: 40,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(50); // (50 + 10 + 20) - (10 + 20) = 50
+    });
+
+    it('should handle missing usage properties', () => {
+      const tokenCountMap = {
+        msg1: 10,
+        msg2: 20,
+        currentMsg: 30,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        output_tokens: 40,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(30); // Original estimate
+    });
+
+    it('should handle empty tokenCountMap', () => {
+      const tokenCountMap = {};
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 50,
+        output_tokens: 40,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(50);
+      expect(Number.isNaN(result)).toBe(false);
+    });
+
+    it('should handle zero values in usage', () => {
+      const tokenCountMap = {
+        msg1: 10,
+        currentMsg: 20,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 0,
+        cache_creation_input_tokens: 0,
+        cache_read_input_tokens: 0,
+        output_tokens: 0,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(20); // Should return original estimate
+      expect(Number.isNaN(result)).toBe(false);
+    });
+
+    it('should handle undefined usage', () => {
+      const tokenCountMap = {
+        msg1: 10,
+        currentMsg: 20,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = undefined;
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(20); // Should return original estimate
+      expect(Number.isNaN(result)).toBe(false);
+    });
+
+    it('should handle non-numeric values in tokenCountMap', () => {
+      const tokenCountMap = {
+        msg1: 'ten',
+        currentMsg: 20,
+      };
+      const currentMessageId = 'currentMsg';
+      const usage = {
+        input_tokens: 30,
+        output_tokens: 10,
+      };
+
+      const result = client.calculateCurrentTokenCount({ tokenCountMap, currentMessageId, usage });
+
+      expect(result).toBe(30); // Should return 30 (input_tokens) - 0 (ignored 'ten') = 30
+      expect(Number.isNaN(result)).toBe(false);
+    });
+  });
 });

api/app/clients/specs/BaseClient.test.js

@@ -1,7 +1,7 @@
 const { Constants } = require('librechat-data-provider');
 const { initializeFakeClient } = require('./FakeClient');
 
-jest.mock('../../../lib/db/connectDb');
+jest.mock('~/lib/db/connectDb');
 jest.mock('~/models', () => ({
   User: jest.fn(),
   Key: jest.fn(),
@@ -565,11 +565,13 @@ describe('BaseClient', () => {
       const getReqData = jest.fn();
       const opts = { getReqData };
       const response = await TestClient.sendMessage('Hello, world!', opts);
-      expect(getReqData).toHaveBeenCalledWith({
-        userMessage: expect.objectContaining({ text: 'Hello, world!' }),
-        conversationId: response.conversationId,
-        responseMessageId: response.messageId,
-      });
+      expect(getReqData).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userMessage: expect.objectContaining({ text: 'Hello, world!' }),
+          conversationId: response.conversationId,
+          responseMessageId: response.messageId,
+        }),
+      );
     });
 
     test('onStart is called with the correct arguments', async () => {
@@ -631,5 +633,32 @@ describe('BaseClient', () => {
         }),
       );
     });
+
+    test('userMessagePromise is awaited before saving response message', async () => {
+      // Mock the saveMessageToDatabase method
+      TestClient.saveMessageToDatabase = jest.fn().mockImplementation(() => {
+        return new Promise((resolve) => setTimeout(resolve, 100)); // Simulate a delay
+      });
+
+      // Send a message
+      const messagePromise = TestClient.sendMessage('Hello, world!');
+
+      // Wait a short time to ensure the user message save has started
+      await new Promise((resolve) => setTimeout(resolve, 50));
+
+      // Check that saveMessageToDatabase has been called once (for the user message)
+      expect(TestClient.saveMessageToDatabase).toHaveBeenCalledTimes(1);
+
+      // Wait for the message to be fully processed
+      await messagePromise;
+
+      // Check that saveMessageToDatabase has been called twice (once for user message, once for response)
+      expect(TestClient.saveMessageToDatabase).toHaveBeenCalledTimes(2);
+
+      // Check the order of calls
+      const calls = TestClient.saveMessageToDatabase.mock.calls;
+      expect(calls[0][0].isCreatedByUser).toBe(true); // First call should be for user message
+      expect(calls[1][0].isCreatedByUser).toBe(false); // Second call should be for response message
+    });
   });
 });

api/app/clients/specs/OpenAIClient.tokens.js

@@ -38,7 +38,12 @@ const run = async () => {
   "On the other hand, we denounce with righteous indignation and dislike men who are so beguiled and demoralized by the charms of pleasure of the moment, so blinded by desire, that they cannot foresee the pain and trouble that are bound to ensue; and equal blame belongs to those who fail in their duty through weakness of will, which is the same as saying through shrinking from toil and pain. These cases are perfectly simple and easy to distinguish. In a free hour, when our power of choice is untrammelled and when nothing prevents our being able to do what we like best, every pleasure is to be welcomed and every pain avoided. But in certain circumstances and owing to the claims of duty or the obligations of business it will frequently occur that pleasures have to be repudiated and annoyances accepted. The wise man therefore always holds in these matters to this principle of selection: he rejects pleasures to secure other greater pleasures, or else he endures pains to avoid worse pains."
   `;
   const model = 'gpt-3.5-turbo';
-  const maxContextTokens = model === 'gpt-4' ? 8191 : model === 'gpt-4-32k' ? 32767 : 4095; // 1 less than maximum
+  let maxContextTokens = 4095;
+  if (model === 'gpt-4') {
+    maxContextTokens = 8191;
+  } else if (model === 'gpt-4-32k') {
+    maxContextTokens = 32767;
+  }
   const clientOptions = {
     reverseProxyUrl: process.env.OPENAI_REVERSE_PROXY || null,
     maxContextTokens,

api/app/clients/tools/structured/GoogleSearch.js

@@ -12,9 +12,15 @@ class GoogleSearchResults extends Tool {
     this.envVarApiKey = 'GOOGLE_SEARCH_API_KEY';
     this.envVarSearchEngineId = 'GOOGLE_CSE_ID';
     this.override = fields.override ?? false;
-    this.apiKey = fields.apiKey ?? getEnvironmentVariable(this.envVarApiKey);
+    this.apiKey = fields[this.envVarApiKey] ?? getEnvironmentVariable(this.envVarApiKey);
     this.searchEngineId =
-      fields.searchEngineId ?? getEnvironmentVariable(this.envVarSearchEngineId);
+      fields[this.envVarSearchEngineId] ?? getEnvironmentVariable(this.envVarSearchEngineId);
+
+    if (!this.override && (!this.apiKey || !this.searchEngineId)) {
+      throw new Error(
+        `Missing ${this.envVarApiKey} or ${this.envVarSearchEngineId} environment variable.`,
+      );
+    }
 
     this.kwargs = fields?.kwargs ?? {};
     this.name = 'google';

api/app/clients/tools/structured/TavilySearch.js

@@ -0,0 +1,78 @@
+const { z } = require('zod');
+const { tool } = require('@langchain/core/tools');
+const { getEnvironmentVariable } = require('@langchain/core/utils/env');
+
+function createTavilySearchTool(fields = {}) {
+  const envVar = 'TAVILY_API_KEY';
+  const override = fields.override ?? false;
+  const apiKey = fields.apiKey ?? getApiKey(envVar, override);
+  const kwargs = fields?.kwargs ?? {};
+
+  function getApiKey(envVar, override) {
+    const key = getEnvironmentVariable(envVar);
+    if (!key && !override) {
+      throw new Error(`Missing ${envVar} environment variable.`);
+    }
+    return key;
+  }
+
+  return tool(
+    async (input) => {
+      const { query, ...rest } = input;
+
+      const requestBody = {
+        api_key: apiKey,
+        query,
+        ...rest,
+        ...kwargs,
+      };
+
+      const response = await fetch('https://api.tavily.com/search', {
+        method: 'POST',
+        headers: {
+          'Content-Type': 'application/json',
+        },
+        body: JSON.stringify(requestBody),
+      });
+
+      const json = await response.json();
+      if (!response.ok) {
+        throw new Error(`Request failed with status ${response.status}: ${json.error}`);
+      }
+
+      return JSON.stringify(json);
+    },
+    {
+      name: 'tavily_search_results_json',
+      description:
+        'A search engine optimized for comprehensive, accurate, and trusted results. Useful for when you need to answer questions about current events.',
+      schema: z.object({
+        query: z.string().min(1).describe('The search query string.'),
+        max_results: z
+          .number()
+          .min(1)
+          .max(10)
+          .optional()
+          .describe('The maximum number of search results to return. Defaults to 5.'),
+        search_depth: z
+          .enum(['basic', 'advanced'])
+          .optional()
+          .describe(
+            'The depth of the search, affecting result quality and response time (`basic` or `advanced`). Default is basic for quick results and advanced for indepth high quality results but longer response time. Advanced calls equals 2 requests.',
+          ),
+        include_images: z
+          .boolean()
+          .optional()
+          .describe(
+            'Whether to include a list of query-related images in the response. Default is False.',
+          ),
+        include_answer: z
+          .boolean()
+          .optional()
+          .describe('Whether to include answers in the search results. Default is False.'),
+      }),
+    },
+  );
+}
+
+module.exports = createTavilySearchTool;

api/app/clients/tools/structured/TavilySearchResults.js

@@ -12,7 +12,7 @@ class TavilySearchResults extends Tool {
     this.envVar = 'TAVILY_API_KEY';
     /* Used to initialize the Tool without necessary variables. */
     this.override = fields.override ?? false;
-    this.apiKey = fields.apiKey ?? this.getApiKey();
+    this.apiKey = fields[this.envVar] ?? this.getApiKey();
 
     this.kwargs = fields?.kwargs ?? {};
     this.name = 'tavily_search_results_json';
@@ -82,7 +82,9 @@ class TavilySearchResults extends Tool {
 
     const json = await response.json();
     if (!response.ok) {
-      throw new Error(`Request failed with status ${response.status}: ${json.error}`);
+      throw new Error(
+        `Request failed with status ${response.status}: ${json?.detail?.error || json?.error}`,
+      );
     }
 
     return JSON.stringify(json);

api/app/clients/tools/structured/specs/GoogleSearch.spec.js

@@ -0,0 +1,50 @@
+const GoogleSearch = require('../GoogleSearch');
+
+jest.mock('node-fetch');
+jest.mock('@langchain/core/utils/env');
+
+describe('GoogleSearch', () => {
+  let originalEnv;
+  const mockApiKey = 'mock_api';
+  const mockSearchEngineId = 'mock_search_engine_id';
+
+  beforeAll(() => {
+    originalEnv = { ...process.env };
+  });
+
+  beforeEach(() => {
+    jest.resetModules();
+    process.env = {
+      ...originalEnv,
+      GOOGLE_SEARCH_API_KEY: mockApiKey,
+      GOOGLE_CSE_ID: mockSearchEngineId,
+    };
+  });
+
+  afterEach(() => {
+    jest.clearAllMocks();
+    process.env = originalEnv;
+  });
+
+  it('should use mockApiKey and mockSearchEngineId when environment variables are not set', () => {
+    const instance = new GoogleSearch({
+      GOOGLE_SEARCH_API_KEY: mockApiKey,
+      GOOGLE_CSE_ID: mockSearchEngineId,
+    });
+    expect(instance.apiKey).toBe(mockApiKey);
+    expect(instance.searchEngineId).toBe(mockSearchEngineId);
+  });
+
+  it('should throw an error if GOOGLE_SEARCH_API_KEY or GOOGLE_CSE_ID is missing', () => {
+    delete process.env.GOOGLE_SEARCH_API_KEY;
+    expect(() => new GoogleSearch()).toThrow(
+      'Missing GOOGLE_SEARCH_API_KEY or GOOGLE_CSE_ID environment variable.',
+    );
+
+    process.env.GOOGLE_SEARCH_API_KEY = mockApiKey;
+    delete process.env.GOOGLE_CSE_ID;
+    expect(() => new GoogleSearch()).toThrow(
+      'Missing GOOGLE_SEARCH_API_KEY or GOOGLE_CSE_ID environment variable.',
+    );
+  });
+});

api/app/clients/tools/structured/specs/TavilySearchResults.spec.js

@@ -0,0 +1,38 @@
+const TavilySearchResults = require('../TavilySearchResults');
+
+jest.mock('node-fetch');
+jest.mock('@langchain/core/utils/env');
+
+describe('TavilySearchResults', () => {
+  let originalEnv;
+  const mockApiKey = 'mock_api_key';
+
+  beforeAll(() => {
+    originalEnv = { ...process.env };
+  });
+
+  beforeEach(() => {
+    jest.resetModules();
+    process.env = {
+      ...originalEnv,
+      TAVILY_API_KEY: mockApiKey,
+    };
+  });
+
+  afterEach(() => {
+    jest.clearAllMocks();
+    process.env = originalEnv;
+  });
+
+  it('should throw an error if TAVILY_API_KEY is missing', () => {
+    delete process.env.TAVILY_API_KEY;
+    expect(() => new TavilySearchResults()).toThrow('Missing TAVILY_API_KEY environment variable.');
+  });
+
+  it('should use mockApiKey when TAVILY_API_KEY is not set in the environment', () => {
+    const instance = new TavilySearchResults({
+      TAVILY_API_KEY: mockApiKey,
+    });
+    expect(instance.apiKey).toBe(mockApiKey);
+  });
+});

api/cache/clearPendingReq.js

@@ -35,7 +35,7 @@ const clearPendingReq = async ({ userId, cache: _cache }) => {
     return;
   }
 
-  const key = `${USE_REDIS ? namespace : ''}:${userId ?? ''}`;
+  const key = `${isEnabled(USE_REDIS) ? namespace : ''}:${userId ?? ''}`;
   const currentReq = +((await cache.get(key)) ?? 0);
 
   if (currentReq && currentReq >= 1) {

api/cache/getLogStores.js

@@ -1,13 +1,11 @@
 const Keyv = require('keyv');
-const { CacheKeys, ViolationTypes } = require('librechat-data-provider');
+const { CacheKeys, ViolationTypes, Time } = require('librechat-data-provider');
 const { logFile, violationFile } = require('./keyvFiles');
 const { math, isEnabled } = require('~/server/utils');
 const keyvRedis = require('./keyvRedis');
 const keyvMongo = require('./keyvMongo');
 
 const { BAN_DURATION, USE_REDIS } = process.env ?? {};
-const THIRTY_MINUTES = 1800000;
-const TEN_MINUTES = 600000;
 
 const duration = math(BAN_DURATION, 7200000);
 
@@ -29,17 +27,21 @@ const roles = isEnabled(USE_REDIS)
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: CacheKeys.ROLES });
 
-const audioRuns = isEnabled(USE_REDIS) // ttl: 30 minutes
-  ? new Keyv({ store: keyvRedis, ttl: TEN_MINUTES })
-  : new Keyv({ namespace: CacheKeys.AUDIO_RUNS, ttl: TEN_MINUTES });
+const audioRuns = isEnabled(USE_REDIS)
+  ? new Keyv({ store: keyvRedis, ttl: Time.TEN_MINUTES })
+  : new Keyv({ namespace: CacheKeys.AUDIO_RUNS, ttl: Time.TEN_MINUTES });
 
-const tokenConfig = isEnabled(USE_REDIS) // ttl: 30 minutes
-  ? new Keyv({ store: keyvRedis, ttl: THIRTY_MINUTES })
-  : new Keyv({ namespace: CacheKeys.TOKEN_CONFIG, ttl: THIRTY_MINUTES });
+const messages = isEnabled(USE_REDIS)
+  ? new Keyv({ store: keyvRedis, ttl: Time.FIVE_MINUTES })
+  : new Keyv({ namespace: CacheKeys.MESSAGES, ttl: Time.FIVE_MINUTES });
 
-const genTitle = isEnabled(USE_REDIS) // ttl: 2 minutes
-  ? new Keyv({ store: keyvRedis, ttl: 120000 })
-  : new Keyv({ namespace: CacheKeys.GEN_TITLE, ttl: 120000 });
+const tokenConfig = isEnabled(USE_REDIS)
+  ? new Keyv({ store: keyvRedis, ttl: Time.THIRTY_MINUTES })
+  : new Keyv({ namespace: CacheKeys.TOKEN_CONFIG, ttl: Time.THIRTY_MINUTES });
+
+const genTitle = isEnabled(USE_REDIS)
+  ? new Keyv({ store: keyvRedis, ttl: Time.TWO_MINUTES })
+  : new Keyv({ namespace: CacheKeys.GEN_TITLE, ttl: Time.TWO_MINUTES });
 
 const modelQueries = isEnabled(process.env.USE_REDIS)
   ? new Keyv({ store: keyvRedis })
@@ -47,7 +49,7 @@ const modelQueries = isEnabled(process.env.USE_REDIS)
 
 const abortKeys = isEnabled(USE_REDIS)
   ? new Keyv({ store: keyvRedis })
-  : new Keyv({ namespace: CacheKeys.ABORT_KEYS, ttl: 600000 });
+  : new Keyv({ namespace: CacheKeys.ABORT_KEYS, ttl: Time.TEN_MINUTES });
 
 const namespaces = {
   [CacheKeys.ROLES]: roles,
@@ -67,6 +69,7 @@ const namespaces = {
   registrations: createViolationInstance('registrations'),
   [ViolationTypes.TTS_LIMIT]: createViolationInstance(ViolationTypes.TTS_LIMIT),
   [ViolationTypes.STT_LIMIT]: createViolationInstance(ViolationTypes.STT_LIMIT),
+  [ViolationTypes.CONVO_ACCESS]: createViolationInstance(ViolationTypes.CONVO_ACCESS),
   [ViolationTypes.FILE_UPLOAD_LIMIT]: createViolationInstance(ViolationTypes.FILE_UPLOAD_LIMIT),
   [ViolationTypes.VERIFY_EMAIL_LIMIT]: createViolationInstance(ViolationTypes.VERIFY_EMAIL_LIMIT),
   [ViolationTypes.RESET_PASSWORD_LIMIT]: createViolationInstance(
@@ -81,6 +84,7 @@ const namespaces = {
   [CacheKeys.GEN_TITLE]: genTitle,
   [CacheKeys.MODEL_QUERIES]: modelQueries,
   [CacheKeys.AUDIO_RUNS]: audioRuns,
+  [CacheKeys.MESSAGES]: messages,
 };
 
 /**

api/config/parsers.js

@@ -109,6 +109,14 @@ const condenseArray = (item) => {
  * @returns {string} - The formatted log message.
  */
 const debugTraverse = winston.format.printf(({ level, message, timestamp, ...metadata }) => {
+  if (!message) {
+    return `${timestamp} ${level}`;
+  }
+
+  if (!message?.trim || typeof message !== 'string') {
+    return `${timestamp} ${level}: ${JSON.stringify(message)}`;
+  }
+
   let msg = `${timestamp} ${level}: ${truncateLongStrings(message?.trim(), 150)}`;
   try {
     if (level !== 'debug') {

api/models/Action.js

@@ -5,17 +5,16 @@ const Action = mongoose.model('action', actionSchema);
 
 /**
  * Update an action with new data without overwriting existing properties,
- * or create a new action if it doesn't exist, within a transaction session if provided.
+ * or create a new action if it doesn't exist.
  *
  * @param {Object} searchParams - The search parameters to find the action to update.
  * @param {string} searchParams.action_id - The ID of the action to update.
  * @param {string} searchParams.user - The user ID of the action's author.
  * @param {Object} updateData - An object containing the properties to update.
- * @param {mongoose.ClientSession} [session] - The transaction session to use.
- * @returns {Promise<Object>} The updated or newly created action document as a plain object.
+ * @returns {Promise<Action>} The updated or newly created action document as a plain object.
  */
-const updateAction = async (searchParams, updateData, session = null) => {
-  const options = { new: true, upsert: true, session };
+const updateAction = async (searchParams, updateData) => {
+  const options = { new: true, upsert: true };
   return await Action.findOneAndUpdate(searchParams, updateData, options).lean();
 };
 
@@ -24,7 +23,7 @@ const updateAction = async (searchParams, updateData, session = null) => {
  *
  * @param {Object} searchParams - The search parameters to find matching actions.
  * @param {boolean} includeSensitive - Flag to include sensitive data in the metadata.
- * @returns {Promise<Array<Object>>} A promise that resolves to an array of action documents as plain objects.
+ * @returns {Promise<Array<Action>>} A promise that resolves to an array of action documents as plain objects.
  */
 const getActions = async (searchParams, includeSensitive = false) => {
   const actions = await Action.find(searchParams).lean();
@@ -49,31 +48,27 @@ const getActions = async (searchParams, includeSensitive = false) => {
 };
 
 /**
- * Deletes an action by params, within a transaction session if provided.
+ * Deletes an action by params.
  *
  * @param {Object} searchParams - The search parameters to find the action to delete.
  * @param {string} searchParams.action_id - The ID of the action to delete.
  * @param {string} searchParams.user - The user ID of the action's author.
- * @param {mongoose.ClientSession} [session] - The transaction session to use (optional).
- * @returns {Promise<Object>} A promise that resolves to the deleted action document as a plain object, or null if no document was found.
+ * @returns {Promise<Action>} A promise that resolves to the deleted action document as a plain object, or null if no document was found.
  */
-const deleteAction = async (searchParams, session = null) => {
-  const options = session ? { session } : {};
-  return await Action.findOneAndDelete(searchParams, options).lean();
+const deleteAction = async (searchParams) => {
+  return await Action.findOneAndDelete(searchParams).lean();
 };
 
 /**
- * Deletes actions by params, within a transaction session if provided.
+ * Deletes actions by params.
  *
  * @param {Object} searchParams - The search parameters to find the actions to delete.
  * @param {string} searchParams.action_id - The ID of the action(s) to delete.
  * @param {string} searchParams.user - The user ID of the action's author.
- * @param {mongoose.ClientSession} [session] - The transaction session to use (optional).
  * @returns {Promise<Number>} A promise that resolves to the number of deleted action documents.
  */
-const deleteActions = async (searchParams, session = null) => {
-  const options = session ? { session } : {};
-  const result = await Action.deleteMany(searchParams, options);
+const deleteActions = async (searchParams) => {
+  const result = await Action.deleteMany(searchParams);
   return result.deletedCount;
 };
 

api/models/Agent.js

@@ -0,0 +1,142 @@
+const mongoose = require('mongoose');
+const { GLOBAL_PROJECT_NAME } = require('librechat-data-provider').Constants;
+const {
+  getProjectByName,
+  addAgentIdsToProject,
+  removeAgentIdsFromProject,
+  removeAgentFromAllProjects,
+} = require('./Project');
+const agentSchema = require('./schema/agent');
+
+const Agent = mongoose.model('agent', agentSchema);
+
+/**
+ * Create an agent with the provided data.
+ * @param {Object} agentData - The agent data to create.
+ * @returns {Promise<Agent>} The created agent document as a plain object.
+ * @throws {Error} If the agent creation fails.
+ */
+const createAgent = async (agentData) => {
+  return await Agent.create(agentData);
+};
+
+/**
+ * Get an agent document based on the provided ID.
+ *
+ * @param {Object} searchParameter - The search parameters to find the agent to update.
+ * @param {string} searchParameter.id - The ID of the agent to update.
+ * @param {string} searchParameter.author - The user ID of the agent's author.
+ * @returns {Promise<Agent|null>} The agent document as a plain object, or null if not found.
+ */
+const getAgent = async (searchParameter) => await Agent.findOne(searchParameter).lean();
+
+/**
+ * Update an agent with new data without overwriting existing
+ *  properties, or create a new agent if it doesn't exist.
+ *
+ * @param {Object} searchParameter - The search parameters to find the agent to update.
+ * @param {string} searchParameter.id - The ID of the agent to update.
+ * @param {string} [searchParameter.author] - The user ID of the agent's author.
+ * @param {Object} updateData - An object containing the properties to update.
+ * @returns {Promise<Agent>} The updated or newly created agent document as a plain object.
+ */
+const updateAgent = async (searchParameter, updateData) => {
+  const options = { new: true, upsert: true };
+  return await Agent.findOneAndUpdate(searchParameter, updateData, options).lean();
+};
+
+/**
+ * Deletes an agent based on the provided ID.
+ *
+ * @param {Object} searchParameter - The search parameters to find the agent to delete.
+ * @param {string} searchParameter.id - The ID of the agent to delete.
+ * @param {string} [searchParameter.author] - The user ID of the agent's author.
+ * @returns {Promise<void>} Resolves when the agent has been successfully deleted.
+ */
+const deleteAgent = async (searchParameter) => {
+  const agent = await Agent.findOneAndDelete(searchParameter);
+  if (agent) {
+    await removeAgentFromAllProjects(agent.id);
+  }
+  return agent;
+};
+
+/**
+ * Get all agents.
+ * @param {Object} searchParameter - The search parameters to find matching agents.
+ * @param {string} searchParameter.author - The user ID of the agent's author.
+ * @returns {Promise<Object>} A promise that resolves to an object containing the agents data and pagination info.
+ */
+const getListAgents = async (searchParameter) => {
+  const { author, ...otherParams } = searchParameter;
+
+  let query = Object.assign({ author }, otherParams);
+
+  const globalProject = await getProjectByName(GLOBAL_PROJECT_NAME, ['agentIds']);
+  if (globalProject && (globalProject.agentIds?.length ?? 0) > 0) {
+    const globalQuery = { id: { $in: globalProject.agentIds }, ...otherParams };
+    delete globalQuery.author;
+    query = { $or: [globalQuery, query] };
+  }
+
+  const agents = await Agent.find(query, {
+    id: 1,
+    name: 1,
+    avatar: 1,
+    projectIds: 1,
+  }).lean();
+
+  const hasMore = agents.length > 0;
+  const firstId = agents.length > 0 ? agents[0].id : null;
+  const lastId = agents.length > 0 ? agents[agents.length - 1].id : null;
+
+  return {
+    data: agents,
+    has_more: hasMore,
+    first_id: firstId,
+    last_id: lastId,
+  };
+};
+
+/**
+ * Updates the projects associated with an agent, adding and removing project IDs as specified.
+ * This function also updates the corresponding projects to include or exclude the agent ID.
+ *
+ * @param {string} agentId - The ID of the agent to update.
+ * @param {string[]} [projectIds] - Array of project IDs to add to the agent.
+ * @param {string[]} [removeProjectIds] - Array of project IDs to remove from the agent.
+ * @returns {Promise<MongoAgent>} The updated agent document.
+ * @throws {Error} If there's an error updating the agent or projects.
+ */
+const updateAgentProjects = async (agentId, projectIds, removeProjectIds) => {
+  const updateOps = {};
+
+  if (removeProjectIds && removeProjectIds.length > 0) {
+    for (const projectId of removeProjectIds) {
+      await removeAgentIdsFromProject(projectId, [agentId]);
+    }
+    updateOps.$pull = { projectIds: { $in: removeProjectIds } };
+  }
+
+  if (projectIds && projectIds.length > 0) {
+    for (const projectId of projectIds) {
+      await addAgentIdsToProject(projectId, [agentId]);
+    }
+    updateOps.$addToSet = { projectIds: { $each: projectIds } };
+  }
+
+  if (Object.keys(updateOps).length === 0) {
+    return await getAgent({ id: agentId });
+  }
+
+  return await updateAgent({ id: agentId }, updateOps);
+};
+
+module.exports = {
+  createAgent,
+  getAgent,
+  updateAgent,
+  deleteAgent,
+  getListAgents,
+  updateAgentProjects,
+};

api/models/Assistant.js

@@ -5,17 +5,16 @@ const Assistant = mongoose.model('assistant', assistantSchema);
 
 /**
  * Update an assistant with new data without overwriting existing properties,
- * or create a new assistant if it doesn't exist, within a transaction session if provided.
+ * or create a new assistant if it doesn't exist.
  *
  * @param {Object} searchParams - The search parameters to find the assistant to update.
  * @param {string} searchParams.assistant_id - The ID of the assistant to update.
  * @param {string} searchParams.user - The user ID of the assistant's author.
  * @param {Object} updateData - An object containing the properties to update.
- * @param {mongoose.ClientSession} [session] - The transaction session to use (optional).
- * @returns {Promise<Object>} The updated or newly created assistant document as a plain object.
+ * @returns {Promise<AssistantDocument>} The updated or newly created assistant document as a plain object.
  */
-const updateAssistantDoc = async (searchParams, updateData, session = null) => {
-  const options = { new: true, upsert: true, session };
+const updateAssistantDoc = async (searchParams, updateData) => {
+  const options = { new: true, upsert: true };
   return await Assistant.findOneAndUpdate(searchParams, updateData, options).lean();
 };
 
@@ -25,7 +24,7 @@ const updateAssistantDoc = async (searchParams, updateData, session = null) => {
  * @param {Object} searchParams - The search parameters to find the assistant to update.
  * @param {string} searchParams.assistant_id - The ID of the assistant to update.
  * @param {string} searchParams.user - The user ID of the assistant's author.
- * @returns {Promise<Object|null>} The assistant document as a plain object, or null if not found.
+ * @returns {Promise<AssistantDocument|null>} The assistant document as a plain object, or null if not found.
  */
 const getAssistant = async (searchParams) => await Assistant.findOne(searchParams).lean();
 
@@ -33,10 +32,17 @@ const getAssistant = async (searchParams) => await Assistant.findOne(searchParam
  * Retrieves all assistants that match the given search parameters.
  *
  * @param {Object} searchParams - The search parameters to find matching assistants.
- * @returns {Promise<Array<Object>>} A promise that resolves to an array of action documents as plain objects.
+ * @param {Object} [select] - Optional. Specifies which document fields to include or exclude.
+ * @returns {Promise<Array<AssistantDocument>>} A promise that resolves to an array of assistant documents as plain objects.
  */
-const getAssistants = async (searchParams) => {
-  return await Assistant.find(searchParams).lean();
+const getAssistants = async (searchParams, select = null) => {
+  let query = Assistant.find(searchParams);
+
+  if (select) {
+    query = query.select(select);
+  }
+
+  return await query.lean();
 };
 
 /**

api/models/Conversation.js

@@ -2,6 +2,20 @@ const Conversation = require('./schema/convoSchema');
 const { getMessages, deleteMessages } = require('./Message');
 const logger = require('~/config/winston');
 
+/**
+ * Searches for a conversation by conversationId and returns a lean document with only conversationId and user.
+ * @param {string} conversationId - The conversation's ID.
+ * @returns {Promise<{conversationId: string, user: string} | null>} The conversation object with selected fields or null if not found.
+ */
+const searchConversation = async (conversationId) => {
+  try {
+    return await Conversation.findOne({ conversationId }, 'conversationId user').lean();
+  } catch (error) {
+    logger.error('[searchConversation] Error searching conversation', error);
+    throw new Error('Error searching conversation');
+  }
+};
+
 /**
  * Retrieves a single conversation for a given user and conversation ID.
  * @param {string} user - The user's ID.
@@ -19,22 +33,40 @@ const getConvo = async (user, conversationId) => {
 
 module.exports = {
   Conversation,
-  saveConvo: async (user, { conversationId, newConversationId, ...convo }) => {
+  searchConversation,
+  /**
+   * Saves a conversation to the database.
+   * @param {Object} req - The request object.
+   * @param {string} conversationId - The conversation's ID.
+   * @param {Object} metadata - Additional metadata to log for operation.
+   * @returns {Promise<TConversation>} The conversation object.
+   */
+  saveConvo: async (req, { conversationId, newConversationId, ...convo }, metadata) => {
     try {
+      if (metadata && metadata?.context) {
+        logger.debug(`[saveConvo] ${metadata.context}`);
+      }
       const messages = await getMessages({ conversationId }, '_id');
-      const update = { ...convo, messages, user };
+      const update = { ...convo, messages, user: req.user.id };
       if (newConversationId) {
         update.conversationId = newConversationId;
       }
 
-      const conversation = await Conversation.findOneAndUpdate({ conversationId, user }, update, {
-        new: true,
-        upsert: true,
-      });
+      const conversation = await Conversation.findOneAndUpdate(
+        { conversationId, user: req.user.id },
+        update,
+        {
+          new: true,
+          upsert: true,
+        },
+      );
 
       return conversation.toObject();
     } catch (error) {
       logger.error('[saveConvo] Error saving conversation', error);
+      if (metadata && metadata?.context) {
+        logger.info(`[saveConvo] ${metadata.context}`);
+      }
       return { message: 'Error saving conversation' };
     }
   },
@@ -56,13 +88,16 @@ module.exports = {
       throw new Error('Failed to save conversations in bulk.');
     }
   },
-  getConvosByPage: async (user, pageNumber = 1, pageSize = 25, isArchived = false) => {
+  getConvosByPage: async (user, pageNumber = 1, pageSize = 25, isArchived = false, tags) => {
     const query = { user };
     if (isArchived) {
       query.isArchived = true;
     } else {
       query.$or = [{ isArchived: false }, { isArchived: { $exists: false } }];
     }
+    if (Array.isArray(tags) && tags.length > 0) {
+      query.tags = { $in: tags };
+    }
     try {
       const totalConvos = (await Conversation.countDocuments(query)) || 1;
       const totalPages = Math.ceil(totalConvos / pageSize);

api/models/ConversationTag.js

@@ -0,0 +1,249 @@
+const ConversationTag = require('./schema/conversationTagSchema');
+const Conversation = require('./schema/convoSchema');
+const logger = require('~/config/winston');
+
+/**
+ * Retrieves all conversation tags for a user.
+ * @param {string} user - The user ID.
+ * @returns {Promise<Array>} An array of conversation tags.
+ */
+const getConversationTags = async (user) => {
+  try {
+    return await ConversationTag.find({ user }).sort({ position: 1 }).lean();
+  } catch (error) {
+    logger.error('[getConversationTags] Error getting conversation tags', error);
+    throw new Error('Error getting conversation tags');
+  }
+};
+
+/**
+ * Creates a new conversation tag.
+ * @param {string} user - The user ID.
+ * @param {Object} data - The tag data.
+ * @param {string} data.tag - The tag name.
+ * @param {string} [data.description] - The tag description.
+ * @param {boolean} [data.addToConversation] - Whether to add the tag to a conversation.
+ * @param {string} [data.conversationId] - The conversation ID to add the tag to.
+ * @returns {Promise<Object>} The created tag.
+ */
+const createConversationTag = async (user, data) => {
+  try {
+    const { tag, description, addToConversation, conversationId } = data;
+
+    const existingTag = await ConversationTag.findOne({ user, tag }).lean();
+    if (existingTag) {
+      return existingTag;
+    }
+
+    const maxPosition = await ConversationTag.findOne({ user }).sort('-position').lean();
+    const position = (maxPosition?.position || 0) + 1;
+
+    const newTag = await ConversationTag.findOneAndUpdate(
+      { tag, user },
+      {
+        tag,
+        user,
+        count: addToConversation ? 1 : 0,
+        position,
+        description,
+        $setOnInsert: { createdAt: new Date() },
+      },
+      {
+        new: true,
+        upsert: true,
+        lean: true,
+      },
+    );
+
+    if (addToConversation && conversationId) {
+      await Conversation.findOneAndUpdate(
+        { user, conversationId },
+        { $addToSet: { tags: tag } },
+        { new: true },
+      );
+    }
+
+    return newTag;
+  } catch (error) {
+    logger.error('[createConversationTag] Error creating conversation tag', error);
+    throw new Error('Error creating conversation tag');
+  }
+};
+
+/**
+ * Updates an existing conversation tag.
+ * @param {string} user - The user ID.
+ * @param {string} oldTag - The current tag name.
+ * @param {Object} data - The updated tag data.
+ * @param {string} [data.tag] - The new tag name.
+ * @param {string} [data.description] - The updated description.
+ * @param {number} [data.position] - The new position.
+ * @returns {Promise<Object>} The updated tag.
+ */
+const updateConversationTag = async (user, oldTag, data) => {
+  try {
+    const { tag: newTag, description, position } = data;
+
+    const existingTag = await ConversationTag.findOne({ user, tag: oldTag }).lean();
+    if (!existingTag) {
+      return null;
+    }
+
+    if (newTag && newTag !== oldTag) {
+      const tagAlreadyExists = await ConversationTag.findOne({ user, tag: newTag }).lean();
+      if (tagAlreadyExists) {
+        throw new Error('Tag already exists');
+      }
+
+      await Conversation.updateMany({ user, tags: oldTag }, { $set: { 'tags.$': newTag } });
+    }
+
+    const updateData = {};
+    if (newTag) {
+      updateData.tag = newTag;
+    }
+    if (description !== undefined) {
+      updateData.description = description;
+    }
+    if (position !== undefined) {
+      await adjustPositions(user, existingTag.position, position);
+      updateData.position = position;
+    }
+
+    return await ConversationTag.findOneAndUpdate({ user, tag: oldTag }, updateData, {
+      new: true,
+      lean: true,
+    });
+  } catch (error) {
+    logger.error('[updateConversationTag] Error updating conversation tag', error);
+    throw new Error('Error updating conversation tag');
+  }
+};
+
+/**
+ * Adjusts positions of tags when a tag's position is changed.
+ * @param {string} user - The user ID.
+ * @param {number} oldPosition - The old position of the tag.
+ * @param {number} newPosition - The new position of the tag.
+ * @returns {Promise<void>}
+ */
+const adjustPositions = async (user, oldPosition, newPosition) => {
+  if (oldPosition === newPosition) {
+    return;
+  }
+
+  const update = oldPosition < newPosition ? { $inc: { position: -1 } } : { $inc: { position: 1 } };
+  const position =
+    oldPosition < newPosition
+      ? {
+        $gt: Math.min(oldPosition, newPosition),
+        $lte: Math.max(oldPosition, newPosition),
+      }
+      : {
+        $gte: Math.min(oldPosition, newPosition),
+        $lt: Math.max(oldPosition, newPosition),
+      };
+
+  await ConversationTag.updateMany(
+    {
+      user,
+      position,
+    },
+    update,
+  );
+};
+
+/**
+ * Deletes a conversation tag.
+ * @param {string} user - The user ID.
+ * @param {string} tag - The tag to delete.
+ * @returns {Promise<Object>} The deleted tag.
+ */
+const deleteConversationTag = async (user, tag) => {
+  try {
+    const deletedTag = await ConversationTag.findOneAndDelete({ user, tag }).lean();
+    if (!deletedTag) {
+      return null;
+    }
+
+    await Conversation.updateMany({ user, tags: tag }, { $pull: { tags: tag } });
+
+    await ConversationTag.updateMany(
+      { user, position: { $gt: deletedTag.position } },
+      { $inc: { position: -1 } },
+    );
+
+    return deletedTag;
+  } catch (error) {
+    logger.error('[deleteConversationTag] Error deleting conversation tag', error);
+    throw new Error('Error deleting conversation tag');
+  }
+};
+
+/**
+ * Updates tags for a specific conversation.
+ * @param {string} user - The user ID.
+ * @param {string} conversationId - The conversation ID.
+ * @param {string[]} tags - The new set of tags for the conversation.
+ * @returns {Promise<string[]>} The updated list of tags for the conversation.
+ */
+const updateTagsForConversation = async (user, conversationId, tags) => {
+  try {
+    const conversation = await Conversation.findOne({ user, conversationId }).lean();
+    if (!conversation) {
+      throw new Error('Conversation not found');
+    }
+
+    const oldTags = new Set(conversation.tags);
+    const newTags = new Set(tags);
+
+    const addedTags = [...newTags].filter((tag) => !oldTags.has(tag));
+    const removedTags = [...oldTags].filter((tag) => !newTags.has(tag));
+
+    const bulkOps = [];
+
+    for (const tag of addedTags) {
+      bulkOps.push({
+        updateOne: {
+          filter: { user, tag },
+          update: { $inc: { count: 1 } },
+          upsert: true,
+        },
+      });
+    }
+
+    for (const tag of removedTags) {
+      bulkOps.push({
+        updateOne: {
+          filter: { user, tag },
+          update: { $inc: { count: -1 } },
+        },
+      });
+    }
+
+    if (bulkOps.length > 0) {
+      await ConversationTag.bulkWrite(bulkOps);
+    }
+
+    const updatedConversation = (
+      await Conversation.findOneAndUpdate(
+        { user, conversationId },
+        { $set: { tags: [...newTags] } },
+        { new: true },
+      )
+    ).toObject();
+
+    return updatedConversation.tags;
+  } catch (error) {
+    logger.error('[updateTagsForConversation] Error updating tags', error);
+    throw new Error('Error updating tags for conversation');
+  }
+};
+
+module.exports = {
+  getConversationTags,
+  createConversationTag,
+  updateConversationTag,
+  deleteConversationTag,
+  updateTagsForConversation,
+};

api/models/Message.js

@@ -1,204 +1,294 @@
 const { z } = require('zod');
 const Message = require('./schema/messageSchema');
-const logger = require('~/config/winston');
+const { logger } = require('~/config');
 
 const idSchema = z.string().uuid();
 
+/**
+ * Saves a message in the database.
+ *
+ * @async
+ * @function saveMessage
+ * @param {Express.Request} req - The request object containing user information.
+ * @param {Object} params - The message data object.
+ * @param {string} params.endpoint - The endpoint where the message originated.
+ * @param {string} params.iconURL - The URL of the sender's icon.
+ * @param {string} params.messageId - The unique identifier for the message.
+ * @param {string} params.newMessageId - The new unique identifier for the message (if applicable).
+ * @param {string} params.conversationId - The identifier of the conversation.
+ * @param {string} [params.parentMessageId] - The identifier of the parent message, if any.
+ * @param {string} params.sender - The identifier of the sender.
+ * @param {string} params.text - The text content of the message.
+ * @param {boolean} params.isCreatedByUser - Indicates if the message was created by the user.
+ * @param {string} [params.error] - Any error associated with the message.
+ * @param {boolean} [params.unfinished] - Indicates if the message is unfinished.
+ * @param {Object[]} [params.files] - An array of files associated with the message.
+ * @param {boolean} [params.isEdited] - Indicates if the message was edited.
+ * @param {string} [params.finish_reason] - Reason for finishing the message.
+ * @param {number} [params.tokenCount] - The number of tokens in the message.
+ * @param {string} [params.plugin] - Plugin associated with the message.
+ * @param {string[]} [params.plugins] - An array of plugins associated with the message.
+ * @param {string} [params.model] - The model used to generate the message.
+ * @param {Object} [metadata] - Additional metadata for this operation
+ * @param {string} [metadata.context] - The context of the operation
+ * @returns {Promise<TMessage>} The updated or newly inserted message document.
+ * @throws {Error} If there is an error in saving the message.
+ */
+async function saveMessage(req, params, metadata) {
+  if (!req?.user?.id) {
+    throw new Error('User not authenticated');
+  }
+
+  const validConvoId = idSchema.safeParse(params.conversationId);
+  if (!validConvoId.success) {
+    logger.warn(`Invalid conversation ID: ${params.conversationId}`);
+    logger.info(`---\`saveMessage\` context: ${metadata?.context}`);
+    logger.info(`---Invalid conversation ID Params: ${JSON.stringify(params, null, 2)}`);
+    return;
+  }
+
+  try {
+    const update = {
+      ...params,
+      user: req.user.id,
+      messageId: params.newMessageId || params.messageId,
+    };
+    const message = await Message.findOneAndUpdate(
+      { messageId: params.messageId, user: req.user.id },
+      update,
+      { upsert: true, new: true },
+    );
+
+    return message.toObject();
+  } catch (err) {
+    logger.error('Error saving message:', err);
+    logger.info(`---\`saveMessage\` context: ${metadata?.context}`);
+    throw err;
+  }
+}
+
+/**
+ * Saves multiple messages in the database in bulk.
+ *
+ * @async
+ * @function bulkSaveMessages
+ * @param {Object[]} messages - An array of message objects to save.
+ * @returns {Promise<Object>} The result of the bulk write operation.
+ * @throws {Error} If there is an error in saving messages in bulk.
+ */
+async function bulkSaveMessages(messages) {
+  try {
+    const bulkOps = messages.map((message) => ({
+      updateOne: {
+        filter: { messageId: message.messageId },
+        update: message,
+        upsert: true,
+      },
+    }));
+
+    const result = await Message.bulkWrite(bulkOps);
+    return result;
+  } catch (err) {
+    logger.error('Error saving messages in bulk:', err);
+    throw err;
+  }
+}
+
+/**
+ * Records a message in the database.
+ *
+ * @async
+ * @function recordMessage
+ * @param {Object} params - The message data object.
+ * @param {string} params.user - The identifier of the user.
+ * @param {string} params.endpoint - The endpoint where the message originated.
+ * @param {string} params.messageId - The unique identifier for the message.
+ * @param {string} params.conversationId - The identifier of the conversation.
+ * @param {string} [params.parentMessageId] - The identifier of the parent message, if any.
+ * @param {Partial<TMessage>} rest - Any additional properties from the TMessage typedef not explicitly listed.
+ * @returns {Promise<Object>} The updated or newly inserted message document.
+ * @throws {Error} If there is an error in saving the message.
+ */
+async function recordMessage({
+  user,
+  endpoint,
+  messageId,
+  conversationId,
+  parentMessageId,
+  ...rest
+}) {
+  try {
+    // No parsing of convoId as may use threadId
+    const message = {
+      user,
+      endpoint,
+      messageId,
+      conversationId,
+      parentMessageId,
+      ...rest,
+    };
+
+    return await Message.findOneAndUpdate({ user, messageId }, message, {
+      upsert: true,
+      new: true,
+    });
+  } catch (err) {
+    logger.error('Error recording message:', err);
+    throw err;
+  }
+}
+
+/**
+ * Updates the text of a message.
+ *
+ * @async
+ * @function updateMessageText
+ * @param {Object} params - The update data object.
+ * @param {Object} req - The request object.
+ * @param {string} params.messageId - The unique identifier for the message.
+ * @param {string} params.text - The new text content of the message.
+ * @returns {Promise<void>}
+ * @throws {Error} If there is an error in updating the message text.
+ */
+async function updateMessageText(req, { messageId, text }) {
+  try {
+    await Message.updateOne({ messageId, user: req.user.id }, { text });
+  } catch (err) {
+    logger.error('Error updating message text:', err);
+    throw err;
+  }
+}
+
+/**
+ * Updates a message.
+ *
+ * @async
+ * @function updateMessage
+ * @param {Object} req - The request object.
+ * @param {Object} message - The message object containing update data.
+ * @param {string} message.messageId - The unique identifier for the message.
+ * @param {string} [message.text] - The new text content of the message.
+ * @param {Object[]} [message.files] - The files associated with the message.
+ * @param {boolean} [message.isCreatedByUser] - Indicates if the message was created by the user.
+ * @param {string} [message.sender] - The identifier of the sender.
+ * @param {number} [message.tokenCount] - The number of tokens in the message.
+ * @param {Object} [metadata] - The operation metadata
+ * @param {string} [metadata.context] - The operation metadata
+ * @returns {Promise<TMessage>} The updated message document.
+ * @throws {Error} If there is an error in updating the message or if the message is not found.
+ */
+async function updateMessage(req, message, metadata) {
+  try {
+    const { messageId, ...update } = message;
+    update.isEdited = true;
+    const updatedMessage = await Message.findOneAndUpdate(
+      { messageId, user: req.user.id },
+      update,
+      {
+        new: true,
+      },
+    );
+
+    if (!updatedMessage) {
+      throw new Error('Message not found or user not authorized.');
+    }
+
+    return {
+      messageId: updatedMessage.messageId,
+      conversationId: updatedMessage.conversationId,
+      parentMessageId: updatedMessage.parentMessageId,
+      sender: updatedMessage.sender,
+      text: updatedMessage.text,
+      isCreatedByUser: updatedMessage.isCreatedByUser,
+      tokenCount: updatedMessage.tokenCount,
+      isEdited: true,
+    };
+  } catch (err) {
+    logger.error('Error updating message:', err);
+    if (metadata && metadata?.context) {
+      logger.info(`---\`updateMessage\` context: ${metadata.context}`);
+    }
+    throw err;
+  }
+}
+
+/**
+ * Deletes messages in a conversation since a specific message.
+ *
+ * @async
+ * @function deleteMessagesSince
+ * @param {Object} params - The parameters object.
+ * @param {Object} req - The request object.
+ * @param {string} params.messageId - The unique identifier for the message.
+ * @param {string} params.conversationId - The identifier of the conversation.
+ * @returns {Promise<Number>} The number of deleted messages.
+ * @throws {Error} If there is an error in deleting messages.
+ */
+async function deleteMessagesSince(req, { messageId, conversationId }) {
+  try {
+    const message = await Message.findOne({ messageId, user: req.user.id }).lean();
+
+    if (message) {
+      const query = Message.find({ conversationId, user: req.user.id });
+      return await query.deleteMany({
+        createdAt: { $gt: message.createdAt },
+      });
+    }
+    return undefined;
+  } catch (err) {
+    logger.error('Error deleting messages:', err);
+    throw err;
+  }
+}
+
+/**
+ * Retrieves messages from the database.
+ * @async
+ * @function getMessages
+ * @param {Record<string, unknown>} filter - The filter criteria.
+ * @param {string | undefined} [select] - The fields to select.
+ * @returns {Promise<TMessage[]>} The messages that match the filter criteria.
+ * @throws {Error} If there is an error in retrieving messages.
+ */
+async function getMessages(filter, select) {
+  try {
+    if (select) {
+      return await Message.find(filter).select(select).sort({ createdAt: 1 }).lean();
+    }
+
+    return await Message.find(filter).sort({ createdAt: 1 }).lean();
+  } catch (err) {
+    logger.error('Error getting messages:', err);
+    throw err;
+  }
+}
+
+/**
+ * Deletes messages from the database.
+ *
+ * @async
+ * @function deleteMessages
+ * @param {Object} filter - The filter criteria to find messages to delete.
+ * @returns {Promise<Object>} The metadata with count of deleted messages.
+ * @throws {Error} If there is an error in deleting messages.
+ */
+async function deleteMessages(filter) {
+  try {
+    return await Message.deleteMany(filter);
+  } catch (err) {
+    logger.error('Error deleting messages:', err);
+    throw err;
+  }
+}
+
 module.exports = {
   Message,
-
-  async saveMessage({
-    user,
-    endpoint,
-    iconURL,
-    messageId,
-    newMessageId,
-    conversationId,
-    parentMessageId,
-    sender,
-    text,
-    isCreatedByUser,
-    error,
-    unfinished,
-    files,
-    isEdited,
-    finish_reason,
-    tokenCount,
-    plugin,
-    plugins,
-    model,
-  }) {
-    try {
-      const validConvoId = idSchema.safeParse(conversationId);
-      if (!validConvoId.success) {
-        return;
-      }
-
-      const update = {
-        user,
-        iconURL,
-        endpoint,
-        messageId: newMessageId || messageId,
-        conversationId,
-        parentMessageId,
-        sender,
-        text,
-        isCreatedByUser,
-        isEdited,
-        finish_reason,
-        error,
-        unfinished,
-        tokenCount,
-        plugin,
-        plugins,
-        model,
-      };
-
-      if (files) {
-        update.files = files;
-      }
-
-      const message = await Message.findOneAndUpdate({ messageId }, update, {
-        upsert: true,
-        new: true,
-      });
-
-      return message.toObject();
-    } catch (err) {
-      logger.error('Error saving message:', err);
-      throw new Error('Failed to save message.');
-    }
-  },
-
-  async bulkSaveMessages(messages) {
-    try {
-      const bulkOps = messages.map((message) => ({
-        updateOne: {
-          filter: { messageId: message.messageId },
-          update: message,
-          upsert: true,
-        },
-      }));
-
-      const result = await Message.bulkWrite(bulkOps);
-      return result;
-    } catch (err) {
-      logger.error('Error saving messages in bulk:', err);
-      throw new Error('Failed to save messages in bulk.');
-    }
-  },
-
-  /**
-   * Records a message in the database.
-   *
-   * @async
-   * @function recordMessage
-   * @param {Object} params - The message data object.
-   * @param {string} params.user - The identifier of the user.
-   * @param {string} params.endpoint - The endpoint where the message originated.
-   * @param {string} params.messageId - The unique identifier for the message.
-   * @param {string} params.conversationId - The identifier of the conversation.
-   * @param {string} [params.parentMessageId] - The identifier of the parent message, if any.
-   * @param {Partial<TMessage>} rest - Any additional properties from the TMessage typedef not explicitly listed.
-   * @returns {Promise<Object>} The updated or newly inserted message document.
-   * @throws {Error} If there is an error in saving the message.
-   */
-  async recordMessage({ user, endpoint, messageId, conversationId, parentMessageId, ...rest }) {
-    try {
-      // No parsing of convoId as may use threadId
-      const message = {
-        user,
-        endpoint,
-        messageId,
-        conversationId,
-        parentMessageId,
-        ...rest,
-      };
-
-      return await Message.findOneAndUpdate({ user, messageId }, message, {
-        upsert: true,
-        new: true,
-      });
-    } catch (err) {
-      logger.error('Error saving message:', err);
-      throw new Error('Failed to save message.');
-    }
-  },
-  async updateMessageText({ messageId, text }) {
-    try {
-      await Message.updateOne({ messageId }, { text });
-    } catch (err) {
-      logger.error('Error updating message text:', err);
-      throw new Error('Failed to update message text.');
-    }
-  },
-  async updateMessage(message) {
-    try {
-      const { messageId, ...update } = message;
-      update.isEdited = true;
-      const updatedMessage = await Message.findOneAndUpdate({ messageId }, update, {
-        new: true,
-      });
-
-      if (!updatedMessage) {
-        throw new Error('Message not found.');
-      }
-
-      return {
-        messageId: updatedMessage.messageId,
-        conversationId: updatedMessage.conversationId,
-        parentMessageId: updatedMessage.parentMessageId,
-        sender: updatedMessage.sender,
-        text: updatedMessage.text,
-        isCreatedByUser: updatedMessage.isCreatedByUser,
-        tokenCount: updatedMessage.tokenCount,
-        isEdited: true,
-      };
-    } catch (err) {
-      logger.error('Error updating message:', err);
-      throw new Error('Failed to update message.');
-    }
-  },
-  async deleteMessagesSince({ messageId, conversationId }) {
-    try {
-      const message = await Message.findOne({ messageId }).lean();
-
-      if (message) {
-        return await Message.find({ conversationId }).deleteMany({
-          createdAt: { $gt: message.createdAt },
-        });
-      }
-    } catch (err) {
-      logger.error('Error deleting messages:', err);
-      throw new Error('Failed to delete messages.');
-    }
-  },
-
-  /**
-   * Retrieves messages from the database.
-   * @param {Record<string, unknown>} filter
-   * @param {string | undefined} [select]
-   * @returns
-   */
-  async getMessages(filter, select) {
-    try {
-      if (select) {
-        return await Message.find(filter).select(select).sort({ createdAt: 1 }).lean();
-      }
-
-      return await Message.find(filter).sort({ createdAt: 1 }).lean();
-    } catch (err) {
-      logger.error('Error getting messages:', err);
-      throw new Error('Failed to get messages.');
-    }
-  },
-
-  async deleteMessages(filter) {
-    try {
-      return await Message.deleteMany(filter);
-    } catch (err) {
-      logger.error('Error deleting messages:', err);
-      throw new Error('Failed to delete messages.');
-    }
-  },
+  saveMessage,
+  bulkSaveMessages,
+  recordMessage,
+  updateMessageText,
+  updateMessage,
+  deleteMessagesSince,
+  getMessages,
+  deleteMessages,
 };

api/models/Message.spec.js

@@ -0,0 +1,239 @@
+const mongoose = require('mongoose');
+const { v4: uuidv4 } = require('uuid');
+
+jest.mock('mongoose');
+
+const mockFindQuery = {
+  select: jest.fn().mockReturnThis(),
+  sort: jest.fn().mockReturnThis(),
+  lean: jest.fn().mockReturnThis(),
+  deleteMany: jest.fn().mockResolvedValue({ deletedCount: 1 }),
+};
+
+const mockSchema = {
+  findOneAndUpdate: jest.fn(),
+  updateOne: jest.fn(),
+  findOne: jest.fn(() => ({
+    lean: jest.fn(),
+  })),
+  find: jest.fn(() => mockFindQuery),
+  deleteMany: jest.fn(),
+};
+
+mongoose.model.mockReturnValue(mockSchema);
+
+jest.mock('~/models/schema/messageSchema', () => mockSchema);
+
+jest.mock('~/config/winston', () => ({
+  error: jest.fn(),
+}));
+
+const {
+  saveMessage,
+  getMessages,
+  updateMessage,
+  deleteMessages,
+  updateMessageText,
+  deleteMessagesSince,
+} = require('~/models/Message');
+
+describe('Message Operations', () => {
+  let mockReq;
+  let mockMessage;
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+
+    mockReq = {
+      user: { id: 'user123' },
+    };
+
+    mockMessage = {
+      messageId: 'msg123',
+      conversationId: uuidv4(),
+      text: 'Hello, world!',
+      user: 'user123',
+    };
+
+    mockSchema.findOneAndUpdate.mockResolvedValue({
+      toObject: () => mockMessage,
+    });
+  });
+
+  describe('saveMessage', () => {
+    it('should save a message for an authenticated user', async () => {
+      const result = await saveMessage(mockReq, mockMessage);
+      expect(result).toEqual(mockMessage);
+      expect(mockSchema.findOneAndUpdate).toHaveBeenCalledWith(
+        { messageId: 'msg123', user: 'user123' },
+        expect.objectContaining({ user: 'user123' }),
+        expect.any(Object),
+      );
+    });
+
+    it('should throw an error for unauthenticated user', async () => {
+      mockReq.user = null;
+      await expect(saveMessage(mockReq, mockMessage)).rejects.toThrow('User not authenticated');
+    });
+
+    it('should throw an error for invalid conversation ID', async () => {
+      mockMessage.conversationId = 'invalid-id';
+      await expect(saveMessage(mockReq, mockMessage)).resolves.toBeUndefined();
+    });
+  });
+
+  describe('updateMessageText', () => {
+    it('should update message text for the authenticated user', async () => {
+      await updateMessageText(mockReq, { messageId: 'msg123', text: 'Updated text' });
+      expect(mockSchema.updateOne).toHaveBeenCalledWith(
+        { messageId: 'msg123', user: 'user123' },
+        { text: 'Updated text' },
+      );
+    });
+  });
+
+  describe('updateMessage', () => {
+    it('should update a message for the authenticated user', async () => {
+      mockSchema.findOneAndUpdate.mockResolvedValue(mockMessage);
+      const result = await updateMessage(mockReq, { messageId: 'msg123', text: 'Updated text' });
+      expect(result).toEqual(
+        expect.objectContaining({
+          messageId: 'msg123',
+          text: 'Hello, world!',
+          isEdited: true,
+        }),
+      );
+    });
+
+    it('should throw an error if message is not found', async () => {
+      mockSchema.findOneAndUpdate.mockResolvedValue(null);
+      await expect(
+        updateMessage(mockReq, { messageId: 'nonexistent', text: 'Test' }),
+      ).rejects.toThrow('Message not found or user not authorized.');
+    });
+  });
+
+  describe('deleteMessagesSince', () => {
+    it('should delete messages only for the authenticated user', async () => {
+      mockSchema.findOne().lean.mockResolvedValueOnce({ createdAt: new Date() });
+      mockFindQuery.deleteMany.mockResolvedValueOnce({ deletedCount: 1 });
+      const result = await deleteMessagesSince(mockReq, {
+        messageId: 'msg123',
+        conversationId: 'convo123',
+      });
+      expect(mockSchema.findOne).toHaveBeenCalledWith({ messageId: 'msg123', user: 'user123' });
+      expect(mockSchema.find).not.toHaveBeenCalled();
+      expect(result).toBeUndefined();
+    });
+
+    it('should return undefined if no message is found', async () => {
+      mockSchema.findOne().lean.mockResolvedValueOnce(null);
+      const result = await deleteMessagesSince(mockReq, {
+        messageId: 'nonexistent',
+        conversationId: 'convo123',
+      });
+      expect(result).toBeUndefined();
+    });
+  });
+
+  describe('getMessages', () => {
+    it('should retrieve messages with the correct filter', async () => {
+      const filter = { conversationId: 'convo123' };
+      await getMessages(filter);
+      expect(mockSchema.find).toHaveBeenCalledWith(filter);
+      expect(mockFindQuery.sort).toHaveBeenCalledWith({ createdAt: 1 });
+      expect(mockFindQuery.lean).toHaveBeenCalled();
+    });
+  });
+
+  describe('deleteMessages', () => {
+    it('should delete messages with the correct filter', async () => {
+      await deleteMessages({ user: 'user123' });
+      expect(mockSchema.deleteMany).toHaveBeenCalledWith({ user: 'user123' });
+    });
+  });
+
+  describe('Conversation Hijacking Prevention', () => {
+    it('should not allow editing a message in another user\'s conversation', async () => {
+      const attackerReq = { user: { id: 'attacker123' } };
+      const victimConversationId = 'victim-convo-123';
+      const victimMessageId = 'victim-msg-123';
+
+      mockSchema.findOneAndUpdate.mockResolvedValue(null);
+
+      await expect(
+        updateMessage(attackerReq, {
+          messageId: victimMessageId,
+          conversationId: victimConversationId,
+          text: 'Hacked message',
+        }),
+      ).rejects.toThrow('Message not found or user not authorized.');
+
+      expect(mockSchema.findOneAndUpdate).toHaveBeenCalledWith(
+        { messageId: victimMessageId, user: 'attacker123' },
+        expect.anything(),
+        expect.anything(),
+      );
+    });
+
+    it('should not allow deleting messages from another user\'s conversation', async () => {
+      const attackerReq = { user: { id: 'attacker123' } };
+      const victimConversationId = 'victim-convo-123';
+      const victimMessageId = 'victim-msg-123';
+
+      mockSchema.findOne().lean.mockResolvedValueOnce(null); // Simulating message not found for this user
+      const result = await deleteMessagesSince(attackerReq, {
+        messageId: victimMessageId,
+        conversationId: victimConversationId,
+      });
+
+      expect(result).toBeUndefined();
+      expect(mockSchema.findOne).toHaveBeenCalledWith({
+        messageId: victimMessageId,
+        user: 'attacker123',
+      });
+    });
+
+    it('should not allow inserting a new message into another user\'s conversation', async () => {
+      const attackerReq = { user: { id: 'attacker123' } };
+      const victimConversationId = uuidv4(); // Use a valid UUID
+
+      await expect(
+        saveMessage(attackerReq, {
+          conversationId: victimConversationId,
+          text: 'Inserted malicious message',
+          messageId: 'new-msg-123',
+        }),
+      ).resolves.not.toThrow(); // It should not throw an error
+
+      // Check that the message was saved with the attacker's user ID
+      expect(mockSchema.findOneAndUpdate).toHaveBeenCalledWith(
+        { messageId: 'new-msg-123', user: 'attacker123' },
+        expect.objectContaining({
+          user: 'attacker123',
+          conversationId: victimConversationId,
+        }),
+        expect.anything(),
+      );
+    });
+
+    it('should allow retrieving messages from any conversation', async () => {
+      const victimConversationId = 'victim-convo-123';
+
+      await getMessages({ conversationId: victimConversationId });
+
+      expect(mockSchema.find).toHaveBeenCalledWith({
+        conversationId: victimConversationId,
+      });
+
+      mockSchema.find.mockReturnValueOnce({
+        select: jest.fn().mockReturnThis(),
+        sort: jest.fn().mockReturnThis(),
+        lean: jest.fn().mockResolvedValue([{ text: 'Test message' }]),
+      });
+
+      const result = await getMessages({ conversationId: victimConversationId });
+      expect(result).toEqual([{ text: 'Test message' }]);
+    });
+  });
+});

api/models/Project.js

@@ -1,4 +1,5 @@
 const { model } = require('mongoose');
+const { GLOBAL_PROJECT_NAME } = require('librechat-data-provider').Constants;
 const projectSchema = require('~/models/schema/projectSchema');
 
 const Project = model('Project', projectSchema);
@@ -33,7 +34,7 @@ const getProjectByName = async function (projectName, fieldsToSelect = null) {
   const update = { $setOnInsert: { name: projectName } };
   const options = {
     new: true,
-    upsert: projectName === 'instance',
+    upsert: projectName === GLOBAL_PROJECT_NAME,
     lean: true,
     select: fieldsToSelect,
   };
@@ -81,10 +82,55 @@ const removeGroupFromAllProjects = async (promptGroupId) => {
   await Project.updateMany({}, { $pull: { promptGroupIds: promptGroupId } });
 };
 
+/**
+ * Add an array of agent IDs to a project's agentIds array, ensuring uniqueness.
+ *
+ * @param {string} projectId - The ID of the project to update.
+ * @param {string[]} agentIds - The array of agent IDs to add to the project.
+ * @returns {Promise<MongoProject>} The updated project document.
+ */
+const addAgentIdsToProject = async function (projectId, agentIds) {
+  return await Project.findByIdAndUpdate(
+    projectId,
+    { $addToSet: { agentIds: { $each: agentIds } } },
+    { new: true },
+  );
+};
+
+/**
+ * Remove an array of agent IDs from a project's agentIds array.
+ *
+ * @param {string} projectId - The ID of the project to update.
+ * @param {string[]} agentIds - The array of agent IDs to remove from the project.
+ * @returns {Promise<MongoProject>} The updated project document.
+ */
+const removeAgentIdsFromProject = async function (projectId, agentIds) {
+  return await Project.findByIdAndUpdate(
+    projectId,
+    { $pull: { agentIds: { $in: agentIds } } },
+    { new: true },
+  );
+};
+
+/**
+ * Remove an agent ID from all projects.
+ *
+ * @param {string} agentId - The ID of the agent to remove from projects.
+ * @returns {Promise<void>}
+ */
+const removeAgentFromAllProjects = async (agentId) => {
+  await Project.updateMany({}, { $pull: { agentIds: agentId } });
+};
+
 module.exports = {
   getProjectById,
   getProjectByName,
+  /* prompts */
   addGroupIdsToProject,
   removeGroupIdsFromProject,
   removeGroupFromAllProjects,
+  /* agents */
+  addAgentIdsToProject,
+  removeAgentIdsFromProject,
+  removeAgentFromAllProjects,
 };

api/models/Prompt.js

@@ -1,5 +1,5 @@
 const { ObjectId } = require('mongodb');
-const { SystemRoles, SystemCategories } = require('librechat-data-provider');
+const { SystemRoles, SystemCategories, Constants } = require('librechat-data-provider');
 const {
   getProjectByName,
   addGroupIdsToProject,
@@ -123,7 +123,7 @@ const getAllPromptGroups = async (req, filter) => {
     let combinedQuery = query;
 
     if (searchShared) {
-      const project = await getProjectByName('instance', 'promptGroupIds');
+      const project = await getProjectByName(Constants.GLOBAL_PROJECT_NAME, 'promptGroupIds');
       if (project && project.promptGroupIds.length > 0) {
         const projectQuery = { _id: { $in: project.promptGroupIds }, ...query };
         delete projectQuery.author;
@@ -177,7 +177,7 @@ const getPromptGroups = async (req, filter) => {
 
     if (searchShared) {
       // const projects = req.user.projects || []; // TODO: handle multiple projects
-      const project = await getProjectByName('instance', 'promptGroupIds');
+      const project = await getProjectByName(Constants.GLOBAL_PROJECT_NAME, 'promptGroupIds');
       if (project && project.promptGroupIds.length > 0) {
         const projectQuery = { _id: { $in: project.promptGroupIds }, ...query };
         delete projectQuery.author;

api/models/Role.js

@@ -1,6 +1,17 @@
-const { SystemRoles, CacheKeys, roleDefaults } = require('librechat-data-provider');
+const {
+  CacheKeys,
+  SystemRoles,
+  roleDefaults,
+  PermissionTypes,
+  removeNullishValues,
+  agentPermissionsSchema,
+  promptPermissionsSchema,
+  bookmarkPermissionsSchema,
+  multiConvoPermissionsSchema,
+} = require('librechat-data-provider');
 const getLogStores = require('~/cache/getLogStores');
 const Role = require('~/models/schema/roleSchema');
+const { logger } = require('~/config');
 
 /**
  * Retrieve a role by name and convert the found role document to a plain object.
@@ -61,9 +72,69 @@ const updateRoleByName = async function (roleName, updates) {
   }
 };
 
+const permissionSchemas = {
+  [PermissionTypes.AGENTS]: agentPermissionsSchema,
+  [PermissionTypes.PROMPTS]: promptPermissionsSchema,
+  [PermissionTypes.BOOKMARKS]: bookmarkPermissionsSchema,
+  [PermissionTypes.MULTI_CONVO]: multiConvoPermissionsSchema,
+};
+
+/**
+ * Updates access permissions for a specific role and multiple permission types.
+ * @param {SystemRoles} roleName - The role to update.
+ * @param {Object.<PermissionTypes, Object.<Permissions, boolean>>} permissionsUpdate - Permissions to update and their values.
+ */
+async function updateAccessPermissions(roleName, permissionsUpdate) {
+  const updates = {};
+  for (const [permissionType, permissions] of Object.entries(permissionsUpdate)) {
+    if (permissionSchemas[permissionType]) {
+      updates[permissionType] = removeNullishValues(permissions);
+    }
+  }
+
+  if (Object.keys(updates).length === 0) {
+    return;
+  }
+
+  try {
+    const role = await getRoleByName(roleName);
+    if (!role) {
+      return;
+    }
+
+    const updatedPermissions = {};
+    let hasChanges = false;
+
+    for (const [permissionType, permissions] of Object.entries(updates)) {
+      const currentPermissions = role[permissionType] || {};
+      updatedPermissions[permissionType] = { ...currentPermissions };
+
+      for (const [permission, value] of Object.entries(permissions)) {
+        if (currentPermissions[permission] !== value) {
+          updatedPermissions[permissionType][permission] = value;
+          hasChanges = true;
+          logger.info(
+            `Updating '${roleName}' role ${permissionType} '${permission}' permission from ${currentPermissions[permission]} to: ${value}`,
+          );
+        }
+      }
+    }
+
+    if (hasChanges) {
+      await updateRoleByName(roleName, updatedPermissions);
+      logger.info(`Updated '${roleName}' role permissions`);
+    } else {
+      logger.info(`No changes needed for '${roleName}' role permissions`);
+    }
+  } catch (error) {
+    logger.error(`Failed to update ${roleName} role permissions:`, error);
+  }
+}
+
 /**
  * Initialize default roles in the system.
  * Creates the default roles (ADMIN, USER) if they don't exist in the database.
+ * Updates existing roles with new permission types if they're missing.
  *
  * @returns {Promise<void>}
  */
@@ -71,16 +142,30 @@ const initializeRoles = async function () {
   const defaultRoles = [SystemRoles.ADMIN, SystemRoles.USER];
 
   for (const roleName of defaultRoles) {
-    let role = await Role.findOne({ name: roleName }).select('name').lean();
+    let role = await Role.findOne({ name: roleName });
+
     if (!role) {
+      // Create new role if it doesn't exist
       role = new Role(roleDefaults[roleName]);
-      await role.save();
+    } else {
+      // Add missing permission types
+      let isUpdated = false;
+      for (const permType of Object.values(PermissionTypes)) {
+        if (!role[permType]) {
+          role[permType] = roleDefaults[roleName][permType];
+          isUpdated = true;
+        }
+      }
+      if (isUpdated) {
+        await role.save();
+      }
     }
+    await role.save();
   }
 };
-
 module.exports = {
   getRoleByName,
   initializeRoles,
   updateRoleByName,
+  updateAccessPermissions,
 };

api/models/Role.spec.js

@@ -0,0 +1,420 @@
+const mongoose = require('mongoose');
+const { MongoMemoryServer } = require('mongodb-memory-server');
+const {
+  SystemRoles,
+  PermissionTypes,
+  roleDefaults,
+  Permissions,
+} = require('librechat-data-provider');
+const { updateAccessPermissions, initializeRoles } = require('~/models/Role');
+const getLogStores = require('~/cache/getLogStores');
+const Role = require('~/models/schema/roleSchema');
+
+// Mock the cache
+jest.mock('~/cache/getLogStores', () => {
+  return jest.fn().mockReturnValue({
+    get: jest.fn(),
+    set: jest.fn(),
+    del: jest.fn(),
+  });
+});
+
+let mongoServer;
+
+beforeAll(async () => {
+  mongoServer = await MongoMemoryServer.create();
+  const mongoUri = mongoServer.getUri();
+  await mongoose.connect(mongoUri);
+});
+
+afterAll(async () => {
+  await mongoose.disconnect();
+  await mongoServer.stop();
+});
+
+beforeEach(async () => {
+  await Role.deleteMany({});
+  getLogStores.mockClear();
+});
+
+describe('updateAccessPermissions', () => {
+  it('should update permissions when changes are needed', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: true,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: true,
+      SHARED_GLOBAL: true,
+    });
+  });
+
+  it('should not update permissions when no changes are needed', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: true,
+      SHARED_GLOBAL: false,
+    });
+  });
+
+  it('should handle non-existent roles', async () => {
+    await updateAccessPermissions('NON_EXISTENT_ROLE', {
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+      },
+    });
+
+    const role = await Role.findOne({ name: 'NON_EXISTENT_ROLE' });
+    expect(role).toBeNull();
+  });
+
+  it('should update only specified permissions', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: {
+        SHARED_GLOBAL: true,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: true,
+      SHARED_GLOBAL: true,
+    });
+  });
+
+  it('should handle partial updates', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: {
+        USE: false,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: false,
+      SHARED_GLOBAL: false,
+    });
+  });
+
+  it('should update multiple permission types at once', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+      [PermissionTypes.BOOKMARKS]: {
+        USE: true,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: { USE: false, SHARED_GLOBAL: true },
+      [PermissionTypes.BOOKMARKS]: { USE: false },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: false,
+      SHARED_GLOBAL: true,
+    });
+    expect(updatedRole[PermissionTypes.BOOKMARKS]).toEqual({
+      USE: false,
+    });
+  });
+
+  it('should handle updates for a single permission type', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: { USE: false, SHARED_GLOBAL: true },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: false,
+      SHARED_GLOBAL: true,
+    });
+  });
+
+  it('should update MULTI_CONVO permissions', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.MULTI_CONVO]: {
+        USE: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.MULTI_CONVO]: {
+        USE: true,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.MULTI_CONVO]).toEqual({
+      USE: true,
+    });
+  });
+
+  it('should update MULTI_CONVO permissions along with other permission types', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        CREATE: true,
+        USE: true,
+        SHARED_GLOBAL: false,
+      },
+      [PermissionTypes.MULTI_CONVO]: {
+        USE: false,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.PROMPTS]: { SHARED_GLOBAL: true },
+      [PermissionTypes.MULTI_CONVO]: { USE: true },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.PROMPTS]).toEqual({
+      CREATE: true,
+      USE: true,
+      SHARED_GLOBAL: true,
+    });
+    expect(updatedRole[PermissionTypes.MULTI_CONVO]).toEqual({
+      USE: true,
+    });
+  });
+
+  it('should not update MULTI_CONVO permissions when no changes are needed', async () => {
+    await new Role({
+      name: SystemRoles.USER,
+      [PermissionTypes.MULTI_CONVO]: {
+        USE: true,
+      },
+    }).save();
+
+    await updateAccessPermissions(SystemRoles.USER, {
+      [PermissionTypes.MULTI_CONVO]: {
+        USE: true,
+      },
+    });
+
+    const updatedRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+    expect(updatedRole[PermissionTypes.MULTI_CONVO]).toEqual({
+      USE: true,
+    });
+  });
+});
+
+describe('initializeRoles', () => {
+  beforeEach(async () => {
+    await Role.deleteMany({});
+  });
+
+  it('should create default roles if they do not exist', async () => {
+    await initializeRoles();
+
+    const adminRole = await Role.findOne({ name: SystemRoles.ADMIN }).lean();
+    const userRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+
+    expect(adminRole).toBeTruthy();
+    expect(userRole).toBeTruthy();
+
+    // Check if all permission types exist
+    Object.values(PermissionTypes).forEach((permType) => {
+      expect(adminRole[permType]).toBeDefined();
+      expect(userRole[permType]).toBeDefined();
+    });
+
+    // Check if permissions match defaults (example for ADMIN role)
+    expect(adminRole[PermissionTypes.PROMPTS].SHARED_GLOBAL).toBe(true);
+    expect(adminRole[PermissionTypes.BOOKMARKS].USE).toBe(true);
+    expect(adminRole[PermissionTypes.AGENTS].CREATE).toBe(true);
+  });
+
+  it('should not modify existing permissions for existing roles', async () => {
+    const customUserRole = {
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: {
+        [Permissions.USE]: false,
+        [Permissions.CREATE]: true,
+        [Permissions.SHARED_GLOBAL]: true,
+      },
+      [PermissionTypes.BOOKMARKS]: {
+        [Permissions.USE]: false,
+      },
+    };
+
+    await new Role(customUserRole).save();
+
+    await initializeRoles();
+
+    const userRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+
+    expect(userRole[PermissionTypes.PROMPTS]).toEqual(customUserRole[PermissionTypes.PROMPTS]);
+    expect(userRole[PermissionTypes.BOOKMARKS]).toEqual(customUserRole[PermissionTypes.BOOKMARKS]);
+    expect(userRole[PermissionTypes.AGENTS]).toBeDefined();
+  });
+
+  it('should add new permission types to existing roles', async () => {
+    const partialUserRole = {
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: roleDefaults[SystemRoles.USER][PermissionTypes.PROMPTS],
+      [PermissionTypes.BOOKMARKS]: roleDefaults[SystemRoles.USER][PermissionTypes.BOOKMARKS],
+    };
+
+    await new Role(partialUserRole).save();
+
+    await initializeRoles();
+
+    const userRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+
+    expect(userRole[PermissionTypes.AGENTS]).toBeDefined();
+    expect(userRole[PermissionTypes.AGENTS].CREATE).toBeDefined();
+    expect(userRole[PermissionTypes.AGENTS].USE).toBeDefined();
+    expect(userRole[PermissionTypes.AGENTS].SHARED_GLOBAL).toBeDefined();
+  });
+
+  it('should handle multiple runs without duplicating or modifying data', async () => {
+    await initializeRoles();
+    await initializeRoles();
+
+    const adminRoles = await Role.find({ name: SystemRoles.ADMIN });
+    const userRoles = await Role.find({ name: SystemRoles.USER });
+
+    expect(adminRoles).toHaveLength(1);
+    expect(userRoles).toHaveLength(1);
+
+    const adminRole = adminRoles[0].toObject();
+    const userRole = userRoles[0].toObject();
+
+    // Check if all permission types exist
+    Object.values(PermissionTypes).forEach((permType) => {
+      expect(adminRole[permType]).toBeDefined();
+      expect(userRole[permType]).toBeDefined();
+    });
+  });
+
+  it('should update roles with missing permission types from roleDefaults', async () => {
+    const partialAdminRole = {
+      name: SystemRoles.ADMIN,
+      [PermissionTypes.PROMPTS]: {
+        [Permissions.USE]: false,
+        [Permissions.CREATE]: false,
+        [Permissions.SHARED_GLOBAL]: false,
+      },
+      [PermissionTypes.BOOKMARKS]: roleDefaults[SystemRoles.ADMIN][PermissionTypes.BOOKMARKS],
+    };
+
+    await new Role(partialAdminRole).save();
+
+    await initializeRoles();
+
+    const adminRole = await Role.findOne({ name: SystemRoles.ADMIN }).lean();
+
+    expect(adminRole[PermissionTypes.PROMPTS]).toEqual(partialAdminRole[PermissionTypes.PROMPTS]);
+    expect(adminRole[PermissionTypes.AGENTS]).toBeDefined();
+    expect(adminRole[PermissionTypes.AGENTS].CREATE).toBeDefined();
+    expect(adminRole[PermissionTypes.AGENTS].USE).toBeDefined();
+    expect(adminRole[PermissionTypes.AGENTS].SHARED_GLOBAL).toBeDefined();
+  });
+
+  it('should include MULTI_CONVO permissions when creating default roles', async () => {
+    await initializeRoles();
+
+    const adminRole = await Role.findOne({ name: SystemRoles.ADMIN }).lean();
+    const userRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+
+    expect(adminRole[PermissionTypes.MULTI_CONVO]).toBeDefined();
+    expect(userRole[PermissionTypes.MULTI_CONVO]).toBeDefined();
+
+    // Check if MULTI_CONVO permissions match defaults
+    expect(adminRole[PermissionTypes.MULTI_CONVO].USE).toBe(
+      roleDefaults[SystemRoles.ADMIN][PermissionTypes.MULTI_CONVO].USE,
+    );
+    expect(userRole[PermissionTypes.MULTI_CONVO].USE).toBe(
+      roleDefaults[SystemRoles.USER][PermissionTypes.MULTI_CONVO].USE,
+    );
+  });
+
+  it('should add MULTI_CONVO permissions to existing roles without them', async () => {
+    const partialUserRole = {
+      name: SystemRoles.USER,
+      [PermissionTypes.PROMPTS]: roleDefaults[SystemRoles.USER][PermissionTypes.PROMPTS],
+      [PermissionTypes.BOOKMARKS]: roleDefaults[SystemRoles.USER][PermissionTypes.BOOKMARKS],
+    };
+
+    await new Role(partialUserRole).save();
+
+    await initializeRoles();
+
+    const userRole = await Role.findOne({ name: SystemRoles.USER }).lean();
+
+    expect(userRole[PermissionTypes.MULTI_CONVO]).toBeDefined();
+    expect(userRole[PermissionTypes.MULTI_CONVO].USE).toBeDefined();
+  });
+});

api/models/Session.js

@@ -1,6 +1,6 @@
-const crypto = require('crypto');
 const mongoose = require('mongoose');
 const signPayload = require('~/server/services/signPayload');
+const { hashToken } = require('~/server/utils/crypto');
 const { logger } = require('~/config');
 
 const { REFRESH_TOKEN_EXPIRY } = process.env ?? {};
@@ -39,8 +39,7 @@ sessionSchema.methods.generateRefreshToken = async function () {
       expirationTime: Math.floor((expiresIn - Date.now()) / 1000),
     });
 
-    const hash = crypto.createHash('sha256');
-    this.refreshTokenHash = hash.update(refreshToken).digest('hex');
+    this.refreshTokenHash = await hashToken(refreshToken);
 
     await this.save();
 

api/models/Share.js

@@ -1,117 +1,252 @@
-const crypto = require('crypto');
-const { getMessages } = require('./Message');
+const { nanoid } = require('nanoid');
+const { Constants } = require('librechat-data-provider');
 const SharedLink = require('./schema/shareSchema');
+const { getMessages } = require('./Message');
 const logger = require('~/config/winston');
 
-module.exports = {
-  SharedLink,
-  getSharedMessages: async (shareId) => {
-    try {
-      const share = await SharedLink.findOne({ shareId })
-        .populate({
-          path: 'messages',
-          select: '-_id -__v -user',
-        })
-        .select('-_id -__v -user')
-        .lean();
+/**
+ * Anonymizes a conversation ID
+ * @returns {string} The anonymized conversation ID
+ */
+function anonymizeConvoId() {
+  return `convo_${nanoid()}`;
+}
 
-      if (!share || !share.conversationId || !share.isPublic) {
-        return null;
-      }
+/**
+ * Anonymizes an assistant ID
+ * @returns {string} The anonymized assistant ID
+ */
+function anonymizeAssistantId() {
+  return `a_${nanoid()}`;
+}
 
-      return share;
-    } catch (error) {
-      logger.error('[getShare] Error getting share link', error);
-      throw new Error('Error getting share link');
+/**
+ * Anonymizes a message ID
+ * @param {string} id - The original message ID
+ * @returns {string} The anonymized message ID
+ */
+function anonymizeMessageId(id) {
+  return id === Constants.NO_PARENT ? id : `msg_${nanoid()}`;
+}
+
+/**
+ * Anonymizes a conversation object
+ * @param {object} conversation - The conversation object
+ * @returns {object} The anonymized conversation object
+ */
+function anonymizeConvo(conversation) {
+  const newConvo = { ...conversation };
+  if (newConvo.assistant_id) {
+    newConvo.assistant_id = anonymizeAssistantId();
+  }
+  return newConvo;
+}
+
+/**
+ * Anonymizes messages in a conversation
+ * @param {TMessage[]} messages - The original messages
+ * @param {string} newConvoId - The new conversation ID
+ * @returns {TMessage[]} The anonymized messages
+ */
+function anonymizeMessages(messages, newConvoId) {
+  const idMap = new Map();
+  return messages.map((message) => {
+    const newMessageId = anonymizeMessageId(message.messageId);
+    idMap.set(message.messageId, newMessageId);
+
+    const anonymizedMessage = Object.assign(message, {
+      messageId: newMessageId,
+      parentMessageId:
+        idMap.get(message.parentMessageId) || anonymizeMessageId(message.parentMessageId),
+      conversationId: newConvoId,
+    });
+
+    if (anonymizedMessage.model && anonymizedMessage.model.startsWith('asst_')) {
+      anonymizedMessage.model = anonymizeAssistantId();
     }
-  },
 
-  getSharedLinks: async (user, pageNumber = 1, pageSize = 25, isPublic = true) => {
-    const query = { user, isPublic };
-    try {
-      const totalConvos = (await SharedLink.countDocuments(query)) || 1;
-      const totalPages = Math.ceil(totalConvos / pageSize);
-      const shares = await SharedLink.find(query)
+    return anonymizedMessage;
+  });
+}
+
+/**
+ * Retrieves shared messages for a given share ID
+ * @param {string} shareId - The share ID
+ * @returns {Promise<object|null>} The shared conversation data or null if not found
+ */
+async function getSharedMessages(shareId) {
+  try {
+    const share = await SharedLink.findOne({ shareId })
+      .populate({
+        path: 'messages',
+        select: '-_id -__v -user',
+      })
+      .select('-_id -__v -user')
+      .lean();
+
+    if (!share || !share.conversationId || !share.isPublic) {
+      return null;
+    }
+
+    const newConvoId = anonymizeConvoId();
+    return Object.assign(share, {
+      conversationId: newConvoId,
+      messages: anonymizeMessages(share.messages, newConvoId),
+    });
+  } catch (error) {
+    logger.error('[getShare] Error getting share link', error);
+    throw new Error('Error getting share link');
+  }
+}
+
+/**
+ * Retrieves shared links for a user
+ * @param {string} user - The user ID
+ * @param {number} [pageNumber=1] - The page number
+ * @param {number} [pageSize=25] - The page size
+ * @param {boolean} [isPublic=true] - Whether to retrieve public links only
+ * @returns {Promise<object>} The shared links and pagination data
+ */
+async function getSharedLinks(user, pageNumber = 1, pageSize = 25, isPublic = true) {
+  const query = { user, isPublic };
+  try {
+    const [totalConvos, sharedLinks] = await Promise.all([
+      SharedLink.countDocuments(query),
+      SharedLink.find(query)
         .sort({ updatedAt: -1 })
         .skip((pageNumber - 1) * pageSize)
         .limit(pageSize)
         .select('-_id -__v -user')
-        .lean();
+        .lean(),
+    ]);
 
-      return { sharedLinks: shares, pages: totalPages, pageNumber, pageSize };
-    } catch (error) {
-      logger.error('[getShareByPage] Error getting shares', error);
-      throw new Error('Error getting shares');
-    }
-  },
+    const totalPages = Math.ceil((totalConvos || 1) / pageSize);
 
-  createSharedLink: async (user, { conversationId, ...shareData }) => {
-    try {
-      const share = await SharedLink.findOne({ conversationId }).select('-_id -__v -user').lean();
-      if (share) {
-        return share;
-      }
+    return {
+      sharedLinks,
+      pages: totalPages,
+      pageNumber,
+      pageSize,
+    };
+  } catch (error) {
+    logger.error('[getShareByPage] Error getting shares', error);
+    throw new Error('Error getting shares');
+  }
+}
 
-      const shareId = crypto.randomUUID();
-      const messages = await getMessages({ conversationId });
-      const update = { ...shareData, shareId, messages, user };
-      return await SharedLink.findOneAndUpdate({ conversationId: conversationId, user }, update, {
-        new: true,
-        upsert: true,
+/**
+ * Creates a new shared link
+ * @param {string} user - The user ID
+ * @param {object} shareData - The share data
+ * @param {string} shareData.conversationId - The conversation ID
+ * @returns {Promise<object>} The created shared link
+ */
+async function createSharedLink(user, { conversationId, ...shareData }) {
+  try {
+    const share = await SharedLink.findOne({ conversationId }).select('-_id -__v -user').lean();
+    if (share) {
+      const newConvoId = anonymizeConvoId();
+      const sharedConvo = anonymizeConvo(share);
+      return Object.assign(sharedConvo, {
+        conversationId: newConvoId,
+        messages: anonymizeMessages(share.messages, newConvoId),
       });
-    } catch (error) {
-      logger.error('[createSharedLink] Error creating shared link', error);
-      throw new Error('Error creating shared link');
     }
-  },
 
-  updateSharedLink: async (user, { conversationId, ...shareData }) => {
-    try {
-      const share = await SharedLink.findOne({ conversationId }).select('-_id -__v -user').lean();
-      if (!share) {
-        return { message: 'Share not found' };
-      }
+    const shareId = nanoid();
+    const messages = await getMessages({ conversationId });
+    const update = { ...shareData, shareId, messages, user };
+    const newShare = await SharedLink.findOneAndUpdate({ conversationId, user }, update, {
+      new: true,
+      upsert: true,
+    }).lean();
 
-      // update messages to the latest
-      const messages = await getMessages({ conversationId });
-      const update = { ...shareData, messages, user };
-      return await SharedLink.findOneAndUpdate({ conversationId: conversationId, user }, update, {
-        new: true,
-        upsert: false,
-      });
-    } catch (error) {
-      logger.error('[updateSharedLink] Error updating shared link', error);
-      throw new Error('Error updating shared link');
-    }
-  },
+    const newConvoId = anonymizeConvoId();
+    const sharedConvo = anonymizeConvo(newShare);
+    return Object.assign(sharedConvo, {
+      conversationId: newConvoId,
+      messages: anonymizeMessages(newShare.messages, newConvoId),
+    });
+  } catch (error) {
+    logger.error('[createSharedLink] Error creating shared link', error);
+    throw new Error('Error creating shared link');
+  }
+}
 
-  deleteSharedLink: async (user, { shareId }) => {
-    try {
-      const share = await SharedLink.findOne({ shareId, user });
-      if (!share) {
-        return { message: 'Share not found' };
-      }
-      return await SharedLink.findOneAndDelete({ shareId, user });
-    } catch (error) {
-      logger.error('[deleteSharedLink] Error deleting shared link', error);
-      throw new Error('Error deleting shared link');
+/**
+ * Updates an existing shared link
+ * @param {string} user - The user ID
+ * @param {object} shareData - The share data to update
+ * @param {string} shareData.conversationId - The conversation ID
+ * @returns {Promise<object>} The updated shared link
+ */
+async function updateSharedLink(user, { conversationId, ...shareData }) {
+  try {
+    const share = await SharedLink.findOne({ conversationId }).select('-_id -__v -user').lean();
+    if (!share) {
+      return { message: 'Share not found' };
     }
-  },
-  /**
-   * Deletes all shared links for a specific user.
-   * @param {string} user - The user ID.
-   * @returns {Promise<{ message: string, deletedCount?: number }>} A result object indicating success or error message.
-   */
-  deleteAllSharedLinks: async (user) => {
-    try {
-      const result = await SharedLink.deleteMany({ user });
-      return {
-        message: 'All shared links have been deleted successfully',
-        deletedCount: result.deletedCount,
-      };
-    } catch (error) {
-      logger.error('[deleteAllSharedLinks] Error deleting shared links', error);
-      throw new Error('Error deleting shared links');
-    }
-  },
+
+    const messages = await getMessages({ conversationId });
+    const update = { ...shareData, messages, user };
+    const updatedShare = await SharedLink.findOneAndUpdate({ conversationId, user }, update, {
+      new: true,
+      upsert: false,
+    }).lean();
+
+    const newConvoId = anonymizeConvoId();
+    const sharedConvo = anonymizeConvo(updatedShare);
+    return Object.assign(sharedConvo, {
+      conversationId: newConvoId,
+      messages: anonymizeMessages(updatedShare.messages, newConvoId),
+    });
+  } catch (error) {
+    logger.error('[updateSharedLink] Error updating shared link', error);
+    throw new Error('Error updating shared link');
+  }
+}
+
+/**
+ * Deletes a shared link
+ * @param {string} user - The user ID
+ * @param {object} params - The deletion parameters
+ * @param {string} params.shareId - The share ID to delete
+ * @returns {Promise<object>} The result of the deletion
+ */
+async function deleteSharedLink(user, { shareId }) {
+  try {
+    const result = await SharedLink.findOneAndDelete({ shareId, user });
+    return result ? { message: 'Share deleted successfully' } : { message: 'Share not found' };
+  } catch (error) {
+    logger.error('[deleteSharedLink] Error deleting shared link', error);
+    throw new Error('Error deleting shared link');
+  }
+}
+
+/**
+ * Deletes all shared links for a specific user
+ * @param {string} user - The user ID
+ * @returns {Promise<object>} The result of the deletion
+ */
+async function deleteAllSharedLinks(user) {
+  try {
+    const result = await SharedLink.deleteMany({ user });
+    return {
+      message: 'All shared links have been deleted successfully',
+      deletedCount: result.deletedCount,
+    };
+  } catch (error) {
+    logger.error('[deleteAllSharedLinks] Error deleting shared links', error);
+    throw new Error('Error deleting shared links');
+  }
+}
+
+module.exports = {
+  SharedLink,
+  getSharedLinks,
+  createSharedLink,
+  updateSharedLink,
+  deleteSharedLink,
+  getSharedMessages,
+  deleteAllSharedLinks,
 };

api/models/Token.js

@@ -0,0 +1,117 @@
+const tokenSchema = require('./schema/tokenSchema');
+const mongoose = require('mongoose');
+const { logger } = require('~/config');
+
+/**
+ * Token model.
+ * @type {mongoose.Model}
+ */
+const Token = mongoose.model('Token', tokenSchema);
+
+/**
+ * Creates a new Token instance.
+ * @param {Object} tokenData - The data for the new Token.
+ * @param {mongoose.Types.ObjectId} tokenData.userId - The user's ID. It is required.
+ * @param {String} tokenData.email - The user's email.
+ * @param {String} tokenData.token - The token. It is required.
+ * @param {Number} tokenData.expiresIn - The number of seconds until the token expires.
+ * @returns {Promise<mongoose.Document>} The new Token instance.
+ * @throws Will throw an error if token creation fails.
+ */
+async function createToken(tokenData) {
+  try {
+    const currentTime = new Date();
+    const expiresAt = new Date(currentTime.getTime() + tokenData.expiresIn * 1000);
+
+    const newTokenData = {
+      ...tokenData,
+      createdAt: currentTime,
+      expiresAt,
+    };
+
+    const newToken = new Token(newTokenData);
+    return await newToken.save();
+  } catch (error) {
+    logger.debug('An error occurred while creating token:', error);
+    throw error;
+  }
+}
+
+/**
+ * Finds a Token document that matches the provided query.
+ * @param {Object} query - The query to match against.
+ * @param {mongoose.Types.ObjectId|String} query.userId - The ID of the user.
+ * @param {String} query.token - The token value.
+ * @param {String} query.email - The email of the user.
+ * @returns {Promise<Object|null>} The matched Token document, or null if not found.
+ * @throws Will throw an error if the find operation fails.
+ */
+async function findToken(query) {
+  try {
+    const conditions = [];
+
+    if (query.userId) {
+      conditions.push({ userId: query.userId });
+    }
+    if (query.token) {
+      conditions.push({ token: query.token });
+    }
+    if (query.email) {
+      conditions.push({ email: query.email });
+    }
+
+    const token = await Token.findOne({
+      $and: conditions,
+    }).lean();
+
+    return token;
+  } catch (error) {
+    logger.debug('An error occurred while finding token:', error);
+    throw error;
+  }
+}
+
+/**
+ * Updates a Token document that matches the provided query.
+ * @param {Object} query - The query to match against.
+ * @param {mongoose.Types.ObjectId|String} query.userId - The ID of the user.
+ * @param {String} query.token - The token value.
+ * @param {Object} updateData - The data to update the Token with.
+ * @returns {Promise<mongoose.Document|null>} The updated Token document, or null if not found.
+ * @throws Will throw an error if the update operation fails.
+ */
+async function updateToken(query, updateData) {
+  try {
+    return await Token.findOneAndUpdate(query, updateData, { new: true });
+  } catch (error) {
+    logger.debug('An error occurred while updating token:', error);
+    throw error;
+  }
+}
+
+/**
+ * Deletes all Token documents that match the provided token, user ID, or email.
+ * @param {Object} query - The query to match against.
+ * @param {mongoose.Types.ObjectId|String} query.userId - The ID of the user.
+ * @param {String} query.token - The token value.
+ * @param {String} query.email - The email of the user.
+ * @returns {Promise<Object>} The result of the delete operation.
+ * @throws Will throw an error if the delete operation fails.
+ */
+async function deleteTokens(query) {
+  try {
+    return await Token.deleteMany({
+      $or: [{ userId: query.userId }, { token: query.token }, { email: query.email }],
+    });
+  } catch (error) {
+    logger.debug('An error occurred while deleting tokens:', error);
+    throw error;
+  }
+}
+
+module.exports = {
+  createToken,
+  findToken,
+  updateToken,
+  deleteTokens,
+};

api/models/Transaction.js

@@ -1,12 +1,12 @@
 const mongoose = require('mongoose');
-const { isEnabled } = require('../server/utils/handleText');
+const { isEnabled } = require('~/server/utils/handleText');
 const transactionSchema = require('./schema/transaction');
-const { getMultiplier } = require('./tx');
+const { getMultiplier, getCacheMultiplier } = require('./tx');
 const { logger } = require('~/config');
 const Balance = require('./Balance');
 const cancelRate = 1.15;
 
-// Method to calculate and set the tokenValue for a transaction
+/** Method to calculate and set the tokenValue for a transaction */
 transactionSchema.methods.calculateTokenValue = function () {
   if (!this.valueKey || !this.tokenType) {
     this.tokenValue = this.rawAmount;
@@ -21,15 +21,17 @@ transactionSchema.methods.calculateTokenValue = function () {
   }
 };
 
-// Static method to create a transaction and update the balance
-transactionSchema.statics.create = async function (transactionData) {
+/**
+ * Static method to create a transaction and update the balance
+ * @param {txData} txData - Transaction data.
+ */
+transactionSchema.statics.create = async function (txData) {
   const Transaction = this;
 
-  const transaction = new Transaction(transactionData);
-  transaction.endpointTokenConfig = transactionData.endpointTokenConfig;
+  const transaction = new Transaction(txData);
+  transaction.endpointTokenConfig = txData.endpointTokenConfig;
   transaction.calculateTokenValue();
 
-  // Save the transaction
   await transaction.save();
 
   if (!isEnabled(process.env.CHECK_BALANCE)) {
@@ -57,6 +59,109 @@ transactionSchema.statics.create = async function (transactionData) {
   };
 };
 
+/**
+ * Static method to create a structured transaction and update the balance
+ * @param {txData} txData - Transaction data.
+ */
+transactionSchema.statics.createStructured = async function (txData) {
+  const Transaction = this;
+
+  const transaction = new Transaction({
+    ...txData,
+    endpointTokenConfig: txData.endpointTokenConfig,
+  });
+
+  transaction.calculateStructuredTokenValue();
+
+  await transaction.save();
+
+  if (!isEnabled(process.env.CHECK_BALANCE)) {
+    return;
+  }
+
+  let balance = await Balance.findOne({ user: transaction.user }).lean();
+  let incrementValue = transaction.tokenValue;
+
+  if (balance && balance?.tokenCredits + incrementValue < 0) {
+    incrementValue = -balance.tokenCredits;
+  }
+
+  balance = await Balance.findOneAndUpdate(
+    { user: transaction.user },
+    { $inc: { tokenCredits: incrementValue } },
+    { upsert: true, new: true },
+  ).lean();
+
+  return {
+    rate: transaction.rate,
+    user: transaction.user.toString(),
+    balance: balance.tokenCredits,
+    [transaction.tokenType]: incrementValue,
+  };
+};
+
+/** Method to calculate token value for structured tokens */
+transactionSchema.methods.calculateStructuredTokenValue = function () {
+  if (!this.tokenType) {
+    this.tokenValue = this.rawAmount;
+    return;
+  }
+
+  const { model, endpointTokenConfig } = this;
+
+  if (this.tokenType === 'prompt') {
+    const inputMultiplier = getMultiplier({ tokenType: 'prompt', model, endpointTokenConfig });
+    const writeMultiplier =
+      getCacheMultiplier({ cacheType: 'write', model, endpointTokenConfig }) ?? inputMultiplier;
+    const readMultiplier =
+      getCacheMultiplier({ cacheType: 'read', model, endpointTokenConfig }) ?? inputMultiplier;
+
+    this.rateDetail = {
+      input: inputMultiplier,
+      write: writeMultiplier,
+      read: readMultiplier,
+    };
+
+    const totalPromptTokens =
+      Math.abs(this.inputTokens || 0) +
+      Math.abs(this.writeTokens || 0) +
+      Math.abs(this.readTokens || 0);
+
+    if (totalPromptTokens > 0) {
+      this.rate =
+        (Math.abs(inputMultiplier * (this.inputTokens || 0)) +
+          Math.abs(writeMultiplier * (this.writeTokens || 0)) +
+          Math.abs(readMultiplier * (this.readTokens || 0))) /
+        totalPromptTokens;
+    } else {
+      this.rate = Math.abs(inputMultiplier); // Default to input rate if no tokens
+    }
+
+    this.tokenValue = -(
+      Math.abs(this.inputTokens || 0) * inputMultiplier +
+      Math.abs(this.writeTokens || 0) * writeMultiplier +
+      Math.abs(this.readTokens || 0) * readMultiplier
+    );
+
+    this.rawAmount = -totalPromptTokens;
+  } else if (this.tokenType === 'completion') {
+    const multiplier = getMultiplier({ tokenType: this.tokenType, model, endpointTokenConfig });
+    this.rate = Math.abs(multiplier);
+    this.tokenValue = -Math.abs(this.rawAmount) * multiplier;
+    this.rawAmount = -Math.abs(this.rawAmount);
+  }
+
+  if (this.context && this.tokenType === 'completion' && this.context === 'incomplete') {
+    this.tokenValue = Math.ceil(this.tokenValue * cancelRate);
+    this.rate *= cancelRate;
+    if (this.rateDetail) {
+      this.rateDetail = Object.fromEntries(
+        Object.entries(this.rateDetail).map(([k, v]) => [k, v * cancelRate]),
+      );
+    }
+  }
+};
+
 const Transaction = mongoose.model('Transaction', transactionSchema);
 
 /**

api/models/Transaction.spec.js

@@ -0,0 +1,348 @@
+const mongoose = require('mongoose');
+const { MongoMemoryServer } = require('mongodb-memory-server');
+const Balance = require('./Balance');
+const { spendTokens, spendStructuredTokens } = require('./spendTokens');
+const { getMultiplier, getCacheMultiplier } = require('./tx');
+
+let mongoServer;
+
+beforeAll(async () => {
+  mongoServer = await MongoMemoryServer.create();
+  const mongoUri = mongoServer.getUri();
+  await mongoose.connect(mongoUri);
+});
+
+afterAll(async () => {
+  await mongoose.disconnect();
+  await mongoServer.stop();
+});
+
+beforeEach(async () => {
+  await mongoose.connection.dropDatabase();
+});
+
+describe('Regular Token Spending Tests', () => {
+  test('Balance should decrease when spending tokens with spendTokens', async () => {
+    // Arrange
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 10000000; // $10.00
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'gpt-3.5-turbo';
+    const txData = {
+      user: userId,
+      conversationId: 'test-conversation-id',
+      model,
+      context: 'test',
+      endpointTokenConfig: null,
+    };
+
+    const tokenUsage = {
+      promptTokens: 100,
+      completionTokens: 50,
+    };
+
+    // Act
+    process.env.CHECK_BALANCE = 'true';
+    await spendTokens(txData, tokenUsage);
+
+    // Assert
+    console.log('Initial Balance:', initialBalance);
+
+    const updatedBalance = await Balance.findOne({ user: userId });
+    console.log('Updated Balance:', updatedBalance.tokenCredits);
+
+    const promptMultiplier = getMultiplier({ model, tokenType: 'prompt' });
+    const completionMultiplier = getMultiplier({ model, tokenType: 'completion' });
+
+    const expectedPromptCost = tokenUsage.promptTokens * promptMultiplier;
+    const expectedCompletionCost = tokenUsage.completionTokens * completionMultiplier;
+    const expectedTotalCost = expectedPromptCost + expectedCompletionCost;
+    const expectedBalance = initialBalance - expectedTotalCost;
+
+    expect(updatedBalance.tokenCredits).toBeLessThan(initialBalance);
+    expect(updatedBalance.tokenCredits).toBeCloseTo(expectedBalance, 0);
+
+    console.log('Expected Total Cost:', expectedTotalCost);
+    console.log('Actual Balance Decrease:', initialBalance - updatedBalance.tokenCredits);
+  });
+
+  test('spendTokens should handle zero completion tokens', async () => {
+    // Arrange
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 10000000; // $10.00
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'gpt-3.5-turbo';
+    const txData = {
+      user: userId,
+      conversationId: 'test-conversation-id',
+      model,
+      context: 'test',
+      endpointTokenConfig: null,
+    };
+
+    const tokenUsage = {
+      promptTokens: 100,
+      completionTokens: 0,
+    };
+
+    // Act
+    process.env.CHECK_BALANCE = 'true';
+    await spendTokens(txData, tokenUsage);
+
+    // Assert
+    const updatedBalance = await Balance.findOne({ user: userId });
+
+    const promptMultiplier = getMultiplier({ model, tokenType: 'prompt' });
+    const expectedCost = tokenUsage.promptTokens * promptMultiplier;
+    expect(updatedBalance.tokenCredits).toBeCloseTo(initialBalance - expectedCost, 0);
+
+    console.log('Initial Balance:', initialBalance);
+    console.log('Updated Balance:', updatedBalance.tokenCredits);
+    console.log('Expected Cost:', expectedCost);
+  });
+
+  test('spendTokens should handle undefined token counts', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 10000000; // $10.00
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'gpt-3.5-turbo';
+    const txData = {
+      user: userId,
+      conversationId: 'test-conversation-id',
+      model,
+      context: 'test',
+      endpointTokenConfig: null,
+    };
+
+    const tokenUsage = {};
+
+    const result = await spendTokens(txData, tokenUsage);
+
+    expect(result).toBeUndefined();
+  });
+
+  test('spendTokens should handle only prompt tokens', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 10000000; // $10.00
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'gpt-3.5-turbo';
+    const txData = {
+      user: userId,
+      conversationId: 'test-conversation-id',
+      model,
+      context: 'test',
+      endpointTokenConfig: null,
+    };
+
+    const tokenUsage = { promptTokens: 100 };
+
+    await spendTokens(txData, tokenUsage);
+
+    const updatedBalance = await Balance.findOne({ user: userId });
+
+    const promptMultiplier = getMultiplier({ model, tokenType: 'prompt' });
+    const expectedCost = 100 * promptMultiplier;
+    expect(updatedBalance.tokenCredits).toBeCloseTo(initialBalance - expectedCost, 0);
+  });
+});
+
+describe('Structured Token Spending Tests', () => {
+  test('Balance should decrease and rawAmount should be set when spending a large number of structured tokens', async () => {
+    // Arrange
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 17613154.55; // $17.61
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'claude-3-5-sonnet';
+    const txData = {
+      user: userId,
+      conversationId: 'c23a18da-706c-470a-ac28-ec87ed065199',
+      model,
+      context: 'message',
+      endpointTokenConfig: null, // We'll use the default rates
+    };
+
+    const tokenUsage = {
+      promptTokens: {
+        input: 11,
+        write: 140522,
+        read: 0,
+      },
+      completionTokens: 5,
+    };
+
+    // Get the actual multipliers
+    const promptMultiplier = getMultiplier({ model, tokenType: 'prompt' });
+    const completionMultiplier = getMultiplier({ model, tokenType: 'completion' });
+    const writeMultiplier = getCacheMultiplier({ model, cacheType: 'write' });
+    const readMultiplier = getCacheMultiplier({ model, cacheType: 'read' });
+
+    console.log('Multipliers:', {
+      promptMultiplier,
+      completionMultiplier,
+      writeMultiplier,
+      readMultiplier,
+    });
+
+    // Act
+    process.env.CHECK_BALANCE = 'true';
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    // Assert
+    console.log('Initial Balance:', initialBalance);
+    console.log('Updated Balance:', result.completion.balance);
+    console.log('Transaction Result:', result);
+
+    const expectedPromptCost =
+      tokenUsage.promptTokens.input * promptMultiplier +
+      tokenUsage.promptTokens.write * writeMultiplier +
+      tokenUsage.promptTokens.read * readMultiplier;
+    const expectedCompletionCost = tokenUsage.completionTokens * completionMultiplier;
+    const expectedTotalCost = expectedPromptCost + expectedCompletionCost;
+    const expectedBalance = initialBalance - expectedTotalCost;
+
+    console.log('Expected Cost:', expectedTotalCost);
+    console.log('Expected Balance:', expectedBalance);
+
+    expect(result.completion.balance).toBeLessThan(initialBalance);
+
+    // Allow for a small difference (e.g., 100 token credits, which is $0.0001)
+    const allowedDifference = 100;
+    expect(Math.abs(result.completion.balance - expectedBalance)).toBeLessThan(allowedDifference);
+
+    // Check if the decrease is approximately as expected
+    const balanceDecrease = initialBalance - result.completion.balance;
+    expect(balanceDecrease).toBeCloseTo(expectedTotalCost, 0);
+
+    // Check token values
+    const expectedPromptTokenValue = -(
+      tokenUsage.promptTokens.input * promptMultiplier +
+      tokenUsage.promptTokens.write * writeMultiplier +
+      tokenUsage.promptTokens.read * readMultiplier
+    );
+    const expectedCompletionTokenValue = -tokenUsage.completionTokens * completionMultiplier;
+
+    expect(result.prompt.prompt).toBeCloseTo(expectedPromptTokenValue, 1);
+    expect(result.completion.completion).toBe(expectedCompletionTokenValue);
+
+    console.log('Expected prompt tokenValue:', expectedPromptTokenValue);
+    console.log('Actual prompt tokenValue:', result.prompt.prompt);
+    console.log('Expected completion tokenValue:', expectedCompletionTokenValue);
+    console.log('Actual completion tokenValue:', result.completion.completion);
+  });
+
+  test('should handle zero completion tokens in structured spending', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 17613154.55;
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'claude-3-5-sonnet';
+    const txData = {
+      user: userId,
+      conversationId: 'test-convo',
+      model,
+      context: 'message',
+    };
+
+    const tokenUsage = {
+      promptTokens: {
+        input: 10,
+        write: 100,
+        read: 5,
+      },
+      completionTokens: 0,
+    };
+
+    process.env.CHECK_BALANCE = 'true';
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    expect(result.prompt).toBeDefined();
+    expect(result.completion).toBeUndefined();
+    expect(result.prompt.prompt).toBeLessThan(0);
+  });
+
+  test('should handle only prompt tokens in structured spending', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 17613154.55;
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'claude-3-5-sonnet';
+    const txData = {
+      user: userId,
+      conversationId: 'test-convo',
+      model,
+      context: 'message',
+    };
+
+    const tokenUsage = {
+      promptTokens: {
+        input: 10,
+        write: 100,
+        read: 5,
+      },
+    };
+
+    process.env.CHECK_BALANCE = 'true';
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    expect(result.prompt).toBeDefined();
+    expect(result.completion).toBeUndefined();
+    expect(result.prompt.prompt).toBeLessThan(0);
+  });
+
+  test('should handle undefined token counts in structured spending', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 17613154.55;
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'claude-3-5-sonnet';
+    const txData = {
+      user: userId,
+      conversationId: 'test-convo',
+      model,
+      context: 'message',
+    };
+
+    const tokenUsage = {};
+
+    process.env.CHECK_BALANCE = 'true';
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    expect(result).toEqual({
+      prompt: undefined,
+      completion: undefined,
+    });
+  });
+
+  test('should handle incomplete context for completion tokens', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 17613154.55;
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'claude-3-5-sonnet';
+    const txData = {
+      user: userId,
+      conversationId: 'test-convo',
+      model,
+      context: 'incomplete',
+    };
+
+    const tokenUsage = {
+      promptTokens: {
+        input: 10,
+        write: 100,
+        read: 5,
+      },
+      completionTokens: 50,
+    };
+
+    process.env.CHECK_BALANCE = 'true';
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    expect(result.completion.completion).toBeCloseTo(-50 * 15 * 1.15, 0); // Assuming multiplier is 15 and cancelRate is 1.15
+  });
+});

api/models/index.js

@@ -1,11 +1,3 @@
-const {
-  getMessages,
-  saveMessage,
-  recordMessage,
-  updateMessage,
-  deleteMessagesSince,
-  deleteMessages,
-} = require('./Message');
 const {
   comparePassword,
   deleteUserById,
@@ -16,8 +8,6 @@ const {
   countUsers,
   findUser,
 } = require('./userMethods');
-const { getConvoTitle, getConvo, saveConvo, deleteConvos } = require('./Conversation');
-const { getPreset, getPresets, savePreset, deletePresets } = require('./Preset');
 const {
   findFileById,
   createFile,
@@ -27,26 +17,40 @@ const {
   getFiles,
   updateFileUsage,
 } = require('./File');
-const Key = require('./Key');
-const User = require('./User');
+const {
+  getMessages,
+  saveMessage,
+  recordMessage,
+  updateMessage,
+  deleteMessagesSince,
+  deleteMessages,
+} = require('./Message');
+const { getConvoTitle, getConvo, saveConvo, deleteConvos } = require('./Conversation');
+const { getPreset, getPresets, savePreset, deletePresets } = require('./Preset');
+const { createToken, findToken, updateToken, deleteTokens } = require('./Token');
 const Session = require('./Session');
 const Balance = require('./Balance');
+const User = require('./User');
+const Key = require('./Key');
 
 module.exports = {
-  User,
-  Key,
-  Session,
-  Balance,
-
   comparePassword,
   deleteUserById,
   generateToken,
   getUserById,
-  countUsers,
-  createUser,
   updateUser,
+  createUser,
+  countUsers,
   findUser,
 
+  findFileById,
+  createFile,
+  updateFile,
+  deleteFile,
+  deleteFiles,
+  getFiles,
+  updateFileUsage,
+
   getMessages,
   saveMessage,
   recordMessage,
@@ -64,11 +68,13 @@ module.exports = {
   savePreset,
   deletePresets,
 
-  findFileById,
-  createFile,
-  updateFile,
-  deleteFile,
-  deleteFiles,
-  getFiles,
-  updateFileUsage,
+  createToken,
+  findToken,
+  updateToken,
+  deleteTokens,
+
+  User,
+  Key,
+  Session,
+  Balance,
 };

api/models/inviteUser.js

@@ -0,0 +1,70 @@
+const crypto = require('crypto');
+const bcrypt = require('bcryptjs');
+const mongoose = require('mongoose');
+const { createToken, findToken } = require('./Token');
+const logger = require('~/config/winston');
+
+/**
+ * @module inviteUser
+ * @description This module provides functions to create and get user invites
+ */
+
+/**
+ * @function createInvite
+ * @description This function creates a new user invite
+ * @param {string} email - The email of the user to invite
+ * @returns {Promise<Object>} A promise that resolves to the saved invite document
+ * @throws {Error} If there is an error creating the invite
+ */
+const createInvite = async (email) => {
+  try {
+    let token = crypto.randomBytes(32).toString('hex');
+    const hash = bcrypt.hashSync(token, 10);
+    const encodedToken = encodeURIComponent(token);
+
+    const fakeUserId = new mongoose.Types.ObjectId();
+
+    await createToken({
+      userId: fakeUserId,
+      email,
+      token: hash,
+      createdAt: Date.now(),
+      expiresIn: 604800,
+    });
+
+    return encodedToken;
+  } catch (error) {
+    logger.error('[createInvite] Error creating invite', error);
+    return { message: 'Error creating invite' };
+  }
+};
+
+/**
+ * @function getInvite
+ * @description This function retrieves a user invite
+ * @param {string} encodedToken - The token of the invite to retrieve
+ * @param {string} email - The email of the user to validate
+ * @returns {Promise<Object>} A promise that resolves to the retrieved invite document
+ * @throws {Error} If there is an error retrieving the invite, if the invite does not exist, or if the email does not match
+ */
+const getInvite = async (encodedToken, email) => {
+  try {
+    const token = decodeURIComponent(encodedToken);
+    const hash = bcrypt.hashSync(token, 10);
+    const invite = await findToken({ token: hash, email });
+
+    if (!invite) {
+      throw new Error('Invite not found or email does not match');
+    }
+
+    return invite;
+  } catch (error) {
+    logger.error('[getInvite] Error getting invite', error);
+    return { error: true, message: error.message };
+  }
+};
+
+module.exports = {
+  createInvite,
+  getInvite,
+};

api/models/schema/action.js

@@ -39,6 +39,7 @@ const actionSchema = new Schema({
     default: 'action_prototype',
   },
   settings: Schema.Types.Mixed,
+  agent_id: String,
   assistant_id: String,
   metadata: {
     api_key: String, // private, encrypted

api/models/schema/agent.js

@@ -0,0 +1,71 @@
+const mongoose = require('mongoose');
+
+const agentSchema = mongoose.Schema(
+  {
+    id: {
+      type: String,
+      index: true,
+      required: true,
+    },
+    name: {
+      type: String,
+    },
+    description: {
+      type: String,
+    },
+    instructions: {
+      type: String,
+    },
+    avatar: {
+      type: {
+        filepath: String,
+        source: String,
+      },
+      default: undefined,
+    },
+    provider: {
+      type: String,
+      required: true,
+    },
+    model: {
+      type: String,
+      required: true,
+    },
+    model_parameters: {
+      type: Object,
+    },
+    access_level: {
+      type: Number,
+    },
+    tools: {
+      type: [String],
+      default: undefined,
+    },
+    tool_kwargs: {
+      type: [{ type: mongoose.Schema.Types.Mixed }],
+    },
+    file_ids: {
+      type: [String],
+      default: undefined,
+    },
+    actions: {
+      type: [String],
+      default: undefined,
+    },
+    author: {
+      type: mongoose.Schema.Types.ObjectId,
+      ref: 'User',
+      required: true,
+    },
+    projectIds: {
+      type: [mongoose.Schema.Types.ObjectId],
+      ref: 'Project',
+      index: true,
+    },
+  },
+  {
+    timestamps: true,
+  },
+);
+
+module.exports = agentSchema;

api/models/schema/assistant.js

@@ -19,6 +19,10 @@ const assistantSchema = mongoose.Schema(
       },
       default: undefined,
     },
+    conversation_starters: {
+      type: [String],
+      default: [],
+    },
     access_level: {
       type: Number,
     },

api/models/schema/conversationTagSchema.js

@@ -0,0 +1,31 @@
+const mongoose = require('mongoose');
+
+const conversationTagSchema = mongoose.Schema(
+  {
+    tag: {
+      type: String,
+      index: true,
+    },
+    user: {
+      type: String,
+      index: true,
+    },
+    description: {
+      type: String,
+      index: true,
+    },
+    count: {
+      type: Number,
+      default: 0,
+    },
+    position: {
+      type: Number,
+      default: 0,
+    },
+  },
+  { timestamps: true },
+);
+
+conversationTagSchema.index({ tag: 1, user: 1 }, { unique: true });
+
+module.exports = mongoose.model('ConversationTag', conversationTagSchema);

api/models/schema/convoSchema.js

@@ -42,6 +42,11 @@ const convoSchema = mongoose.Schema(
     invocationId: {
       type: Number,
     },
+    tags: {
+      type: [String],
+      default: [],
+      meiliIndex: true,
+    },
   },
   { timestamps: true },
 );
@@ -56,6 +61,7 @@ if (process.env.MEILI_HOST && process.env.MEILI_MASTER_KEY) {
 }
 
 convoSchema.index({ createdAt: 1, updatedAt: 1 });
+convoSchema.index({ conversationId: 1, user: 1 }, { unique: true });
 
 const Conversation = mongoose.models.Conversation || mongoose.model('Conversation', convoSchema);
 

api/models/schema/defaults.js

@@ -13,6 +13,11 @@ const conversationPreset = {
     type: String,
     required: false,
   },
+  // for bedrock only
+  region: {
+    type: String,
+    required: false,
+  },
   // for azureOpenAI, openAI only
   chatGptLabel: {
     type: String,
@@ -74,6 +79,13 @@ const conversationPreset = {
   resendImages: {
     type: Boolean,
   },
+  /* Anthropic only */
+  promptCache: {
+    type: Boolean,
+  },
+  system: {
+    type: String,
+  },
   // files
   resendFiles: {
     type: Boolean,
@@ -103,6 +115,10 @@ const conversationPreset = {
   spec: {
     type: String,
   },
+  tags: {
+    type: [String],
+    default: [],
+  },
   tools: { type: [{ type: String }], default: undefined },
   maxContextTokens: {
     type: Number,

api/models/schema/messageSchema.js

@@ -129,7 +129,9 @@ if (process.env.MEILI_HOST && process.env.MEILI_MASTER_KEY) {
 }
 
 messageSchema.index({ createdAt: 1 });
+messageSchema.index({ messageId: 1, user: 1 }, { unique: true });
 
+/** @type {mongoose.Model<TMessage>} */
 const Message = mongoose.models.Message || mongoose.model('Message', messageSchema);
 
 module.exports = Message;

api/models/schema/projectSchema.js

@@ -21,6 +21,11 @@ const projectSchema = new Schema(
       ref: 'PromptGroup',
       default: [],
     },
+    agentIds: {
+      type: [String],
+      ref: 'Agent',
+      default: [],
+    },
   },
   {
     timestamps: true,

api/models/schema/roleSchema.js

@@ -8,6 +8,12 @@ const roleSchema = new mongoose.Schema({
     unique: true,
     index: true,
   },
+  [PermissionTypes.BOOKMARKS]: {
+    [Permissions.USE]: {
+      type: Boolean,
+      default: true,
+    },
+  },
   [PermissionTypes.PROMPTS]: {
     [Permissions.SHARED_GLOBAL]: {
       type: Boolean,
@@ -22,6 +28,26 @@ const roleSchema = new mongoose.Schema({
       default: true,
     },
   },
+  [PermissionTypes.AGENTS]: {
+    [Permissions.SHARED_GLOBAL]: {
+      type: Boolean,
+      default: false,
+    },
+    [Permissions.USE]: {
+      type: Boolean,
+      default: true,
+    },
+    [Permissions.CREATE]: {
+      type: Boolean,
+      default: true,
+    },
+  },
+  [PermissionTypes.MULTI_CONVO]: {
+    [Permissions.USE]: {
+      type: Boolean,
+      default: true,
+    },
+  },
 });
 
 const Role = mongoose.model('Role', roleSchema);

api/models/schema/tokenSchema.js

@@ -18,8 +18,13 @@ const tokenSchema = new Schema({
     type: Date,
     required: true,
     default: Date.now,
-    expires: 900,
+  },
+  expiresAt: {
+    type: Date,
+    required: true,
   },
 });
 
-module.exports = mongoose.model('Token', tokenSchema);
+tokenSchema.index({ expiresAt: 1 }, { expireAfterSeconds: 0 });
+
+module.exports = tokenSchema;

api/models/schema/transaction.js

@@ -30,6 +30,9 @@ const transactionSchema = mongoose.Schema(
     rate: Number,
     rawAmount: Number,
     tokenValue: Number,
+    inputTokens: { type: Number },
+    writeTokens: { type: Number },
+    readTokens: { type: Number },
   },
   {
     timestamps: true,

api/models/schema/userSchema.js

@@ -122,7 +122,12 @@ const userSchema = mongoose.Schema(
       type: Date,
       expires: 604800, // 7 days in seconds
     },
+    termsAccepted: {
+      type: Boolean,
+      default: false,
+    },
   },
+
   { timestamps: true },
 );
 

api/models/spendTokens.js

@@ -11,7 +11,7 @@ const { logger } = require('~/config');
  * @param {String} txData.conversationId - The ID of the conversation.
  * @param {String} txData.model - The model name.
  * @param {String} txData.context - The context in which the transaction is made.
- * @param {String} [txData.endpointTokenConfig] - The current endpoint token config.
+ * @param {EndpointTokenConfig} [txData.endpointTokenConfig] - The current endpoint token config.
  * @param {String} [txData.valueKey] - The value key (optional).
  * @param {Object} tokenUsage - The number of tokens used.
  * @param {Number} tokenUsage.promptTokens - The number of prompt tokens used.
@@ -32,38 +32,109 @@ const spendTokens = async (txData, tokenUsage) => {
   );
   let prompt, completion;
   try {
-    if (promptTokens >= 0) {
+    if (promptTokens !== undefined) {
       prompt = await Transaction.create({
         ...txData,
         tokenType: 'prompt',
-        rawAmount: -promptTokens,
+        rawAmount: -Math.max(promptTokens, 0),
       });
     }
 
-    if (!completionTokens && isNaN(completionTokens)) {
-      logger.debug('[spendTokens] !completionTokens', { prompt, completion });
-      return;
+    if (completionTokens !== undefined) {
+      completion = await Transaction.create({
+        ...txData,
+        tokenType: 'completion',
+        rawAmount: -Math.max(completionTokens, 0),
+      });
     }
 
-    completion = await Transaction.create({
-      ...txData,
-      tokenType: 'completion',
-      rawAmount: -completionTokens,
-    });
-
-    prompt &&
-      completion &&
+    if (prompt || completion) {
       logger.debug('[spendTokens] Transaction data record against balance:', {
         user: txData.user,
-        prompt: prompt.prompt,
-        promptRate: prompt.rate,
-        completion: completion.completion,
-        completionRate: completion.rate,
-        balance: completion.balance,
+        prompt: prompt?.prompt,
+        promptRate: prompt?.rate,
+        completion: completion?.completion,
+        completionRate: completion?.rate,
+        balance: completion?.balance ?? prompt?.balance,
       });
+    } else {
+      logger.debug('[spendTokens] No transactions incurred against balance');
+    }
   } catch (err) {
     logger.error('[spendTokens]', err);
   }
 };
 
-module.exports = spendTokens;
+/**
+ * Creates transactions to record the spending of structured tokens.
+ *
+ * @function
+ * @async
+ * @param {Object} txData - Transaction data.
+ * @param {mongoose.Schema.Types.ObjectId} txData.user - The user ID.
+ * @param {String} txData.conversationId - The ID of the conversation.
+ * @param {String} txData.model - The model name.
+ * @param {String} txData.context - The context in which the transaction is made.
+ * @param {EndpointTokenConfig} [txData.endpointTokenConfig] - The current endpoint token config.
+ * @param {String} [txData.valueKey] - The value key (optional).
+ * @param {Object} tokenUsage - The number of tokens used.
+ * @param {Object} tokenUsage.promptTokens - The number of prompt tokens used.
+ * @param {Number} tokenUsage.promptTokens.input - The number of input tokens.
+ * @param {Number} tokenUsage.promptTokens.write - The number of write tokens.
+ * @param {Number} tokenUsage.promptTokens.read - The number of read tokens.
+ * @param {Number} tokenUsage.completionTokens - The number of completion tokens used.
+ * @returns {Promise<void>} - Returns nothing.
+ * @throws {Error} - Throws an error if there's an issue creating the transactions.
+ */
+const spendStructuredTokens = async (txData, tokenUsage) => {
+  const { promptTokens, completionTokens } = tokenUsage;
+  logger.debug(
+    `[spendStructuredTokens] conversationId: ${txData.conversationId}${
+      txData?.context ? ` | Context: ${txData?.context}` : ''
+    } | Token usage: `,
+    {
+      promptTokens,
+      completionTokens,
+    },
+  );
+  let prompt, completion;
+  try {
+    if (promptTokens) {
+      const { input = 0, write = 0, read = 0 } = promptTokens;
+      prompt = await Transaction.createStructured({
+        ...txData,
+        tokenType: 'prompt',
+        inputTokens: -input,
+        writeTokens: -write,
+        readTokens: -read,
+      });
+    }
+
+    if (completionTokens) {
+      completion = await Transaction.create({
+        ...txData,
+        tokenType: 'completion',
+        rawAmount: -completionTokens,
+      });
+    }
+
+    if (prompt || completion) {
+      logger.debug('[spendStructuredTokens] Transaction data record against balance:', {
+        user: txData.user,
+        prompt: prompt?.prompt,
+        promptRate: prompt?.rate,
+        completion: completion?.completion,
+        completionRate: completion?.rate,
+        balance: completion?.balance ?? prompt?.balance,
+      });
+    } else {
+      logger.debug('[spendStructuredTokens] No transactions incurred against balance');
+    }
+  } catch (err) {
+    logger.error('[spendStructuredTokens]', err);
+  }
+
+  return { prompt, completion };
+};
+
+module.exports = { spendTokens, spendStructuredTokens };

api/models/spendTokens.spec.js

@@ -0,0 +1,197 @@
+const mongoose = require('mongoose');
+
+jest.mock('./Transaction', () => ({
+  Transaction: {
+    create: jest.fn(),
+    createStructured: jest.fn(),
+  },
+}));
+
+jest.mock('./Balance', () => ({
+  findOne: jest.fn(),
+  findOneAndUpdate: jest.fn(),
+}));
+
+jest.mock('~/config', () => ({
+  logger: {
+    debug: jest.fn(),
+    error: jest.fn(),
+  },
+}));
+
+// Import after mocking
+const { spendTokens, spendStructuredTokens } = require('./spendTokens');
+const { Transaction } = require('./Transaction');
+const Balance = require('./Balance');
+describe('spendTokens', () => {
+  beforeEach(() => {
+    jest.clearAllMocks();
+    process.env.CHECK_BALANCE = 'true';
+  });
+
+  it('should create transactions for both prompt and completion tokens', async () => {
+    const txData = {
+      user: new mongoose.Types.ObjectId(),
+      conversationId: 'test-convo',
+      model: 'gpt-3.5-turbo',
+      context: 'test',
+    };
+    const tokenUsage = {
+      promptTokens: 100,
+      completionTokens: 50,
+    };
+
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'prompt', rawAmount: -100 });
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'completion', rawAmount: -50 });
+    Balance.findOne.mockResolvedValue({ tokenCredits: 10000 });
+    Balance.findOneAndUpdate.mockResolvedValue({ tokenCredits: 9850 });
+
+    await spendTokens(txData, tokenUsage);
+
+    expect(Transaction.create).toHaveBeenCalledTimes(2);
+    expect(Transaction.create).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'prompt',
+        rawAmount: -100,
+      }),
+    );
+    expect(Transaction.create).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'completion',
+        rawAmount: -50,
+      }),
+    );
+  });
+
+  it('should handle zero completion tokens', async () => {
+    const txData = {
+      user: new mongoose.Types.ObjectId(),
+      conversationId: 'test-convo',
+      model: 'gpt-3.5-turbo',
+      context: 'test',
+    };
+    const tokenUsage = {
+      promptTokens: 100,
+      completionTokens: 0,
+    };
+
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'prompt', rawAmount: -100 });
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'completion', rawAmount: -0 });
+    Balance.findOne.mockResolvedValue({ tokenCredits: 10000 });
+    Balance.findOneAndUpdate.mockResolvedValue({ tokenCredits: 9850 });
+
+    await spendTokens(txData, tokenUsage);
+
+    expect(Transaction.create).toHaveBeenCalledTimes(2);
+    expect(Transaction.create).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'prompt',
+        rawAmount: -100,
+      }),
+    );
+    expect(Transaction.create).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'completion',
+        rawAmount: -0, // Changed from 0 to -0
+      }),
+    );
+  });
+
+  it('should handle undefined token counts', async () => {
+    const txData = {
+      user: new mongoose.Types.ObjectId(),
+      conversationId: 'test-convo',
+      model: 'gpt-3.5-turbo',
+      context: 'test',
+    };
+    const tokenUsage = {};
+
+    await spendTokens(txData, tokenUsage);
+
+    expect(Transaction.create).not.toHaveBeenCalled();
+  });
+
+  it('should not update balance when CHECK_BALANCE is false', async () => {
+    process.env.CHECK_BALANCE = 'false';
+    const txData = {
+      user: new mongoose.Types.ObjectId(),
+      conversationId: 'test-convo',
+      model: 'gpt-3.5-turbo',
+      context: 'test',
+    };
+    const tokenUsage = {
+      promptTokens: 100,
+      completionTokens: 50,
+    };
+
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'prompt', rawAmount: -100 });
+    Transaction.create.mockResolvedValueOnce({ tokenType: 'completion', rawAmount: -50 });
+
+    await spendTokens(txData, tokenUsage);
+
+    expect(Transaction.create).toHaveBeenCalledTimes(2);
+    expect(Balance.findOne).not.toHaveBeenCalled();
+    expect(Balance.findOneAndUpdate).not.toHaveBeenCalled();
+  });
+
+  it('should create structured transactions for both prompt and completion tokens', async () => {
+    const txData = {
+      user: new mongoose.Types.ObjectId(),
+      conversationId: 'test-convo',
+      model: 'claude-3-5-sonnet',
+      context: 'test',
+    };
+    const tokenUsage = {
+      promptTokens: {
+        input: 10,
+        write: 100,
+        read: 5,
+      },
+      completionTokens: 50,
+    };
+
+    Transaction.createStructured.mockResolvedValueOnce({
+      rate: 3.75,
+      user: txData.user.toString(),
+      balance: 9570,
+      prompt: -430,
+    });
+    Transaction.create.mockResolvedValueOnce({
+      rate: 15,
+      user: txData.user.toString(),
+      balance: 8820,
+      completion: -750,
+    });
+
+    const result = await spendStructuredTokens(txData, tokenUsage);
+
+    expect(Transaction.createStructured).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'prompt',
+        inputTokens: -10,
+        writeTokens: -100,
+        readTokens: -5,
+      }),
+    );
+    expect(Transaction.create).toHaveBeenCalledWith(
+      expect.objectContaining({
+        tokenType: 'completion',
+        rawAmount: -50,
+      }),
+    );
+    expect(result).toEqual({
+      prompt: expect.objectContaining({
+        rate: 3.75,
+        user: txData.user.toString(),
+        balance: 9570,
+        prompt: -430,
+      }),
+      completion: expect.objectContaining({
+        rate: 15,
+        user: txData.user.toString(),
+        balance: 8820,
+        completion: -750,
+      }),
+    });
+  });
+});

api/models/tx.js

@@ -1,36 +1,77 @@
 const { matchModelName } = require('../utils');
 const defaultRate = 6;
 
+/** AWS Bedrock pricing */
+const bedrockValues = {
+  'llama2-13b': { prompt: 0.75, completion: 1.0 },
+  'llama2-70b': { prompt: 1.95, completion: 2.56 },
+  'llama3-8b': { prompt: 0.3, completion: 0.6 },
+  'llama3-70b': { prompt: 2.65, completion: 3.5 },
+  'llama3-1-8b': { prompt: 0.3, completion: 0.6 },
+  'llama3-1-70b': { prompt: 2.65, completion: 3.5 },
+  'llama3-1-405b': { prompt: 5.32, completion: 16.0 },
+  'mistral-7b': { prompt: 0.15, completion: 0.2 },
+  'mistral-small': { prompt: 0.15, completion: 0.2 },
+  'mixtral-8x7b': { prompt: 0.45, completion: 0.7 },
+  'mistral-large-2402': { prompt: 4.0, completion: 12.0 },
+  'mistral-large-2407': { prompt: 3.0, completion: 9.0 },
+  'command-text': { prompt: 1.5, completion: 2.0 },
+  'command-light': { prompt: 0.3, completion: 0.6 },
+  'ai21.j2-mid-v1': { prompt: 12.5, completion: 12.5 },
+  'ai21.j2-ultra-v1': { prompt: 18.8, completion: 18.8 },
+  'ai21.jamba-instruct-v1:0': { prompt: 0.5, completion: 0.7 },
+  'amazon.titan-text-lite-v1': { prompt: 0.15, completion: 0.2 },
+  'amazon.titan-text-express-v1': { prompt: 0.2, completion: 0.6 },
+  'amazon.titan-text-premier-v1:0': { prompt: 0.5, completion: 1.5 },
+};
+
 /**
  * Mapping of model token sizes to their respective multipliers for prompt and completion.
  * The rates are 1 USD per 1M tokens.
  * @type {Object.<string, {prompt: number, completion: number}>}
  */
-const tokenValues = {
-  '8k': { prompt: 30, completion: 60 },
-  '32k': { prompt: 60, completion: 120 },
-  '4k': { prompt: 1.5, completion: 2 },
-  '16k': { prompt: 3, completion: 4 },
-  'gpt-3.5-turbo-1106': { prompt: 1, completion: 2 },
-  'gpt-4o': { prompt: 5, completion: 15 },
-  'gpt-4-1106': { prompt: 10, completion: 30 },
-  'gpt-3.5-turbo-0125': { prompt: 0.5, completion: 1.5 },
-  'claude-3-opus': { prompt: 15, completion: 75 },
-  'claude-3-sonnet': { prompt: 3, completion: 15 },
-  'claude-3-5-sonnet': { prompt: 3, completion: 15 },
-  'claude-3-haiku': { prompt: 0.25, completion: 1.25 },
-  'claude-2.1': { prompt: 8, completion: 24 },
-  'claude-2': { prompt: 8, completion: 24 },
-  'claude-': { prompt: 0.8, completion: 2.4 },
-  'command-r-plus': { prompt: 3, completion: 15 },
-  'command-r': { prompt: 0.5, completion: 1.5 },
-  /* cohere doesn't have rates for the older command models,
+const tokenValues = Object.assign(
+  {
+    '8k': { prompt: 30, completion: 60 },
+    '32k': { prompt: 60, completion: 120 },
+    '4k': { prompt: 1.5, completion: 2 },
+    '16k': { prompt: 3, completion: 4 },
+    'gpt-3.5-turbo-1106': { prompt: 1, completion: 2 },
+    'gpt-4o-2024-08-06': { prompt: 2.5, completion: 10 },
+    'gpt-4o-mini': { prompt: 0.15, completion: 0.6 },
+    'gpt-4o': { prompt: 5, completion: 15 },
+    'gpt-4-1106': { prompt: 10, completion: 30 },
+    'gpt-3.5-turbo-0125': { prompt: 0.5, completion: 1.5 },
+    'claude-3-opus': { prompt: 15, completion: 75 },
+    'claude-3-sonnet': { prompt: 3, completion: 15 },
+    'claude-3-5-sonnet': { prompt: 3, completion: 15 },
+    'claude-3.5-sonnet': { prompt: 3, completion: 15 },
+    'claude-3-haiku': { prompt: 0.25, completion: 1.25 },
+    'claude-2.1': { prompt: 8, completion: 24 },
+    'claude-2': { prompt: 8, completion: 24 },
+    'claude-instant': { prompt: 0.8, completion: 2.4 },
+    'claude-': { prompt: 0.8, completion: 2.4 },
+    'command-r-plus': { prompt: 3, completion: 15 },
+    'command-r': { prompt: 0.5, completion: 1.5 },
+    /* cohere doesn't have rates for the older command models,
   so this was from https://artificialanalysis.ai/models/command-light/providers */
-  command: { prompt: 0.38, completion: 0.38 },
-  // 'gemini-1.5': { prompt: 7, completion: 21 }, // May 2nd, 2024 pricing
-  // 'gemini': { prompt: 0.5, completion: 1.5 }, // May 2nd, 2024 pricing
-  'gemini-1.5': { prompt: 0, completion: 0 }, // currently free
-  gemini: { prompt: 0, completion: 0 }, // currently free
+    command: { prompt: 0.38, completion: 0.38 },
+    'gemini-1.5': { prompt: 7, completion: 21 }, // May 2nd, 2024 pricing
+    gemini: { prompt: 0.5, completion: 1.5 }, // May 2nd, 2024 pricing
+  },
+  bedrockValues,
+);
+
+/**
+ * Mapping of model token sizes to their respective multipliers for cached input, read and write.
+ * See Anthropic's documentation on this: https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching#pricing
+ * The rates are 1 USD per 1M tokens.
+ * @type {Object.<string, {write: number, read: number }>}
+ */
+const cacheTokenValues = {
+  'claude-3.5-sonnet': { write: 3.75, read: 0.3 },
+  'claude-3-5-sonnet': { write: 3.75, read: 0.3 },
+  'claude-3-haiku': { write: 0.3, read: 0.03 },
 };
 
 /**
@@ -54,6 +95,10 @@ const getValueKey = (model, endpoint) => {
     return 'gpt-3.5-turbo-1106';
   } else if (modelName.includes('gpt-3.5')) {
     return '4k';
+  } else if (modelName.includes('gpt-4o-2024-08-06')) {
+    return 'gpt-4o-2024-08-06';
+  } else if (modelName.includes('gpt-4o-mini')) {
+    return 'gpt-4o-mini';
   } else if (modelName.includes('gpt-4o')) {
     return 'gpt-4o';
   } else if (modelName.includes('gpt-4-vision')) {
@@ -81,7 +126,7 @@ const getValueKey = (model, endpoint) => {
  *
  * @param {Object} params - The parameters for the function.
  * @param {string} [params.valueKey] - The key corresponding to the model name.
- * @param {string} [params.tokenType] - The type of token (e.g., 'prompt' or 'completion').
+ * @param {'prompt' | 'completion'} [params.tokenType] - The type of token (e.g., 'prompt' or 'completion').
  * @param {string} [params.model] - The model name to derive the value key from if not provided.
  * @param {string} [params.endpoint] - The endpoint name to derive the value key from if not provided.
  * @param {EndpointTokenConfig} [params.endpointTokenConfig] - The token configuration for the endpoint.
@@ -106,7 +151,41 @@ const getMultiplier = ({ valueKey, tokenType, model, endpoint, endpointTokenConf
   }
 
   // If we got this far, and values[tokenType] is undefined somehow, return a rough average of default multipliers
-  return tokenValues[valueKey][tokenType] ?? defaultRate;
+  return tokenValues[valueKey]?.[tokenType] ?? defaultRate;
 };
 
-module.exports = { tokenValues, getValueKey, getMultiplier, defaultRate };
+/**
+ * Retrieves the cache multiplier for a given value key and token type. If no value key is provided,
+ * it attempts to derive it from the model name.
+ *
+ * @param {Object} params - The parameters for the function.
+ * @param {string} [params.valueKey] - The key corresponding to the model name.
+ * @param {'write' | 'read'} [params.cacheType] - The type of token (e.g., 'write' or 'read').
+ * @param {string} [params.model] - The model name to derive the value key from if not provided.
+ * @param {string} [params.endpoint] - The endpoint name to derive the value key from if not provided.
+ * @param {EndpointTokenConfig} [params.endpointTokenConfig] - The token configuration for the endpoint.
+ * @returns {number | null} The multiplier for the given parameters, or `null` if not found.
+ */
+const getCacheMultiplier = ({ valueKey, cacheType, model, endpoint, endpointTokenConfig }) => {
+  if (endpointTokenConfig) {
+    return endpointTokenConfig?.[model]?.[cacheType] ?? null;
+  }
+
+  if (valueKey && cacheType) {
+    return cacheTokenValues[valueKey]?.[cacheType] ?? null;
+  }
+
+  if (!cacheType || !model) {
+    return null;
+  }
+
+  valueKey = getValueKey(model, endpoint);
+  if (!valueKey) {
+    return null;
+  }
+
+  // If we got this far, and values[cacheType] is undefined somehow, return a rough average of default multipliers
+  return cacheTokenValues[valueKey]?.[cacheType] ?? null;
+};
+
+module.exports = { tokenValues, getValueKey, getMultiplier, getCacheMultiplier, defaultRate };

api/models/tx.spec.js

@@ -1,4 +1,11 @@
-const { getValueKey, getMultiplier, defaultRate, tokenValues } = require('./tx');
+const { EModelEndpoint } = require('librechat-data-provider');
+const {
+  defaultRate,
+  tokenValues,
+  getValueKey,
+  getMultiplier,
+  getCacheMultiplier,
+} = require('./tx');
 
 describe('getValueKey', () => {
   it('should return "16k" for model name containing "gpt-3.5-turbo-16k"', () => {
@@ -49,12 +56,40 @@ describe('getValueKey', () => {
     expect(getValueKey('gpt-4o-0125')).toBe('gpt-4o');
   });
 
+  it('should return "gpt-4o-mini" for model type of "gpt-4o-mini"', () => {
+    expect(getValueKey('gpt-4o-mini-2024-07-18')).toBe('gpt-4o-mini');
+    expect(getValueKey('openai/gpt-4o-mini')).toBe('gpt-4o-mini');
+    expect(getValueKey('gpt-4o-mini-0718')).toBe('gpt-4o-mini');
+    expect(getValueKey('gpt-4o-2024-08-06-0718')).not.toBe('gpt-4o');
+  });
+
+  it('should return "gpt-4o-2024-08-06" for model type of "gpt-4o-2024-08-06"', () => {
+    expect(getValueKey('gpt-4o-2024-08-06-2024-07-18')).toBe('gpt-4o-2024-08-06');
+    expect(getValueKey('openai/gpt-4o-2024-08-06')).toBe('gpt-4o-2024-08-06');
+    expect(getValueKey('gpt-4o-2024-08-06-0718')).toBe('gpt-4o-2024-08-06');
+    expect(getValueKey('gpt-4o-2024-08-06-0718')).not.toBe('gpt-4o');
+  });
+
+  it('should return "gpt-4o" for model type of "chatgpt-4o"', () => {
+    expect(getValueKey('chatgpt-4o-latest')).toBe('gpt-4o');
+    expect(getValueKey('openai/chatgpt-4o-latest')).toBe('gpt-4o');
+    expect(getValueKey('chatgpt-4o-latest-0916')).toBe('gpt-4o');
+    expect(getValueKey('chatgpt-4o-latest-0718')).toBe('gpt-4o');
+  });
+
   it('should return "claude-3-5-sonnet" for model type of "claude-3-5-sonnet-"', () => {
     expect(getValueKey('claude-3-5-sonnet-20240620')).toBe('claude-3-5-sonnet');
     expect(getValueKey('anthropic/claude-3-5-sonnet')).toBe('claude-3-5-sonnet');
     expect(getValueKey('claude-3-5-sonnet-turbo')).toBe('claude-3-5-sonnet');
     expect(getValueKey('claude-3-5-sonnet-0125')).toBe('claude-3-5-sonnet');
   });
+
+  it('should return "claude-3.5-sonnet" for model type of "claude-3.5-sonnet-"', () => {
+    expect(getValueKey('claude-3.5-sonnet-20240620')).toBe('claude-3.5-sonnet');
+    expect(getValueKey('anthropic/claude-3.5-sonnet')).toBe('claude-3.5-sonnet');
+    expect(getValueKey('claude-3.5-sonnet-turbo')).toBe('claude-3.5-sonnet');
+    expect(getValueKey('claude-3.5-sonnet-0125')).toBe('claude-3.5-sonnet');
+  });
 });
 
 describe('getMultiplier', () => {
@@ -109,6 +144,30 @@ describe('getMultiplier', () => {
     );
   });
 
+  it('should return the correct multiplier for gpt-4o-mini', () => {
+    const valueKey = getValueKey('gpt-4o-mini-2024-07-18');
+    expect(getMultiplier({ valueKey, tokenType: 'prompt' })).toBe(
+      tokenValues['gpt-4o-mini'].prompt,
+    );
+    expect(getMultiplier({ valueKey, tokenType: 'completion' })).toBe(
+      tokenValues['gpt-4o-mini'].completion,
+    );
+    expect(getMultiplier({ valueKey, tokenType: 'completion' })).not.toBe(
+      tokenValues['gpt-4-1106'].completion,
+    );
+  });
+
+  it('should return the correct multiplier for chatgpt-4o-latest', () => {
+    const valueKey = getValueKey('chatgpt-4o-latest');
+    expect(getMultiplier({ valueKey, tokenType: 'prompt' })).toBe(tokenValues['gpt-4o'].prompt);
+    expect(getMultiplier({ valueKey, tokenType: 'completion' })).toBe(
+      tokenValues['gpt-4o'].completion,
+    );
+    expect(getMultiplier({ valueKey, tokenType: 'completion' })).not.toBe(
+      tokenValues['gpt-4o-mini'].completion,
+    );
+  });
+
   it('should derive the valueKey from the model if not provided for new models', () => {
     expect(
       getMultiplier({ tokenType: 'prompt', model: 'gpt-3.5-turbo-1106-some-other-info' }),
@@ -133,3 +192,125 @@ describe('getMultiplier', () => {
     );
   });
 });
+
+describe('AWS Bedrock Model Tests', () => {
+  const awsModels = [
+    'anthropic.claude-3-haiku-20240307-v1:0',
+    'anthropic.claude-3-sonnet-20240229-v1:0',
+    'anthropic.claude-3-opus-20240229-v1:0',
+    'anthropic.claude-3-5-sonnet-20240620-v1:0',
+    'anthropic.claude-v2:1',
+    'anthropic.claude-instant-v1',
+    'meta.llama2-13b-chat-v1',
+    'meta.llama2-70b-chat-v1',
+    'meta.llama3-8b-instruct-v1:0',
+    'meta.llama3-70b-instruct-v1:0',
+    'meta.llama3-1-8b-instruct-v1:0',
+    'meta.llama3-1-70b-instruct-v1:0',
+    'meta.llama3-1-405b-instruct-v1:0',
+    'mistral.mistral-7b-instruct-v0:2',
+    'mistral.mistral-small-2402-v1:0',
+    'mistral.mixtral-8x7b-instruct-v0:1',
+    'mistral.mistral-large-2402-v1:0',
+    'mistral.mistral-large-2407-v1:0',
+    'cohere.command-text-v14',
+    'cohere.command-light-text-v14',
+    'cohere.command-r-v1:0',
+    'cohere.command-r-plus-v1:0',
+    'ai21.j2-mid-v1',
+    'ai21.j2-ultra-v1',
+    'amazon.titan-text-lite-v1',
+    'amazon.titan-text-express-v1',
+  ];
+
+  it('should return the correct prompt multipliers for all models', () => {
+    const results = awsModels.map((model) => {
+      const valueKey = getValueKey(model, EModelEndpoint.bedrock);
+      const multiplier = getMultiplier({ valueKey, tokenType: 'prompt' });
+      return tokenValues[valueKey].prompt && multiplier === tokenValues[valueKey].prompt;
+    });
+    expect(results.every(Boolean)).toBe(true);
+  });
+
+  it('should return the correct completion multipliers for all models', () => {
+    const results = awsModels.map((model) => {
+      const valueKey = getValueKey(model, EModelEndpoint.bedrock);
+      const multiplier = getMultiplier({ valueKey, tokenType: 'completion' });
+      return tokenValues[valueKey].completion && multiplier === tokenValues[valueKey].completion;
+    });
+    expect(results.every(Boolean)).toBe(true);
+  });
+});
+
+describe('getCacheMultiplier', () => {
+  it('should return the correct cache multiplier for a given valueKey and cacheType', () => {
+    expect(getCacheMultiplier({ valueKey: 'claude-3-5-sonnet', cacheType: 'write' })).toBe(3.75);
+    expect(getCacheMultiplier({ valueKey: 'claude-3-5-sonnet', cacheType: 'read' })).toBe(0.3);
+    expect(getCacheMultiplier({ valueKey: 'claude-3-haiku', cacheType: 'write' })).toBe(0.3);
+    expect(getCacheMultiplier({ valueKey: 'claude-3-haiku', cacheType: 'read' })).toBe(0.03);
+  });
+
+  it('should return null if cacheType is provided but not found in cacheTokenValues', () => {
+    expect(
+      getCacheMultiplier({ valueKey: 'claude-3-5-sonnet', cacheType: 'unknownType' }),
+    ).toBeNull();
+  });
+
+  it('should derive the valueKey from the model if not provided', () => {
+    expect(getCacheMultiplier({ cacheType: 'write', model: 'claude-3-5-sonnet-20240620' })).toBe(
+      3.75,
+    );
+    expect(getCacheMultiplier({ cacheType: 'read', model: 'claude-3-haiku-20240307' })).toBe(0.03);
+  });
+
+  it('should return null if only model or cacheType is missing', () => {
+    expect(getCacheMultiplier({ cacheType: 'write' })).toBeNull();
+    expect(getCacheMultiplier({ model: 'claude-3-5-sonnet' })).toBeNull();
+  });
+
+  it('should return null if derived valueKey does not match any known patterns', () => {
+    expect(getCacheMultiplier({ cacheType: 'write', model: 'gpt-4-some-other-info' })).toBeNull();
+  });
+
+  it('should handle endpointTokenConfig if provided', () => {
+    const endpointTokenConfig = {
+      'custom-model': {
+        write: 5,
+        read: 1,
+      },
+    };
+    expect(
+      getCacheMultiplier({ model: 'custom-model', cacheType: 'write', endpointTokenConfig }),
+    ).toBe(5);
+    expect(
+      getCacheMultiplier({ model: 'custom-model', cacheType: 'read', endpointTokenConfig }),
+    ).toBe(1);
+  });
+
+  it('should return null if model is not found in endpointTokenConfig', () => {
+    const endpointTokenConfig = {
+      'custom-model': {
+        write: 5,
+        read: 1,
+      },
+    };
+    expect(
+      getCacheMultiplier({ model: 'unknown-model', cacheType: 'write', endpointTokenConfig }),
+    ).toBeNull();
+  });
+
+  it('should handle models with "bedrock/" prefix', () => {
+    expect(
+      getCacheMultiplier({
+        model: 'bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0',
+        cacheType: 'write',
+      }),
+    ).toBe(3.75);
+    expect(
+      getCacheMultiplier({
+        model: 'bedrock/anthropic.claude-3-haiku-20240307-v1:0',
+        cacheType: 'read',
+      }),
+    ).toBe(0.03);
+  });
+});

api/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@librechat/backend",
-  "version": "0.7.4-rc1",
+  "version": "v0.7.5-rc2",
   "description": "",
   "scripts": {
     "start": "echo 'please run this from the root directory'",
@@ -12,6 +12,7 @@
     "list-balances": "node ./list-balances.js",
     "user-stats": "node ./user-stats.js",
     "create-user": "node ./create-user.js",
+    "invite-user": "node ./invite-user.js",
     "ban-user": "node ./ban-user.js",
     "delete-user": "node ./delete-user.js"
   },
@@ -39,15 +40,20 @@
     "@keyv/mongo": "^2.1.8",
     "@keyv/redis": "^2.8.1",
     "@langchain/community": "^0.0.46",
+    "@langchain/core": "^0.2.18",
     "@langchain/google-genai": "^0.0.11",
     "@langchain/google-vertexai": "^0.0.17",
+    "@librechat/agents": "^1.5.2",
     "axios": "^1.3.4",
     "bcryptjs": "^2.4.3",
     "cheerio": "^1.0.0-rc.12",
     "cohere-ai": "^7.9.1",
+    "compression": "^1.7.4",
     "connect-redis": "^7.1.0",
     "cookie": "^0.5.0",
+    "cookie-parser": "^1.4.6",
     "cors": "^2.8.5",
+    "dedent": "^1.5.3",
     "dotenv": "^16.0.3",
     "express": "^4.18.2",
     "express-mongo-sanitize": "^2.2.0",
@@ -72,6 +78,7 @@
     "module-alias": "^2.2.3",
     "mongoose": "^7.1.1",
     "multer": "^1.4.5-lts.1",
+    "nanoid": "^3.3.7",
     "nodejs-gpt": "^1.37.4",
     "nodemailer": "^6.9.4",
     "ollama": "^0.5.0",
@@ -98,7 +105,8 @@
     "zod": "^3.22.4"
   },
   "devDependencies": {
-    "jest": "^29.5.0",
+    "jest": "^29.7.0",
+    "mongodb-memory-server": "^10.0.0",
     "nodemon": "^3.0.1",
     "supertest": "^6.3.3"
   }

api/server/controllers/AskController.js

@@ -1,7 +1,8 @@
 const throttle = require('lodash/throttle');
-const { getResponseSender, Constants, EModelEndpoint } = require('librechat-data-provider');
+const { getResponseSender, Constants, CacheKeys, Time } = require('librechat-data-provider');
 const { createAbortController, handleAbortError } = require('~/server/middleware');
 const { sendMessage, createOnProgress } = require('~/server/utils');
+const { getLogStores } = require('~/cache');
 const { saveMessage } = require('~/models');
 const { logger } = require('~/config');
 
@@ -51,11 +52,13 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
 
   try {
     const { client } = await initializeClient({ req, res, endpointOption });
-    const unfinished = endpointOption.endpoint === EModelEndpoint.google ? false : true;
+    const messageCache = getLogStores(CacheKeys.MESSAGES);
     const { onProgress: progressCallback, getPartialText } = createOnProgress({
       onProgress: throttle(
         ({ text: partialText }) => {
-          saveMessage({
+          /*
+              const unfinished = endpointOption.endpoint === EModelEndpoint.google ? false : true;
+          messageCache.set(responseMessageId, {
             messageId: responseMessageId,
             sender,
             conversationId,
@@ -65,7 +68,10 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
             unfinished,
             error: false,
             user,
-          });
+          }, Time.FIVE_MINUTES);
+          */
+
+          messageCache.set(responseMessageId, partialText, Time.FIVE_MINUTES);
         },
         3000,
         { trailing: false },
@@ -117,11 +123,6 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
     };
 
     let response = await client.sendMessage(text, messageOptions);
-
-    if (overrideParentMessageId) {
-      response.parentMessageId = overrideParentMessageId;
-    }
-
     response.endpoint = endpointOption.endpoint;
 
     const { conversation = {} } = await client.responsePromise;
@@ -144,11 +145,17 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
       });
       res.end();
 
-      await saveMessage({ ...response, user });
+      await saveMessage(
+        req,
+        { ...response, user },
+        { context: 'api/server/controllers/AskController.js - response end' },
+      );
     }
 
     if (!client.skipSaveUserMessage) {
-      await saveMessage(userMessage);
+      await saveMessage(req, userMessage, {
+        context: 'api/server/controllers/AskController.js - don\'t skip saving user message',
+      });
     }
 
     if (addTitle && parentMessageId === Constants.NO_PARENT && newConvo) {

api/server/controllers/AuthController.js

@@ -1,4 +1,3 @@
-const crypto = require('crypto');
 const cookies = require('cookie');
 const jwt = require('jsonwebtoken');
 const {
@@ -7,6 +6,7 @@ const {
   setAuthTokens,
   requestPasswordReset,
 } = require('~/server/services/AuthService');
+const { hashToken } = require('~/server/utils/crypto');
 const { Session, getUserById } = require('~/models');
 const { logger } = require('~/config');
 
@@ -74,8 +74,7 @@ const refreshController = async (req, res) => {
     }
 
     // Hash the refresh token
-    const hash = crypto.createHash('sha256');
-    const hashedToken = hash.update(refreshToken).digest('hex');
+    const hashedToken = await hashToken(refreshToken);
 
     // Find the session with the hashed refresh token
     const session = await Session.findOne({ user: userId, refreshTokenHash: hashedToken });

api/server/controllers/Balance.js

@@ -1,4 +1,4 @@
-const Balance = require('../../models/Balance');
+const Balance = require('~/models/Balance');
 
 async function balanceController(req, res) {
   const { tokenCredits: balance = '' } =

api/server/controllers/EditController.js

@@ -1,7 +1,8 @@
 const throttle = require('lodash/throttle');
-const { getResponseSender, EModelEndpoint } = require('librechat-data-provider');
+const { getResponseSender, CacheKeys, Time } = require('librechat-data-provider');
 const { createAbortController, handleAbortError } = require('~/server/middleware');
 const { sendMessage, createOnProgress } = require('~/server/utils');
+const { getLogStores } = require('~/cache');
 const { saveMessage } = require('~/models');
 const { logger } = require('~/config');
 
@@ -51,12 +52,14 @@ const EditController = async (req, res, next, initializeClient) => {
     }
   };
 
-  const unfinished = endpointOption.endpoint === EModelEndpoint.google ? false : true;
+  const messageCache = getLogStores(CacheKeys.MESSAGES);
   const { onProgress: progressCallback, getPartialText } = createOnProgress({
     generation,
     onProgress: throttle(
       ({ text: partialText }) => {
-        saveMessage({
+        /*
+          const unfinished = endpointOption.endpoint === EModelEndpoint.google ? false : true;
+        {
           messageId: responseMessageId,
           sender,
           conversationId,
@@ -67,7 +70,8 @@ const EditController = async (req, res, next, initializeClient) => {
           isEdited: true,
           error: false,
           user,
-        });
+        } */
+        messageCache.set(responseMessageId, partialText, Time.FIVE_MINUTES);
       },
       3000,
       { trailing: false },
@@ -141,7 +145,11 @@ const EditController = async (req, res, next, initializeClient) => {
       });
       res.end();
 
-      await saveMessage({ ...response, user });
+      await saveMessage(
+        req,
+        { ...response, user },
+        { context: 'api/server/controllers/EditController.js - response end' },
+      );
     }
   } catch (error) {
     const partialText = getPartialText();

api/server/controllers/EndpointController.js

@@ -44,6 +44,14 @@ async function endpointController(req, res) {
     };
   }
 
+  if (mergedConfig[EModelEndpoint.bedrock] && req.app.locals?.[EModelEndpoint.bedrock]) {
+    const { availableRegions } = req.app.locals[EModelEndpoint.bedrock];
+    mergedConfig[EModelEndpoint.bedrock] = {
+      ...mergedConfig[EModelEndpoint.bedrock],
+      availableRegions,
+    };
+  }
+
   const endpointsConfig = orderEndpointsConfig(mergedConfig);
 
   await cache.set(CacheKeys.ENDPOINT_CONFIG, endpointsConfig);

api/server/controllers/ModelController.js

@@ -2,6 +2,9 @@ const { CacheKeys } = require('librechat-data-provider');
 const { loadDefaultModels, loadConfigModels } = require('~/server/services/Config');
 const { getLogStores } = require('~/cache');
 
+/**
+ * @param {ServerRequest} req
+ */
 const getModelsConfig = async (req) => {
   const cache = getLogStores(CacheKeys.CONFIG_STORE);
   let modelsConfig = await cache.get(CacheKeys.MODELS_CONFIG);
@@ -14,7 +17,7 @@ const getModelsConfig = async (req) => {
 
 /**
  * Loads the models from the config.
- * @param {Express.Request} req - The Express request object.
+ * @param {ServerRequest} req - The Express request object.
  * @returns {Promise<TModelsConfig>} The models config.
  */
 async function loadModels(req) {

api/server/controllers/UserController.js

@@ -8,6 +8,7 @@ const {
   deleteMessages,
   deleteUserById,
 } = require('~/models');
+const User = require('~/models/User');
 const { updateUserPluginAuth, deleteUserPluginAuth } = require('~/server/services/PluginService');
 const { updateUserPluginsService, deleteUserKey } = require('~/server/services/UserService');
 const { verifyEmail, resendVerificationEmail } = require('~/server/services/AuthService');
@@ -20,6 +21,32 @@ const getUserController = async (req, res) => {
   res.status(200).send(req.user);
 };
 
+const getTermsStatusController = async (req, res) => {
+  try {
+    const user = await User.findById(req.user.id);
+    if (!user) {
+      return res.status(404).json({ message: 'User not found' });
+    }
+    res.status(200).json({ termsAccepted: !!user.termsAccepted });
+  } catch (error) {
+    logger.error('Error fetching terms acceptance status:', error);
+    res.status(500).json({ message: 'Error fetching terms acceptance status' });
+  }
+};
+
+const acceptTermsController = async (req, res) => {
+  try {
+    const user = await User.findByIdAndUpdate(req.user.id, { termsAccepted: true }, { new: true });
+    if (!user) {
+      return res.status(404).json({ message: 'User not found' });
+    }
+    res.status(200).json({ message: 'Terms accepted successfully' });
+  } catch (error) {
+    logger.error('Error accepting terms:', error);
+    res.status(500).json({ message: 'Error accepting terms' });
+  }
+};
+
 const deleteUserFiles = async (req) => {
   try {
     const userFiles = await getFiles({ user: req.user.id });
@@ -135,6 +162,8 @@ const resendVerificationController = async (req, res) => {
 
 module.exports = {
   getUserController,
+  getTermsStatusController,
+  acceptTermsController,
   deleteUserController,
   verifyEmailController,
   updateUserPluginsController,

api/server/controllers/agents/callbacks.js

@@ -0,0 +1,127 @@
+const { GraphEvents, ToolEndHandler, ChatModelStreamHandler } = require('@librechat/agents');
+
+/** @typedef {import('@librechat/agents').Graph} Graph */
+/** @typedef {import('@librechat/agents').EventHandler} EventHandler */
+/** @typedef {import('@librechat/agents').ModelEndData} ModelEndData */
+/** @typedef {import('@librechat/agents').ChatModelStreamHandler} ChatModelStreamHandler */
+/** @typedef {import('@librechat/agents').ContentAggregatorResult['aggregateContent']} ContentAggregator */
+/** @typedef {import('@librechat/agents').GraphEvents} GraphEvents */
+
+/**
+ * Sends message data in Server Sent Events format.
+ * @param {ServerResponse} res - The server response.
+ * @param {{ data: string | Record<string, unknown>, event?: string }} event - The message event.
+ * @param {string} event.event - The type of event.
+ * @param {string} event.data - The message to be sent.
+ */
+const sendEvent = (res, event) => {
+  if (typeof event.data === 'string' && event.data.length === 0) {
+    return;
+  }
+  res.write(`event: message\ndata: ${JSON.stringify(event)}\n\n`);
+};
+
+class ModelEndHandler {
+  /**
+   * @param {Array<UsageMetadata>} collectedUsage
+   */
+  constructor(collectedUsage) {
+    if (!Array.isArray(collectedUsage)) {
+      throw new Error('collectedUsage must be an array');
+    }
+    this.collectedUsage = collectedUsage;
+  }
+
+  /**
+   * @param {string} event
+   * @param {ModelEndData | undefined} data
+   * @param {Record<string, unknown> | undefined} metadata
+   * @param {Graph} graph
+   * @returns
+   */
+  handle(event, data, metadata, graph) {
+    if (!graph || !metadata) {
+      console.warn(`Graph or metadata not found in ${event} event`);
+      return;
+    }
+
+    const usage = data?.output?.usage_metadata;
+
+    if (usage) {
+      this.collectedUsage.push(usage);
+    }
+  }
+}
+
+/**
+ * Get default handlers for stream events.
+ * @param {Object} options - The options object.
+ * @param {ServerResponse} options.res - The options object.
+ * @param {ContentAggregator} options.aggregateContent - The options object.
+ * @param {Array<UsageMetadata>} options.collectedUsage - The list of collected usage metadata.
+ * @returns {Record<string, t.EventHandler>} The default handlers.
+ * @throws {Error} If the request is not found.
+ */
+function getDefaultHandlers({ res, aggregateContent, collectedUsage }) {
+  if (!res || !aggregateContent) {
+    throw new Error(
+      `[getDefaultHandlers] Missing required options: res: ${!res}, aggregateContent: ${!aggregateContent}`,
+    );
+  }
+  const handlers = {
+    [GraphEvents.CHAT_MODEL_END]: new ModelEndHandler(collectedUsage),
+    [GraphEvents.TOOL_END]: new ToolEndHandler(),
+    [GraphEvents.CHAT_MODEL_STREAM]: new ChatModelStreamHandler(),
+    [GraphEvents.ON_RUN_STEP]: {
+      /**
+       * Handle ON_RUN_STEP event.
+       * @param {string} event - The event name.
+       * @param {StreamEventData} data - The event data.
+       */
+      handle: (event, data) => {
+        sendEvent(res, { event, data });
+        aggregateContent({ event, data });
+      },
+    },
+    [GraphEvents.ON_RUN_STEP_DELTA]: {
+      /**
+       * Handle ON_RUN_STEP_DELTA event.
+       * @param {string} event - The event name.
+       * @param {StreamEventData} data - The event data.
+       */
+      handle: (event, data) => {
+        sendEvent(res, { event, data });
+        aggregateContent({ event, data });
+      },
+    },
+    [GraphEvents.ON_RUN_STEP_COMPLETED]: {
+      /**
+       * Handle ON_RUN_STEP_COMPLETED event.
+       * @param {string} event - The event name.
+       * @param {StreamEventData & { result: ToolEndData }} data - The event data.
+       */
+      handle: (event, data) => {
+        sendEvent(res, { event, data });
+        aggregateContent({ event, data });
+      },
+    },
+    [GraphEvents.ON_MESSAGE_DELTA]: {
+      /**
+       * Handle ON_MESSAGE_DELTA event.
+       * @param {string} event - The event name.
+       * @param {StreamEventData} data - The event data.
+       */
+      handle: (event, data) => {
+        sendEvent(res, { event, data });
+        aggregateContent({ event, data });
+      },
+    },
+  };
+
+  return handlers;
+}
+
+module.exports = {
+  sendEvent,
+  getDefaultHandlers,
+};

api/server/controllers/agents/client.js

@@ -0,0 +1,608 @@
+// const { HttpsProxyAgent } = require('https-proxy-agent');
+// const {
+// Constants,
+// ImageDetail,
+// EModelEndpoint,
+// resolveHeaders,
+// validateVisionModel,
+// mapModelToAzureConfig,
+// } = require('librechat-data-provider');
+const { Callback, createMetadataAggregator } = require('@librechat/agents');
+const {
+  Constants,
+  EModelEndpoint,
+  bedrockOutputParser,
+  providerEndpointMap,
+  removeNullishValues,
+} = require('librechat-data-provider');
+const {
+  extractBaseURL,
+  // constructAzureURL,
+  // genAzureChatCompletion,
+} = require('~/utils');
+const {
+  formatMessage,
+  formatAgentMessages,
+  createContextHandlers,
+} = require('~/app/clients/prompts');
+const { encodeAndFormat } = require('~/server/services/Files/images/encode');
+const Tokenizer = require('~/server/services/Tokenizer');
+const { spendTokens } = require('~/models/spendTokens');
+const BaseClient = require('~/app/clients/BaseClient');
+// const { sleep } = require('~/server/utils');
+const { createRun } = require('./run');
+const { logger } = require('~/config');
+
+/** @typedef {import('@librechat/agents').MessageContentComplex} MessageContentComplex */
+
+// const providerSchemas = {
+// [EModelEndpoint.bedrock]: true,
+// };
+
+const providerParsers = {
+  [EModelEndpoint.bedrock]: bedrockOutputParser,
+};
+
+class AgentClient extends BaseClient {
+  constructor(options = {}) {
+    super(null, options);
+
+    /** @type {'discard' | 'summarize'} */
+    this.contextStrategy = 'discard';
+
+    /** @deprecated @type {true} - Is a Chat Completion Request */
+    this.isChatCompletion = true;
+
+    /** @type {AgentRun} */
+    this.run;
+
+    const {
+      maxContextTokens,
+      modelOptions = {},
+      contentParts,
+      collectedUsage,
+      ...clientOptions
+    } = options;
+
+    this.modelOptions = modelOptions;
+    this.maxContextTokens = maxContextTokens;
+    /** @type {MessageContentComplex[]} */
+    this.contentParts = contentParts;
+    /** @type {Array<UsageMetadata>} */
+    this.collectedUsage = collectedUsage;
+    this.options = Object.assign({ endpoint: options.endpoint }, clientOptions);
+  }
+
+  /**
+   * Returns the aggregated content parts for the current run.
+   * @returns {MessageContentComplex[]} */
+  getContentParts() {
+    return this.contentParts;
+  }
+
+  setOptions(options) {
+    logger.info('[api/server/controllers/agents/client.js] setOptions', options);
+  }
+
+  /**
+   *
+   * Checks if the model is a vision model based on request attachments and sets the appropriate options:
+   * - Sets `this.modelOptions.model` to `gpt-4-vision-preview` if the request is a vision request.
+   * - Sets `this.isVisionModel` to `true` if vision request.
+   * - Deletes `this.modelOptions.stop` if vision request.
+   * @param {MongoFile[]} attachments
+   */
+  checkVisionRequest(attachments) {
+    logger.info(
+      '[api/server/controllers/agents/client.js #checkVisionRequest] not implemented',
+      attachments,
+    );
+    // if (!attachments) {
+    //   return;
+    // }
+
+    // const availableModels = this.options.modelsConfig?.[this.options.endpoint];
+    // if (!availableModels) {
+    //   return;
+    // }
+
+    // let visionRequestDetected = false;
+    // for (const file of attachments) {
+    //   if (file?.type?.includes('image')) {
+    //     visionRequestDetected = true;
+    //     break;
+    //   }
+    // }
+    // if (!visionRequestDetected) {
+    //   return;
+    // }
+
+    // this.isVisionModel = validateVisionModel({ model: this.modelOptions.model, availableModels });
+    // if (this.isVisionModel) {
+    //   delete this.modelOptions.stop;
+    //   return;
+    // }
+
+    // for (const model of availableModels) {
+    //   if (!validateVisionModel({ model, availableModels })) {
+    //     continue;
+    //   }
+    //   this.modelOptions.model = model;
+    //   this.isVisionModel = true;
+    //   delete this.modelOptions.stop;
+    //   return;
+    // }
+
+    // if (!availableModels.includes(this.defaultVisionModel)) {
+    //   return;
+    // }
+    // if (!validateVisionModel({ model: this.defaultVisionModel, availableModels })) {
+    //   return;
+    // }
+
+    // this.modelOptions.model = this.defaultVisionModel;
+    // this.isVisionModel = true;
+    // delete this.modelOptions.stop;
+  }
+
+  getSaveOptions() {
+    const parseOptions = providerParsers[this.options.endpoint];
+    let runOptions =
+      this.options.endpoint === EModelEndpoint.agents
+        ? {
+          model: undefined,
+          // TODO:
+          // would need to be override settings; otherwise, model needs to be undefined
+          // model: this.override.model,
+          // instructions: this.override.instructions,
+          // additional_instructions: this.override.additional_instructions,
+        }
+        : {};
+
+    if (parseOptions) {
+      runOptions = parseOptions(this.modelOptions);
+    }
+
+    return removeNullishValues(
+      Object.assign(
+        {
+          endpoint: this.options.endpoint,
+          agent_id: this.options.agent.id,
+          modelLabel: this.options.modelLabel,
+          maxContextTokens: this.options.maxContextTokens,
+          resendFiles: this.options.resendFiles,
+          imageDetail: this.options.imageDetail,
+          spec: this.options.spec,
+        },
+        // TODO: PARSE OPTIONS BY PROVIDER, MAY CONTAIN SENSITIVE DATA
+        runOptions,
+      ),
+    );
+  }
+
+  getBuildMessagesOptions(opts) {
+    return {
+      instructions: opts.instructions,
+      additional_instructions: opts.additional_instructions,
+    };
+  }
+
+  async addImageURLs(message, attachments) {
+    const { files, image_urls } = await encodeAndFormat(
+      this.options.req,
+      attachments,
+      this.options.agent.provider,
+    );
+    message.image_urls = image_urls.length ? image_urls : undefined;
+    return files;
+  }
+
+  async buildMessages(
+    messages,
+    parentMessageId,
+    { instructions = null, additional_instructions = null },
+    opts,
+  ) {
+    let orderedMessages = this.constructor.getMessagesForConversation({
+      messages,
+      parentMessageId,
+      summary: this.shouldSummarize,
+    });
+
+    let payload;
+    /** @type {{ role: string; name: string; content: string } | undefined} */
+    let systemMessage;
+    /** @type {number | undefined} */
+    let promptTokens;
+
+    /** @type {string} */
+    let systemContent = `${instructions ?? ''}${additional_instructions ?? ''}`;
+
+    if (this.options.attachments) {
+      const attachments = await this.options.attachments;
+
+      if (this.message_file_map) {
+        this.message_file_map[orderedMessages[orderedMessages.length - 1].messageId] = attachments;
+      } else {
+        this.message_file_map = {
+          [orderedMessages[orderedMessages.length - 1].messageId]: attachments,
+        };
+      }
+
+      const files = await this.addImageURLs(
+        orderedMessages[orderedMessages.length - 1],
+        attachments,
+      );
+
+      this.options.attachments = files;
+    }
+
+    if (this.message_file_map) {
+      this.contextHandlers = createContextHandlers(
+        this.options.req,
+        orderedMessages[orderedMessages.length - 1].text,
+      );
+    }
+
+    const formattedMessages = orderedMessages.map((message, i) => {
+      const formattedMessage = formatMessage({
+        message,
+        userName: this.options?.name,
+        assistantName: this.options?.modelLabel,
+      });
+
+      const needsTokenCount = this.contextStrategy && !orderedMessages[i].tokenCount;
+
+      /* If tokens were never counted, or, is a Vision request and the message has files, count again */
+      if (needsTokenCount || (this.isVisionModel && (message.image_urls || message.files))) {
+        orderedMessages[i].tokenCount = this.getTokenCountForMessage(formattedMessage);
+      }
+
+      /* If message has files, calculate image token cost */
+      // if (this.message_file_map && this.message_file_map[message.messageId]) {
+      //   const attachments = this.message_file_map[message.messageId];
+      //   for (const file of attachments) {
+      //     if (file.embedded) {
+      //       this.contextHandlers?.processFile(file);
+      //       continue;
+      //     }
+
+      //     orderedMessages[i].tokenCount += this.calculateImageTokenCost({
+      //       width: file.width,
+      //       height: file.height,
+      //       detail: this.options.imageDetail ?? ImageDetail.auto,
+      //     });
+      //   }
+      // }
+
+      return formattedMessage;
+    });
+
+    if (this.contextHandlers) {
+      this.augmentedPrompt = await this.contextHandlers.createContext();
+      systemContent = this.augmentedPrompt + systemContent;
+    }
+
+    if (systemContent) {
+      systemContent = `${systemContent.trim()}`;
+      systemMessage = {
+        role: 'system',
+        name: 'instructions',
+        content: systemContent,
+      };
+
+      if (this.contextStrategy) {
+        const instructionTokens = this.getTokenCountForMessage(systemMessage);
+        if (instructionTokens >= 0) {
+          const firstMessageTokens = orderedMessages[0].tokenCount ?? 0;
+          orderedMessages[0].tokenCount = firstMessageTokens + instructionTokens;
+        }
+      }
+    }
+
+    if (this.contextStrategy) {
+      ({ payload, promptTokens, messages } = await this.handleContextStrategy({
+        orderedMessages,
+        formattedMessages,
+        /* prefer usage_metadata from final message */
+        buildTokenMap: false,
+      }));
+    }
+
+    const result = {
+      prompt: payload,
+      promptTokens,
+      messages,
+    };
+
+    if (promptTokens >= 0 && typeof opts?.getReqData === 'function') {
+      opts.getReqData({ promptTokens });
+    }
+
+    return result;
+  }
+
+  /** @type {sendCompletion} */
+  async sendCompletion(payload, opts = {}) {
+    this.modelOptions.user = this.user;
+    await this.chatCompletion({
+      payload,
+      onProgress: opts.onProgress,
+      abortController: opts.abortController,
+    });
+    return this.contentParts;
+  }
+
+  /**
+   * @param {Object} params
+   * @param {string} [params.model]
+   * @param {string} [params.context='message']
+   * @param {UsageMetadata[]} [params.collectedUsage=this.collectedUsage]
+   */
+  async recordCollectedUsage({ model, context = 'message', collectedUsage = this.collectedUsage }) {
+    for (const usage of collectedUsage) {
+      await spendTokens(
+        {
+          context,
+          model: model ?? this.modelOptions.model,
+          conversationId: this.conversationId,
+          user: this.user ?? this.options.req.user?.id,
+          endpointTokenConfig: this.options.endpointTokenConfig,
+        },
+        { promptTokens: usage.input_tokens, completionTokens: usage.output_tokens },
+      );
+    }
+  }
+
+  async chatCompletion({ payload, abortController = null }) {
+    try {
+      if (!abortController) {
+        abortController = new AbortController();
+      }
+
+      const baseURL = extractBaseURL(this.completionsUrl);
+      logger.debug('[api/server/controllers/agents/client.js] chatCompletion', {
+        baseURL,
+        payload,
+      });
+
+      // if (this.useOpenRouter) {
+      //   opts.defaultHeaders = {
+      //     'HTTP-Referer': 'https://librechat.ai',
+      //     'X-Title': 'LibreChat',
+      //   };
+      // }
+
+      // if (this.options.headers) {
+      //   opts.defaultHeaders = { ...opts.defaultHeaders, ...this.options.headers };
+      // }
+
+      // if (this.options.proxy) {
+      //   opts.httpAgent = new HttpsProxyAgent(this.options.proxy);
+      // }
+
+      // if (this.isVisionModel) {
+      //   modelOptions.max_tokens = 4000;
+      // }
+
+      // /** @type {TAzureConfig | undefined} */
+      // const azureConfig = this.options?.req?.app?.locals?.[EModelEndpoint.azureOpenAI];
+
+      // if (
+      //   (this.azure && this.isVisionModel && azureConfig) ||
+      //   (azureConfig && this.isVisionModel && this.options.endpoint === EModelEndpoint.azureOpenAI)
+      // ) {
+      //   const { modelGroupMap, groupMap } = azureConfig;
+      //   const {
+      //     azureOptions,
+      //     baseURL,
+      //     headers = {},
+      //     serverless,
+      //   } = mapModelToAzureConfig({
+      //     modelName: modelOptions.model,
+      //     modelGroupMap,
+      //     groupMap,
+      //   });
+      //   opts.defaultHeaders = resolveHeaders(headers);
+      //   this.langchainProxy = extractBaseURL(baseURL);
+      //   this.apiKey = azureOptions.azureOpenAIApiKey;
+
+      //   const groupName = modelGroupMap[modelOptions.model].group;
+      //   this.options.addParams = azureConfig.groupMap[groupName].addParams;
+      //   this.options.dropParams = azureConfig.groupMap[groupName].dropParams;
+      //   // Note: `forcePrompt` not re-assigned as only chat models are vision models
+
+      //   this.azure = !serverless && azureOptions;
+      //   this.azureEndpoint =
+      //     !serverless && genAzureChatCompletion(this.azure, modelOptions.model, this);
+      // }
+
+      // if (this.azure || this.options.azure) {
+      //   /* Azure Bug, extremely short default `max_tokens` response */
+      //   if (!modelOptions.max_tokens && modelOptions.model === 'gpt-4-vision-preview') {
+      //     modelOptions.max_tokens = 4000;
+      //   }
+
+      //   /* Azure does not accept `model` in the body, so we need to remove it. */
+      //   delete modelOptions.model;
+
+      //   opts.baseURL = this.langchainProxy
+      //     ? constructAzureURL({
+      //       baseURL: this.langchainProxy,
+      //       azureOptions: this.azure,
+      //     })
+      //     : this.azureEndpoint.split(/(?<!\/)\/(chat|completion)\//)[0];
+
+      //   opts.defaultQuery = { 'api-version': this.azure.azureOpenAIApiVersion };
+      //   opts.defaultHeaders = { ...opts.defaultHeaders, 'api-key': this.apiKey };
+      // }
+
+      // if (process.env.OPENAI_ORGANIZATION) {
+      //   opts.organization = process.env.OPENAI_ORGANIZATION;
+      // }
+
+      // if (this.options.addParams && typeof this.options.addParams === 'object') {
+      //   modelOptions = {
+      //     ...modelOptions,
+      //     ...this.options.addParams,
+      //   };
+      //   logger.debug('[api/server/controllers/agents/client.js #chatCompletion] added params', {
+      //     addParams: this.options.addParams,
+      //     modelOptions,
+      //   });
+      // }
+
+      // if (this.options.dropParams && Array.isArray(this.options.dropParams)) {
+      //   this.options.dropParams.forEach((param) => {
+      //     delete modelOptions[param];
+      //   });
+      //   logger.debug('[api/server/controllers/agents/client.js #chatCompletion] dropped params', {
+      //     dropParams: this.options.dropParams,
+      //     modelOptions,
+      //   });
+      // }
+
+      const run = await createRun({
+        req: this.options.req,
+        agent: this.options.agent,
+        tools: this.options.tools,
+        toolMap: this.options.toolMap,
+        runId: this.responseMessageId,
+        modelOptions: this.modelOptions,
+        customHandlers: this.options.eventHandlers,
+      });
+
+      const config = {
+        configurable: {
+          provider: providerEndpointMap[this.options.agent.provider],
+          thread_id: this.conversationId,
+        },
+        run_id: this.responseMessageId,
+        signal: abortController.signal,
+        streamMode: 'values',
+        version: 'v2',
+      };
+
+      if (!run) {
+        throw new Error('Failed to create run');
+      }
+
+      this.run = run;
+
+      const messages = formatAgentMessages(payload);
+      await run.processStream({ messages }, config, {
+        [Callback.TOOL_ERROR]: (graph, error, toolId) => {
+          logger.error(
+            '[api/server/controllers/agents/client.js #chatCompletion] Tool Error',
+            error,
+            toolId,
+          );
+        },
+      });
+      this.recordCollectedUsage({ context: 'message' }).catch((err) => {
+        logger.error(
+          '[api/server/controllers/agents/client.js #chatCompletion] Error recording collected usage',
+          err,
+        );
+      });
+    } catch (err) {
+      if (!abortController.signal.aborted) {
+        logger.error(
+          '[api/server/controllers/agents/client.js #sendCompletion] Unhandled error type',
+          err,
+        );
+        throw err;
+      }
+
+      logger.warn(
+        '[api/server/controllers/agents/client.js #sendCompletion] Operation aborted',
+        err,
+      );
+    }
+  }
+
+  /**
+   *
+   * @param {Object} params
+   * @param {string} params.text
+   * @param {string} params.conversationId
+   */
+  async titleConvo({ text }) {
+    if (!this.run) {
+      throw new Error('Run not initialized');
+    }
+    const { handleLLMEnd, collected: collectedMetadata } = createMetadataAggregator();
+    const clientOptions = {};
+    const providerConfig = this.options.req.app.locals[this.options.agent.provider];
+    if (
+      providerConfig &&
+      providerConfig.titleModel &&
+      providerConfig.titleModel !== Constants.CURRENT_MODEL
+    ) {
+      clientOptions.model = providerConfig.titleModel;
+    }
+    try {
+      const titleResult = await this.run.generateTitle({
+        inputText: text,
+        contentParts: this.contentParts,
+        clientOptions,
+        chainOptions: {
+          callbacks: [
+            {
+              handleLLMEnd,
+            },
+          ],
+        },
+      });
+
+      const collectedUsage = collectedMetadata.map((item) => {
+        let input_tokens, output_tokens;
+
+        if (item.usage) {
+          input_tokens = item.usage.input_tokens || item.usage.inputTokens;
+          output_tokens = item.usage.output_tokens || item.usage.outputTokens;
+        } else if (item.tokenUsage) {
+          input_tokens = item.tokenUsage.promptTokens;
+          output_tokens = item.tokenUsage.completionTokens;
+        }
+
+        return {
+          input_tokens: input_tokens,
+          output_tokens: output_tokens,
+        };
+      });
+
+      this.recordCollectedUsage({
+        model: clientOptions.model,
+        context: 'title',
+        collectedUsage,
+      }).catch((err) => {
+        logger.error(
+          '[api/server/controllers/agents/client.js #titleConvo] Error recording collected usage',
+          err,
+        );
+      });
+
+      return titleResult.title;
+    } catch (err) {
+      logger.error('[api/server/controllers/agents/client.js #titleConvo] Error', err);
+      return;
+    }
+  }
+
+  getEncoding() {
+    return this.modelOptions.model?.includes('gpt-4o') ? 'o200k_base' : 'cl100k_base';
+  }
+
+  /**
+   * Returns the token count of a given text. It also checks and resets the tokenizers if necessary.
+   * @param {string} text - The text to get the token count for.
+   * @returns {number} The token count of the given text.
+   */
+  getTokenCount(text) {
+    const encoding = this.getEncoding();
+    return Tokenizer.getTokenCount(text, encoding);
+  }
+}
+
+module.exports = AgentClient;

api/server/controllers/agents/errors.js

@@ -0,0 +1,153 @@
+// errorHandler.js
+const { logger } = require('~/config');
+const getLogStores = require('~/cache/getLogStores');
+const { CacheKeys, ViolationTypes } = require('librechat-data-provider');
+const { recordUsage } = require('~/server/services/Threads');
+const { getConvo } = require('~/models/Conversation');
+const { sendResponse } = require('~/server/utils');
+
+/**
+ * @typedef {Object} ErrorHandlerContext
+ * @property {OpenAIClient} openai - The OpenAI client
+ * @property {string} run_id - The run ID
+ * @property {boolean} completedRun - Whether the run has completed
+ * @property {string} assistant_id - The assistant ID
+ * @property {string} conversationId - The conversation ID
+ * @property {string} parentMessageId - The parent message ID
+ * @property {string} responseMessageId - The response message ID
+ * @property {string} endpoint - The endpoint being used
+ * @property {string} cacheKey - The cache key for the current request
+ */
+
+/**
+ * @typedef {Object} ErrorHandlerDependencies
+ * @property {Express.Request} req - The Express request object
+ * @property {Express.Response} res - The Express response object
+ * @property {() => ErrorHandlerContext} getContext - Function to get the current context
+ * @property {string} [originPath] - The origin path for the error handler
+ */
+
+/**
+ * Creates an error handler function with the given dependencies
+ * @param {ErrorHandlerDependencies} dependencies - The dependencies for the error handler
+ * @returns {(error: Error) => Promise<void>} The error handler function
+ */
+const createErrorHandler = ({ req, res, getContext, originPath = '/assistants/chat/' }) => {
+  const cache = getLogStores(CacheKeys.ABORT_KEYS);
+
+  /**
+   * Handles errors that occur during the chat process
+   * @param {Error} error - The error that occurred
+   * @returns {Promise<void>}
+   */
+  return async (error) => {
+    const {
+      openai,
+      run_id,
+      endpoint,
+      cacheKey,
+      completedRun,
+      assistant_id,
+      conversationId,
+      parentMessageId,
+      responseMessageId,
+    } = getContext();
+
+    const defaultErrorMessage =
+      'The Assistant run failed to initialize. Try sending a message in a new conversation.';
+    const messageData = {
+      assistant_id,
+      conversationId,
+      parentMessageId,
+      sender: 'System',
+      user: req.user.id,
+      shouldSaveMessage: false,
+      messageId: responseMessageId,
+      endpoint,
+    };
+
+    if (error.message === 'Run cancelled') {
+      return res.end();
+    } else if (error.message === 'Request closed' && completedRun) {
+      return;
+    } else if (error.message === 'Request closed') {
+      logger.debug(`[${originPath}] Request aborted on close`);
+    } else if (/Files.*are invalid/.test(error.message)) {
+      const errorMessage = `Files are invalid, or may not have uploaded yet.${
+        endpoint === 'azureAssistants'
+          ? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
+          : ''
+      }`;
+      return sendResponse(req, res, messageData, errorMessage);
+    } else if (error?.message?.includes('string too long')) {
+      return sendResponse(
+        req,
+        res,
+        messageData,
+        'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
+      );
+    } else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
+      return sendResponse(req, res, messageData, error.message);
+    } else {
+      logger.error(`[${originPath}]`, error);
+    }
+
+    if (!openai || !run_id) {
+      return sendResponse(req, res, messageData, defaultErrorMessage);
+    }
+
+    await new Promise((resolve) => setTimeout(resolve, 2000));
+
+    try {
+      const status = await cache.get(cacheKey);
+      if (status === 'cancelled') {
+        logger.debug(`[${originPath}] Run already cancelled`);
+        return res.end();
+      }
+      await cache.delete(cacheKey);
+      // const cancelledRun = await openai.beta.threads.runs.cancel(thread_id, run_id);
+      // logger.debug(`[${originPath}] Cancelled run:`, cancelledRun);
+    } catch (error) {
+      logger.error(`[${originPath}] Error cancelling run`, error);
+    }
+
+    await new Promise((resolve) => setTimeout(resolve, 2000));
+
+    let run;
+    try {
+      // run = await openai.beta.threads.runs.retrieve(thread_id, run_id);
+      await recordUsage({
+        ...run.usage,
+        model: run.model,
+        user: req.user.id,
+        conversationId,
+      });
+    } catch (error) {
+      logger.error(`[${originPath}] Error fetching or processing run`, error);
+    }
+
+    let finalEvent;
+    try {
+      // const errorContentPart = {
+      //   text: {
+      //     value:
+      //       error?.message ?? 'There was an error processing your request. Please try again later.',
+      //   },
+      //   type: ContentTypes.ERROR,
+      // };
+
+      finalEvent = {
+        final: true,
+        conversation: await getConvo(req.user.id, conversationId),
+        // runMessages,
+      };
+    } catch (error) {
+      logger.error(`[${originPath}] Error finalizing error process`, error);
+      return sendResponse(req, res, messageData, 'The Assistant run failed');
+    }
+
+    return sendResponse(req, res, finalEvent);
+  };
+};
+
+module.exports = { createErrorHandler };

api/server/controllers/agents/llm.js

@@ -0,0 +1,106 @@
+const { HttpsProxyAgent } = require('https-proxy-agent');
+const { resolveHeaders } = require('librechat-data-provider');
+const { createLLM } = require('~/app/clients/llm');
+
+/**
+ * Initializes and returns a Language Learning Model (LLM) instance.
+ *
+ * @param {Object} options - Configuration options for the LLM.
+ * @param {string} options.model - The model identifier.
+ * @param {string} options.modelName - The specific name of the model.
+ * @param {number} options.temperature - The temperature setting for the model.
+ * @param {number} options.presence_penalty - The presence penalty for the model.
+ * @param {number} options.frequency_penalty - The frequency penalty for the model.
+ * @param {number} options.max_tokens - The maximum number of tokens for the model output.
+ * @param {boolean} options.streaming - Whether to use streaming for the model output.
+ * @param {Object} options.context - The context for the conversation.
+ * @param {number} options.tokenBuffer - The token buffer size.
+ * @param {number} options.initialMessageCount - The initial message count.
+ * @param {string} options.conversationId - The ID of the conversation.
+ * @param {string} options.user - The user identifier.
+ * @param {string} options.langchainProxy - The langchain proxy URL.
+ * @param {boolean} options.useOpenRouter - Whether to use OpenRouter.
+ * @param {Object} options.options - Additional options.
+ * @param {Object} options.options.headers - Custom headers for the request.
+ * @param {string} options.options.proxy - Proxy URL.
+ * @param {Object} options.options.req - The request object.
+ * @param {Object} options.options.res - The response object.
+ * @param {boolean} options.options.debug - Whether to enable debug mode.
+ * @param {string} options.apiKey - The API key for authentication.
+ * @param {Object} options.azure - Azure-specific configuration.
+ * @param {Object} options.abortController - The AbortController instance.
+ * @returns {Object} The initialized LLM instance.
+ */
+function initializeLLM(options) {
+  const {
+    model,
+    modelName,
+    temperature,
+    presence_penalty,
+    frequency_penalty,
+    max_tokens,
+    streaming,
+    user,
+    langchainProxy,
+    useOpenRouter,
+    options: { headers, proxy },
+    apiKey,
+    azure,
+  } = options;
+
+  const modelOptions = {
+    modelName: modelName || model,
+    temperature,
+    presence_penalty,
+    frequency_penalty,
+    user,
+  };
+
+  if (max_tokens) {
+    modelOptions.max_tokens = max_tokens;
+  }
+
+  const configOptions = {};
+
+  if (langchainProxy) {
+    configOptions.basePath = langchainProxy;
+  }
+
+  if (useOpenRouter) {
+    configOptions.basePath = 'https://openrouter.ai/api/v1';
+    configOptions.baseOptions = {
+      headers: {
+        'HTTP-Referer': 'https://librechat.ai',
+        'X-Title': 'LibreChat',
+      },
+    };
+  }
+
+  if (headers && typeof headers === 'object' && !Array.isArray(headers)) {
+    configOptions.baseOptions = {
+      headers: resolveHeaders({
+        ...headers,
+        ...configOptions?.baseOptions?.headers,
+      }),
+    };
+  }
+
+  if (proxy) {
+    configOptions.httpAgent = new HttpsProxyAgent(proxy);
+    configOptions.httpsAgent = new HttpsProxyAgent(proxy);
+  }
+
+  const llm = createLLM({
+    modelOptions,
+    configOptions,
+    openAIApiKey: apiKey,
+    azure,
+    streaming,
+  });
+
+  return llm;
+}
+
+module.exports = {
+  initializeLLM,
+};

api/server/controllers/agents/request.js

@@ -0,0 +1,142 @@
+const { Constants } = require('librechat-data-provider');
+const { createAbortController, handleAbortError } = require('~/server/middleware');
+const { sendMessage } = require('~/server/utils');
+const { saveMessage } = require('~/models');
+const { logger } = require('~/config');
+
+const AgentController = async (req, res, next, initializeClient, addTitle) => {
+  let {
+    text,
+    endpointOption,
+    conversationId,
+    parentMessageId = null,
+    overrideParentMessageId = null,
+  } = req.body;
+
+  let sender;
+  let userMessage;
+  let promptTokens;
+  let userMessageId;
+  let responseMessageId;
+  let userMessagePromise;
+
+  const newConvo = !conversationId;
+  const user = req.user.id;
+
+  const getReqData = (data = {}) => {
+    for (let key in data) {
+      if (key === 'userMessage') {
+        userMessage = data[key];
+        userMessageId = data[key].messageId;
+      } else if (key === 'userMessagePromise') {
+        userMessagePromise = data[key];
+      } else if (key === 'responseMessageId') {
+        responseMessageId = data[key];
+      } else if (key === 'promptTokens') {
+        promptTokens = data[key];
+      } else if (key === 'sender') {
+        sender = data[key];
+      } else if (!conversationId && key === 'conversationId') {
+        conversationId = data[key];
+      }
+    }
+  };
+
+  try {
+    /** @type {{ client: TAgentClient }} */
+    const { client } = await initializeClient({ req, res, endpointOption });
+
+    const getAbortData = () => ({
+      sender,
+      userMessage,
+      promptTokens,
+      conversationId,
+      userMessagePromise,
+      messageId: responseMessageId,
+      content: client.getContentParts(),
+      parentMessageId: overrideParentMessageId ?? userMessageId,
+    });
+
+    const { abortController, onStart } = createAbortController(req, res, getAbortData, getReqData);
+
+    res.on('close', () => {
+      logger.debug('[AgentController] Request closed');
+      if (!abortController) {
+        return;
+      } else if (abortController.signal.aborted) {
+        return;
+      } else if (abortController.requestCompleted) {
+        return;
+      }
+
+      abortController.abort();
+      logger.debug('[AgentController] Request aborted on close');
+    });
+
+    const messageOptions = {
+      user,
+      onStart,
+      getReqData,
+      conversationId,
+      parentMessageId,
+      abortController,
+      overrideParentMessageId,
+      progressOptions: {
+        res,
+        // parentMessageId: overrideParentMessageId || userMessageId,
+      },
+    };
+
+    let response = await client.sendMessage(text, messageOptions);
+    response.endpoint = endpointOption.endpoint;
+
+    const { conversation = {} } = await client.responsePromise;
+    conversation.title =
+      conversation && !conversation.title ? null : conversation?.title || 'New Chat';
+
+    if (client.options.attachments) {
+      userMessage.files = client.options.attachments;
+      delete userMessage.image_urls;
+    }
+
+    if (!abortController.signal.aborted) {
+      sendMessage(res, {
+        final: true,
+        conversation,
+        title: conversation.title,
+        requestMessage: userMessage,
+        responseMessage: response,
+      });
+      res.end();
+
+      await saveMessage(
+        req,
+        { ...response, user },
+        { context: 'api/server/controllers/agents/request.js - response end' },
+      );
+    }
+
+    if (!client.skipSaveUserMessage) {
+      await saveMessage(req, userMessage, {
+        context: 'api/server/controllers/agents/request.js - don\'t skip saving user message',
+      });
+    }
+
+    if (addTitle && parentMessageId === Constants.NO_PARENT && newConvo) {
+      addTitle(req, {
+        text,
+        response,
+        client,
+      });
+    }
+  } catch (error) {
+    handleAbortError(res, req, error, {
+      conversationId,
+      sender,
+      messageId: responseMessageId,
+      parentMessageId: userMessageId ?? parentMessageId,
+    });
+  }
+};
+
+module.exports = AgentController;

api/server/controllers/agents/run.js

@@ -0,0 +1,67 @@
+const { Run, Providers } = require('@librechat/agents');
+const { providerEndpointMap } = require('librechat-data-provider');
+
+/**
+ * @typedef {import('@librechat/agents').t} t
+ * @typedef {import('@librechat/agents').StreamEventData} StreamEventData
+ * @typedef {import('@librechat/agents').ClientOptions} ClientOptions
+ * @typedef {import('@librechat/agents').EventHandler} EventHandler
+ * @typedef {import('@librechat/agents').GraphEvents} GraphEvents
+ * @typedef {import('@librechat/agents').IState} IState
+ */
+
+/**
+ * Creates a new Run instance with custom handlers and configuration.
+ *
+ * @param {Object} options - The options for creating the Run instance.
+ * @param {ServerRequest} [options.req] - The server request.
+ * @param {string | undefined} [options.runId] - Optional run ID; otherwise, a new run ID will be generated.
+ * @param {Agent} options.agent - The agent for this run.
+ * @param {StructuredTool[] | undefined} [options.tools] - The tools to use in the run.
+ * @param {Record<string, StructuredTool[]> | undefined} [options.toolMap] - The tool map for the run.
+ * @param {Record<GraphEvents, EventHandler> | undefined} [options.customHandlers] - Custom event handlers.
+ * @param {ClientOptions} [options.modelOptions] - Optional model to use; if not provided, it will use the default from modelMap.
+ * @param {boolean} [options.streaming=true] - Whether to use streaming.
+ * @param {boolean} [options.streamUsage=true] - Whether to stream usage information.
+ * @returns {Promise<Run<IState>>} A promise that resolves to a new Run instance.
+ */
+async function createRun({
+  runId,
+  tools,
+  agent,
+  toolMap,
+  modelOptions,
+  customHandlers,
+  streaming = true,
+  streamUsage = true,
+}) {
+  const llmConfig = Object.assign(
+    {
+      provider: providerEndpointMap[agent.provider],
+      streaming,
+      streamUsage,
+    },
+    modelOptions,
+  );
+
+  const graphConfig = {
+    runId,
+    llmConfig,
+    tools,
+    toolMap,
+    instructions: agent.instructions,
+    additional_instructions: agent.additional_instructions,
+  };
+
+  // TEMPORARY FOR TESTING
+  if (agent.provider === Providers.ANTHROPIC) {
+    graphConfig.streamBuffer = 2000;
+  }
+
+  return Run.create({
+    graphConfig,
+    customHandlers,
+  });
+}
+
+module.exports = { createRun };

api/server/controllers/agents/v1.js

@@ -0,0 +1,239 @@
+const { nanoid } = require('nanoid');
+const { FileContext, Constants } = require('librechat-data-provider');
+const {
+  getAgent,
+  createAgent,
+  updateAgent,
+  deleteAgent,
+  getListAgents,
+} = require('~/models/Agent');
+const { getStrategyFunctions } = require('~/server/services/Files/strategies');
+const { uploadImageBuffer } = require('~/server/services/Files/process');
+const { getProjectByName } = require('~/models/Project');
+const { updateAgentProjects } = require('~/models/Agent');
+const { deleteFileByFilter } = require('~/models/File');
+const { logger } = require('~/config');
+
+/**
+ * Creates an Agent.
+ * @route POST /Agents
+ * @param {ServerRequest} req - The request object.
+ * @param {AgentCreateParams} req.body - The request body.
+ * @param {ServerResponse} res - The response object.
+ * @returns {Agent} 201 - success response - application/json
+ */
+const createAgentHandler = async (req, res) => {
+  try {
+    const { tools = [], provider, name, description, instructions, model, ...agentData } = req.body;
+    const { id: userId } = req.user;
+
+    agentData.tools = tools
+      .map((tool) => (typeof tool === 'string' ? req.app.locals.availableTools[tool] : tool))
+      .filter(Boolean);
+
+    Object.assign(agentData, {
+      author: userId,
+      name,
+      description,
+      instructions,
+      provider,
+      model,
+    });
+
+    agentData.id = `agent_${nanoid()}`;
+    const agent = await createAgent(agentData);
+    res.status(201).json(agent);
+  } catch (error) {
+    logger.error('[/Agents] Error creating agent', error);
+    res.status(500).json({ error: error.message });
+  }
+};
+
+/**
+ * Retrieves an Agent by ID.
+ * @route GET /Agents/:id
+ * @param {object} req - Express Request
+ * @param {object} req.params - Request params
+ * @param {string} req.params.id - Agent identifier.
+ * @param {object} req.user - Authenticated user information
+ * @param {string} req.user.id - User ID
+ * @returns {Promise<Agent>} 200 - success response - application/json
+ * @returns {Error} 404 - Agent not found
+ */
+const getAgentHandler = async (req, res) => {
+  try {
+    const id = req.params.id;
+    const author = req.user.id;
+
+    let query = { id, author };
+
+    const globalProject = await getProjectByName(Constants.GLOBAL_PROJECT_NAME, ['agentIds']);
+    if (globalProject && (globalProject.agentIds?.length ?? 0) > 0) {
+      query = {
+        $or: [{ id, $in: globalProject.agentIds }, query],
+      };
+    }
+
+    const agent = await getAgent(query);
+
+    if (!agent) {
+      return res.status(404).json({ error: 'Agent not found' });
+    }
+
+    if (agent.author !== author) {
+      delete agent.author;
+    }
+
+    return res.status(200).json(agent);
+  } catch (error) {
+    logger.error('[/Agents/:id] Error retrieving agent', error);
+    res.status(500).json({ error: error.message });
+  }
+};
+
+/**
+ * Updates an Agent.
+ * @route PATCH /Agents/:id
+ * @param {object} req - Express Request
+ * @param {object} req.params - Request params
+ * @param {string} req.params.id - Agent identifier.
+ * @param {AgentUpdateParams} req.body - The Agent update parameters.
+ * @returns {Agent} 200 - success response - application/json
+ */
+const updateAgentHandler = async (req, res) => {
+  try {
+    const id = req.params.id;
+    const { projectIds, removeProjectIds, ...updateData } = req.body;
+
+    let updatedAgent;
+    if (Object.keys(updateData).length > 0) {
+      updatedAgent = await updateAgent({ id, author: req.user.id }, updateData);
+    }
+
+    if (projectIds || removeProjectIds) {
+      updatedAgent = await updateAgentProjects(id, projectIds, removeProjectIds);
+    }
+
+    return res.json(updatedAgent);
+  } catch (error) {
+    logger.error('[/Agents/:id] Error updating Agent', error);
+    res.status(500).json({ error: error.message });
+  }
+};
+
+/**
+ * Deletes an Agent based on the provided ID.
+ * @route DELETE /Agents/:id
+ * @param {object} req - Express Request
+ * @param {object} req.params - Request params
+ * @param {string} req.params.id - Agent identifier.
+ * @returns {Agent} 200 - success response - application/json
+ */
+const deleteAgentHandler = async (req, res) => {
+  try {
+    const id = req.params.id;
+    const agent = await getAgent({ id });
+    if (!agent) {
+      return res.status(404).json({ error: 'Agent not found' });
+    }
+    await deleteAgent({ id, author: req.user.id });
+    return res.json({ message: 'Agent deleted' });
+  } catch (error) {
+    logger.error('[/Agents/:id] Error deleting Agent', error);
+    res.status(500).json({ error: error.message });
+  }
+};
+
+/**
+ *
+ * @route GET /Agents
+ * @param {object} req - Express Request
+ * @param {object} req.query - Request query
+ * @param {string} [req.query.user] - The user ID of the agent's author.
+ * @returns {Promise<AgentListResponse>} 200 - success response - application/json
+ */
+const getListAgentsHandler = async (req, res) => {
+  try {
+    const data = await getListAgents({
+      author: req.user.id,
+    });
+    return res.json(data);
+  } catch (error) {
+    logger.error('[/Agents] Error listing Agents', error);
+    res.status(500).json({ error: error.message });
+  }
+};
+
+/**
+ * Uploads and updates an avatar for a specific agent.
+ * @route POST /avatar/:agent_id
+ * @param {object} req - Express Request
+ * @param {object} req.params - Request params
+ * @param {string} req.params.agent_id - The ID of the agent.
+ * @param {Express.Multer.File} req.file - The avatar image file.
+ * @param {object} req.body - Request body
+ * @param {string} [req.body.avatar] - Optional avatar for the agent's avatar.
+ * @returns {Object} 200 - success response - application/json
+ */
+const uploadAgentAvatarHandler = async (req, res) => {
+  try {
+    const { agent_id } = req.params;
+    if (!agent_id) {
+      return res.status(400).json({ message: 'Agent ID is required' });
+    }
+
+    let { avatar: _avatar = '{}' } = req.body;
+
+    const image = await uploadImageBuffer({
+      req,
+      context: FileContext.avatar,
+      metadata: {
+        buffer: req.file.buffer,
+      },
+    });
+
+    try {
+      _avatar = JSON.parse(_avatar);
+    } catch (error) {
+      logger.error('[/avatar/:agent_id] Error parsing avatar', error);
+      _avatar = {};
+    }
+
+    if (_avatar && _avatar.source) {
+      const { deleteFile } = getStrategyFunctions(_avatar.source);
+      try {
+        await deleteFile(req, { filepath: _avatar.filepath });
+        await deleteFileByFilter({ filepath: _avatar.filepath });
+      } catch (error) {
+        logger.error('[/avatar/:agent_id] Error deleting old avatar', error);
+      }
+    }
+
+    const promises = [];
+
+    const data = {
+      avatar: {
+        filepath: image.filepath,
+        source: req.app.locals.fileStrategy,
+      },
+    };
+
+    promises.push(await updateAgent({ id: agent_id, author: req.user.id }, data));
+
+    const resolved = await Promise.all(promises);
+    res.status(201).json(resolved[0]);
+  } catch (error) {
+    const message = 'An error occurred while updating the Agent Avatar';
+    logger.error(message, error);
+    res.status(500).json({ message });
+  }
+};
+
+module.exports = {
+  createAgent: createAgentHandler,
+  getAgent: getAgentHandler,
+  updateAgent: updateAgentHandler,
+  deleteAgent: deleteAgentHandler,
+  getListAgents: getListAgentsHandler,
+  uploadAgentAvatar: uploadAgentAvatarHandler,
+};

api/server/controllers/assistants/chatV1.js

@@ -54,6 +54,7 @@ const chatV1 = async (req, res) => {
     promptPrefix,
     assistant_id,
     instructions,
+    endpointOption,
     thread_id: _thread_id,
     messageId: _messageId,
     conversationId: convoId,
@@ -120,21 +121,22 @@ const chatV1 = async (req, res) => {
           ? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
           : ''
       }`;
-      return sendResponse(res, messageData, errorMessage);
+      return sendResponse(req, res, messageData, errorMessage);
     } else if (error?.message?.includes('string too long')) {
       return sendResponse(
+        req,
         res,
         messageData,
         'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
       );
     } else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
-      return sendResponse(res, messageData, error.message);
+      return sendResponse(req, res, messageData, error.message);
     } else {
       logger.error('[/assistants/chat/]', error);
     }
 
     if (!openai || !thread_id || !run_id) {
-      return sendResponse(res, messageData, defaultErrorMessage);
+      return sendResponse(req, res, messageData, defaultErrorMessage);
     }
 
     await sleep(2000);
@@ -221,10 +223,10 @@ const chatV1 = async (req, res) => {
       };
     } catch (error) {
       logger.error('[/assistants/chat/] Error finalizing error process', error);
-      return sendResponse(res, messageData, 'The Assistant run failed');
+      return sendResponse(req, res, messageData, 'The Assistant run failed');
     }
 
-    return sendResponse(res, finalEvent);
+    return sendResponse(req, res, finalEvent);
   };
 
   try {
@@ -282,7 +284,7 @@ const chatV1 = async (req, res) => {
     const { openai: _openai, client } = await getOpenAIClient({
       req,
       res,
-      endpointOption: req.body.endpointOption,
+      endpointOption,
       initAppClient: true,
     });
 
@@ -311,6 +313,10 @@ const chatV1 = async (req, res) => {
       body.additional_instructions = promptPrefix;
     }
 
+    if (typeof endpointOption.artifactsPrompt === 'string' && endpointOption.artifactsPrompt) {
+      body.additional_instructions = `${body.additional_instructions ?? ''}\n${endpointOption.artifactsPrompt}`.trim();
+    }
+
     if (instructions) {
       body.instructions = instructions;
     }
@@ -332,12 +338,12 @@ const chatV1 = async (req, res) => {
     };
 
     const addVisionPrompt = async () => {
-      if (!req.body.endpointOption.attachments) {
+      if (!endpointOption.attachments) {
         return;
       }
 
       /** @type {MongoFile[]} */
-      const attachments = await req.body.endpointOption.attachments;
+      const attachments = await endpointOption.attachments;
       if (attachments && attachments.every((attachment) => checkOpenAIStorage(attachment.source))) {
         return;
       }
@@ -382,6 +388,9 @@ const chatV1 = async (req, res) => {
       return files;
     };
 
+    /** @type {Promise<Run>|undefined} */
+    let userMessagePromise;
+
     const initializeThread = async () => {
       /** @type {[ undefined | MongoFile[]]}*/
       const [processedFiles] = await Promise.all([addVisionPrompt(), getRequestFileIds()]);
@@ -438,7 +447,7 @@ const chatV1 = async (req, res) => {
       previousMessages.push(requestMessage);
 
       /* asynchronous */
-      saveUserMessage({ ...requestMessage, model });
+      userMessagePromise = saveUserMessage(req, { ...requestMessage, model });
 
       conversation = {
         conversationId,
@@ -582,7 +591,10 @@ const chatV1 = async (req, res) => {
     });
     res.end();
 
-    await saveAssistantMessage({ ...responseMessage, model });
+    if (userMessagePromise) {
+      await userMessagePromise;
+    }
+    await saveAssistantMessage(req, { ...responseMessage, model });
 
     if (parentMessageId === Constants.NO_PARENT && !_thread_id) {
       addTitle(req, {

api/server/controllers/assistants/chatV2.js

@@ -1,12 +1,12 @@
 const { v4 } = require('uuid');
 const {
+  Time,
   Constants,
   RunStatus,
   CacheKeys,
   ContentTypes,
   ToolCallTypes,
   EModelEndpoint,
-  ViolationTypes,
   retrievalMimeTypes,
   AssistantStreamEvents,
 } = require('librechat-data-provider');
@@ -14,12 +14,12 @@ const {
   initThread,
   recordUsage,
   saveUserMessage,
-  checkMessageGaps,
   addThreadMetadata,
   saveAssistantMessage,
 } = require('~/server/services/Threads');
-const { sendResponse, sendMessage, sleep, isEnabled, countTokens } = require('~/server/utils');
 const { runAssistant, createOnTextProgress } = require('~/server/services/AssistantService');
+const { sendMessage, sleep, isEnabled, countTokens } = require('~/server/utils');
+const { createErrorHandler } = require('~/server/controllers/assistants/errors');
 const validateAuthor = require('~/server/middleware/assistants/validateAuthor');
 const { createRun, StreamRunManager } = require('~/server/services/Runs');
 const { addTitle } = require('~/server/services/Endpoints/assistants');
@@ -44,7 +44,7 @@ const ten_minutes = 1000 * 60 * 10;
 const chatV2 = async (req, res) => {
   logger.debug('[/assistants/chat/] req.body', req.body);
 
-  /** @type {{ files: MongoFile[]}} */
+  /** @type {{files: MongoFile[]}} */
   const {
     text,
     model,
@@ -53,6 +53,7 @@ const chatV2 = async (req, res) => {
     promptPrefix,
     assistant_id,
     instructions,
+    endpointOption,
     thread_id: _thread_id,
     messageId: _messageId,
     conversationId: convoId,
@@ -90,139 +91,20 @@ const chatV2 = async (req, res) => {
   /** @type {Run | undefined} - The completed run, undefined if incomplete */
   let completedRun;
 
-  const handleError = async (error) => {
-    const defaultErrorMessage =
-      'The Assistant run failed to initialize. Try sending a message in a new conversation.';
-    const messageData = {
-      thread_id,
-      assistant_id,
-      conversationId,
-      parentMessageId,
-      sender: 'System',
-      user: req.user.id,
-      shouldSaveMessage: false,
-      messageId: responseMessageId,
-      endpoint,
-    };
+  const getContext = () => ({
+    openai,
+    run_id,
+    endpoint,
+    cacheKey,
+    thread_id,
+    completedRun,
+    assistant_id,
+    conversationId,
+    parentMessageId,
+    responseMessageId,
+  });
 
-    if (error.message === 'Run cancelled') {
-      return res.end();
-    } else if (error.message === 'Request closed' && completedRun) {
-      return;
-    } else if (error.message === 'Request closed') {
-      logger.debug('[/assistants/chat/] Request aborted on close');
-    } else if (/Files.*are invalid/.test(error.message)) {
-      const errorMessage = `Files are invalid, or may not have uploaded yet.${
-        endpoint === EModelEndpoint.azureAssistants
-          ? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
-          : ''
-      }`;
-      return sendResponse(res, messageData, errorMessage);
-    } else if (error?.message?.includes('string too long')) {
-      return sendResponse(
-        res,
-        messageData,
-        'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
-      );
-    } else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
-      return sendResponse(res, messageData, error.message);
-    } else {
-      logger.error('[/assistants/chat/]', error);
-    }
-
-    if (!openai || !thread_id || !run_id) {
-      return sendResponse(res, messageData, defaultErrorMessage);
-    }
-
-    await sleep(2000);
-
-    try {
-      const status = await cache.get(cacheKey);
-      if (status === 'cancelled') {
-        logger.debug('[/assistants/chat/] Run already cancelled');
-        return res.end();
-      }
-      await cache.delete(cacheKey);
-      const cancelledRun = await openai.beta.threads.runs.cancel(thread_id, run_id);
-      logger.debug('[/assistants/chat/] Cancelled run:', cancelledRun);
-    } catch (error) {
-      logger.error('[/assistants/chat/] Error cancelling run', error);
-    }
-
-    await sleep(2000);
-
-    let run;
-    try {
-      run = await openai.beta.threads.runs.retrieve(thread_id, run_id);
-      await recordUsage({
-        ...run.usage,
-        model: run.model,
-        user: req.user.id,
-        conversationId,
-      });
-    } catch (error) {
-      logger.error('[/assistants/chat/] Error fetching or processing run', error);
-    }
-
-    let finalEvent;
-    try {
-      const runMessages = await checkMessageGaps({
-        openai,
-        run_id,
-        endpoint,
-        thread_id,
-        conversationId,
-        latestMessageId: responseMessageId,
-      });
-
-      const errorContentPart = {
-        text: {
-          value:
-            error?.message ?? 'There was an error processing your request. Please try again later.',
-        },
-        type: ContentTypes.ERROR,
-      };
-
-      if (!Array.isArray(runMessages[runMessages.length - 1]?.content)) {
-        runMessages[runMessages.length - 1].content = [errorContentPart];
-      } else {
-        const contentParts = runMessages[runMessages.length - 1].content;
-        for (let i = 0; i < contentParts.length; i++) {
-          const currentPart = contentParts[i];
-          /** @type {CodeToolCall | RetrievalToolCall | FunctionToolCall | undefined} */
-          const toolCall = currentPart?.[ContentTypes.TOOL_CALL];
-          if (
-            toolCall &&
-            toolCall?.function &&
-            !(toolCall?.function?.output || toolCall?.function?.output?.length)
-          ) {
-            contentParts[i] = {
-              ...currentPart,
-              [ContentTypes.TOOL_CALL]: {
-                ...toolCall,
-                function: {
-                  ...toolCall.function,
-                  output: 'error processing tool',
-                },
-              },
-            };
-          }
-        }
-        runMessages[runMessages.length - 1].content.push(errorContentPart);
-      }
-
-      finalEvent = {
-        final: true,
-        conversation: await getConvo(req.user.id, conversationId),
-        runMessages,
-      };
-    } catch (error) {
-      logger.error('[/assistants/chat/] Error finalizing error process', error);
-      return sendResponse(res, messageData, 'The Assistant run failed');
-    }
-
-    return sendResponse(res, finalEvent);
-  };
+  const handleError = createErrorHandler({ req, res, getContext });
 
   try {
     res.on('close', async () => {
@@ -279,7 +161,7 @@ const chatV2 = async (req, res) => {
     const { openai: _openai, client } = await getOpenAIClient({
       req,
       res,
-      endpointOption: req.body.endpointOption,
+      endpointOption,
       initAppClient: true,
     });
 
@@ -313,6 +195,10 @@ const chatV2 = async (req, res) => {
       body.additional_instructions = promptPrefix;
     }
 
+    if (typeof endpointOption.artifactsPrompt === 'string' && endpointOption.artifactsPrompt) {
+      body.additional_instructions = `${body.additional_instructions ?? ''}\n${endpointOption.artifactsPrompt}`.trim();
+    }
+
     if (instructions) {
       body.instructions = instructions;
     }
@@ -365,6 +251,9 @@ const chatV2 = async (req, res) => {
       }
     };
 
+    /** @type {Promise<Run>|undefined} */
+    let userMessagePromise;
+
     const initializeThread = async () => {
       await getRequestFileIds();
 
@@ -407,7 +296,7 @@ const chatV2 = async (req, res) => {
       previousMessages.push(requestMessage);
 
       /* asynchronous */
-      saveUserMessage({ ...requestMessage, model });
+      userMessagePromise = saveUserMessage(req, { ...requestMessage, model });
 
       conversation = {
         conversationId,
@@ -489,6 +378,11 @@ const chatV2 = async (req, res) => {
         },
       };
 
+      /** @type {undefined | TAssistantEndpoint} */
+      const config = req.app.locals[endpoint] ?? {};
+      /** @type {undefined | TBaseEndpoint} */
+      const allConfig = req.app.locals.all;
+
       const streamRunManager = new StreamRunManager({
         req,
         res,
@@ -498,6 +392,7 @@ const chatV2 = async (req, res) => {
         attachedFileIds,
         parentMessageId: userMessageId,
         responseMessage: openai.responseMessage,
+        streamRate: allConfig?.streamRate ?? config.streamRate,
         // streamOptions: {
 
         // },
@@ -510,6 +405,16 @@ const chatV2 = async (req, res) => {
 
       response = streamRunManager;
       response.text = streamRunManager.intermediateText;
+
+      const messageCache = getLogStores(CacheKeys.MESSAGES);
+      messageCache.set(
+        responseMessageId,
+        {
+          complete: true,
+          text: response.text,
+        },
+        Time.FIVE_MINUTES,
+      );
     };
 
     await processRun();
@@ -552,7 +457,10 @@ const chatV2 = async (req, res) => {
     });
     res.end();
 
-    await saveAssistantMessage({ ...responseMessage, model });
+    if (userMessagePromise) {
+      await userMessagePromise;
+    }
+    await saveAssistantMessage(req, { ...responseMessage, model });
 
     if (parentMessageId === Constants.NO_PARENT && !_thread_id) {
       addTitle(req, {

api/server/controllers/assistants/errors.js

@@ -0,0 +1,193 @@
+// errorHandler.js
+const { sendResponse } = require('~/server/utils');
+const { logger } = require('~/config');
+const getLogStores = require('~/cache/getLogStores');
+const { CacheKeys, ViolationTypes, ContentTypes } = require('librechat-data-provider');
+const { getConvo } = require('~/models/Conversation');
+const { recordUsage, checkMessageGaps } = require('~/server/services/Threads');
+
+/**
+ * @typedef {Object} ErrorHandlerContext
+ * @property {OpenAIClient} openai - The OpenAI client
+ * @property {string} thread_id - The thread ID
+ * @property {string} run_id - The run ID
+ * @property {boolean} completedRun - Whether the run has completed
+ * @property {string} assistant_id - The assistant ID
+ * @property {string} conversationId - The conversation ID
+ * @property {string} parentMessageId - The parent message ID
+ * @property {string} responseMessageId - The response message ID
+ * @property {string} endpoint - The endpoint being used
+ * @property {string} cacheKey - The cache key for the current request
+ */
+
+/**
+ * @typedef {Object} ErrorHandlerDependencies
+ * @property {Express.Request} req - The Express request object
+ * @property {Express.Response} res - The Express response object
+ * @property {() => ErrorHandlerContext} getContext - Function to get the current context
+ * @property {string} [originPath] - The origin path for the error handler
+ */
+
+/**
+ * Creates an error handler function with the given dependencies
+ * @param {ErrorHandlerDependencies} dependencies - The dependencies for the error handler
+ * @returns {(error: Error) => Promise<void>} The error handler function
+ */
+const createErrorHandler = ({ req, res, getContext, originPath = '/assistants/chat/' }) => {
+  const cache = getLogStores(CacheKeys.ABORT_KEYS);
+
+  /**
+   * Handles errors that occur during the chat process
+   * @param {Error} error - The error that occurred
+   * @returns {Promise<void>}
+   */
+  return async (error) => {
+    const {
+      openai,
+      run_id,
+      endpoint,
+      cacheKey,
+      thread_id,
+      completedRun,
+      assistant_id,
+      conversationId,
+      parentMessageId,
+      responseMessageId,
+    } = getContext();
+
+    const defaultErrorMessage =
+      'The Assistant run failed to initialize. Try sending a message in a new conversation.';
+    const messageData = {
+      thread_id,
+      assistant_id,
+      conversationId,
+      parentMessageId,
+      sender: 'System',
+      user: req.user.id,
+      shouldSaveMessage: false,
+      messageId: responseMessageId,
+      endpoint,
+    };
+
+    if (error.message === 'Run cancelled') {
+      return res.end();
+    } else if (error.message === 'Request closed' && completedRun) {
+      return;
+    } else if (error.message === 'Request closed') {
+      logger.debug(`[${originPath}] Request aborted on close`);
+    } else if (/Files.*are invalid/.test(error.message)) {
+      const errorMessage = `Files are invalid, or may not have uploaded yet.${
+        endpoint === 'azureAssistants'
+          ? ' If using Azure OpenAI, files are only available in the region of the assistant\'s model at the time of upload.'
+          : ''
+      }`;
+      return sendResponse(req, res, messageData, errorMessage);
+    } else if (error?.message?.includes('string too long')) {
+      return sendResponse(
+        req,
+        res,
+        messageData,
+        'Message too long. The Assistants API has a limit of 32,768 characters per message. Please shorten it and try again.',
+      );
+    } else if (error?.message?.includes(ViolationTypes.TOKEN_BALANCE)) {
+      return sendResponse(req, res, messageData, error.message);
+    } else {
+      logger.error(`[${originPath}]`, error);
+    }
+
+    if (!openai || !thread_id || !run_id) {
+      return sendResponse(req, res, messageData, defaultErrorMessage);
+    }
+
+    await new Promise((resolve) => setTimeout(resolve, 2000));
+
+    try {
+      const status = await cache.get(cacheKey);
+      if (status === 'cancelled') {
+        logger.debug(`[${originPath}] Run already cancelled`);
+        return res.end();
+      }
+      await cache.delete(cacheKey);
+      const cancelledRun = await openai.beta.threads.runs.cancel(thread_id, run_id);
+      logger.debug(`[${originPath}] Cancelled run:`, cancelledRun);
+    } catch (error) {
+      logger.error(`[${originPath}] Error cancelling run`, error);
+    }
+
+    await new Promise((resolve) => setTimeout(resolve, 2000));
+
+    let run;
+    try {
+      run = await openai.beta.threads.runs.retrieve(thread_id, run_id);
+      await recordUsage({
+        ...run.usage,
+        model: run.model,
+        user: req.user.id,
+        conversationId,
+      });
+    } catch (error) {
+      logger.error(`[${originPath}] Error fetching or processing run`, error);
+    }
+
+    let finalEvent;
+    try {
+      const runMessages = await checkMessageGaps({
+        openai,
+        run_id,
+        endpoint,
+        thread_id,
+        conversationId,
+        latestMessageId: responseMessageId,
+      });
+
+      const errorContentPart = {
+        text: {
+          value:
+            error?.message ?? 'There was an error processing your request. Please try again later.',
+        },
+        type: ContentTypes.ERROR,
+      };
+
+      if (!Array.isArray(runMessages[runMessages.length - 1]?.content)) {
+        runMessages[runMessages.length - 1].content = [errorContentPart];
+      } else {
+        const contentParts = runMessages[runMessages.length - 1].content;
+        for (let i = 0; i < contentParts.length; i++) {
+          const currentPart = contentParts[i];
+          /** @type {CodeToolCall | RetrievalToolCall | FunctionToolCall | undefined} */
+          const toolCall = currentPart?.[ContentTypes.TOOL_CALL];
+          if (
+            toolCall &&
+            toolCall?.function &&
+            !(toolCall?.function?.output || toolCall?.function?.output?.length)
+          ) {
+            contentParts[i] = {
+              ...currentPart,
+              [ContentTypes.TOOL_CALL]: {
+                ...toolCall,
+                function: {
+                  ...toolCall.function,
+                  output: 'error processing tool',
+                },
+              },
+            };
+          }
+        }
+        runMessages[runMessages.length - 1].content.push(errorContentPart);
+      }
+
+      finalEvent = {
+        final: true,
+        conversation: await getConvo(req.user.id, conversationId),
+        runMessages,
+      };
+    } catch (error) {
+      logger.error(`[${originPath}] Error finalizing error process`, error);
+      return sendResponse(req, res, messageData, 'The Assistant run failed');
+    }
+
+    return sendResponse(req, res, finalEvent);
+  };
+};
+
+module.exports = { createErrorHandler };

api/server/controllers/assistants/helpers.js

@@ -64,7 +64,7 @@ const _listAssistants = async ({ req, res, version, query }) => {
  * @param {object} params.res - The response object, used for initializing the client.
  * @param {string} params.version - The API version to use.
  * @param {Omit<AssistantListParams, 'endpoint'>} params.query - The query parameters to list assistants (e.g., limit, order).
- * @returns {Promise<object>} A promise that resolves to the response from the `openai.beta.assistants.list` method call.
+ * @returns {Promise<Array<Assistant>>} A promise that resolves to the response from the `openai.beta.assistants.list` method call.
  */
 const listAllAssistants = async ({ req, res, version, query }) => {
   /** @type {{ openai: OpenAIClient }} */

api/server/controllers/assistants/v1.js

@@ -18,7 +18,9 @@ const createAssistant = async (req, res) => {
   try {
     const { openai } = await getOpenAIClient({ req, res });
 
-    const { tools = [], endpoint, ...assistantData } = req.body;
+    const { tools = [], endpoint, conversation_starters, ...assistantData } = req.body;
+    delete assistantData.conversation_starters;
+
     assistantData.tools = tools
       .map((tool) => {
         if (typeof tool !== 'string') {
@@ -41,11 +43,22 @@ const createAssistant = async (req, res) => {
     };
 
     const assistant = await openai.beta.assistants.create(assistantData);
-    const promise = updateAssistantDoc({ assistant_id: assistant.id }, { user: req.user.id });
+
+    const createData = { user: req.user.id };
+    if (conversation_starters) {
+      createData.conversation_starters = conversation_starters;
+    }
+
+    const document = await updateAssistantDoc({ assistant_id: assistant.id }, createData);
+
     if (azureModelIdentifier) {
       assistant.model = azureModelIdentifier;
     }
-    await promise;
+
+    if (document.conversation_starters) {
+      assistant.conversation_starters = document.conversation_starters;
+    }
+
     logger.debug('/assistants/', assistant);
     res.status(201).json(assistant);
   } catch (error) {
@@ -88,7 +101,7 @@ const patchAssistant = async (req, res) => {
     await validateAuthor({ req, openai });
 
     const assistant_id = req.params.id;
-    const { endpoint: _e, ...updateData } = req.body;
+    const { endpoint: _e, conversation_starters, ...updateData } = req.body;
     updateData.tools = (updateData.tools ?? [])
       .map((tool) => {
         if (typeof tool !== 'string') {
@@ -104,6 +117,15 @@ const patchAssistant = async (req, res) => {
     }
 
     const updatedAssistant = await openai.beta.assistants.update(assistant_id, updateData);
+
+    if (conversation_starters !== undefined) {
+      const conversationStartersUpdate = await updateAssistantDoc(
+        { assistant_id },
+        { conversation_starters },
+      );
+      updatedAssistant.conversation_starters = conversationStartersUpdate.conversation_starters;
+    }
+
     res.json(updatedAssistant);
   } catch (error) {
     logger.error('[/assistants/:id] Error updating assistant', error);
@@ -153,6 +175,32 @@ const listAssistants = async (req, res) => {
   }
 };
 
+/**
+ * Filter assistants based on configuration.
+ *
+ * @param {object} params - The parameters object.
+ * @param {string} params.userId -  The user ID to filter private assistants.
+ * @param {AssistantDocument[]} params.assistants - The list of assistants to filter.
+ * @param {Partial<TAssistantEndpoint>} [params.assistantsConfig] -  The assistant configuration.
+ * @returns {AssistantDocument[]} - The filtered list of assistants.
+ */
+function filterAssistantDocs({ documents, userId, assistantsConfig = {} }) {
+  const { supportedIds, excludedIds, privateAssistants } = assistantsConfig;
+  const removeUserId = (doc) => {
+    const { user: _u, ...document } = doc;
+    return document;
+  };
+
+  if (privateAssistants) {
+    return documents.filter((doc) => userId === doc.user.toString()).map(removeUserId);
+  } else if (supportedIds?.length) {
+    return documents.filter((doc) => supportedIds.includes(doc.assistant_id)).map(removeUserId);
+  } else if (excludedIds?.length) {
+    return documents.filter((doc) => !excludedIds.includes(doc.assistant_id)).map(removeUserId);
+  }
+  return documents.map(removeUserId);
+}
+
 /**
  * Returns a list of the user's assistant documents (metadata saved to database).
  * @route GET /assistants/documents
@@ -160,7 +208,25 @@ const listAssistants = async (req, res) => {
  */
 const getAssistantDocuments = async (req, res) => {
   try {
-    res.json(await getAssistants({ user: req.user.id }));
+    const endpoint = req.query;
+    const assistantsConfig = req.app.locals[endpoint];
+    const documents = await getAssistants(
+      {},
+      {
+        user: 1,
+        assistant_id: 1,
+        conversation_starters: 1,
+        createdAt: 1,
+        updatedAt: 1,
+      },
+    );
+
+    const docs = filterAssistantDocs({
+      documents,
+      userId: req.user.id,
+      assistantsConfig,
+    });
+    res.json(docs);
   } catch (error) {
     logger.error('[/assistants/documents] Error listing assistant documents', error);
     res.status(500).json({ error: error.message });

api/server/controllers/assistants/v2.js

@@ -16,7 +16,9 @@ const createAssistant = async (req, res) => {
     /** @type {{ openai: OpenAIClient }} */
     const { openai } = await getOpenAIClient({ req, res });
 
-    const { tools = [], endpoint, ...assistantData } = req.body;
+    const { tools = [], endpoint, conversation_starters, ...assistantData } = req.body;
+    delete assistantData.conversation_starters;
+
     assistantData.tools = tools
       .map((tool) => {
         if (typeof tool !== 'string') {
@@ -39,11 +41,22 @@ const createAssistant = async (req, res) => {
     };
 
     const assistant = await openai.beta.assistants.create(assistantData);
-    const promise = updateAssistantDoc({ assistant_id: assistant.id }, { user: req.user.id });
+
+    const createData = { user: req.user.id };
+    if (conversation_starters) {
+      createData.conversation_starters = conversation_starters;
+    }
+
+    const document = await updateAssistantDoc({ assistant_id: assistant.id }, createData);
+
     if (azureModelIdentifier) {
       assistant.model = azureModelIdentifier;
     }
-    await promise;
+
+    if (document.conversation_starters) {
+      assistant.conversation_starters = document.conversation_starters;
+    }
+
     logger.debug('/assistants/', assistant);
     res.status(201).json(assistant);
   } catch (error) {
@@ -64,6 +77,17 @@ const createAssistant = async (req, res) => {
 const updateAssistant = async ({ req, openai, assistant_id, updateData }) => {
   await validateAuthor({ req, openai });
   const tools = [];
+  let conversation_starters = null;
+
+  if (updateData?.conversation_starters) {
+    const conversationStartersUpdate = await updateAssistantDoc(
+      { assistant_id: assistant_id },
+      { conversation_starters: updateData.conversation_starters },
+    );
+    conversation_starters = conversationStartersUpdate.conversation_starters;
+
+    delete updateData.conversation_starters;
+  }
 
   let hasFileSearch = false;
   for (const tool of updateData.tools ?? []) {
@@ -108,7 +132,13 @@ const updateAssistant = async ({ req, openai, assistant_id, updateData }) => {
     updateData.model = openai.locals.azureOptions.azureOpenAIApiDeploymentName;
   }
 
-  return await openai.beta.assistants.update(assistant_id, updateData);
+  const assistant = await openai.beta.assistants.update(assistant_id, updateData);
+
+  if (conversation_starters) {
+    assistant.conversation_starters = conversation_starters;
+  }
+
+  return assistant;
 };
 
 /**

api/server/index.js

@@ -4,8 +4,11 @@ require('module-alias')({ base: path.resolve(__dirname, '..') });
 const cors = require('cors');
 const axios = require('axios');
 const express = require('express');
+const compression = require('compression');
 const passport = require('passport');
 const mongoSanitize = require('express-mongo-sanitize');
+const fs = require('fs');
+const cookieParser = require('cookie-parser');
 const { jwtLogin, passportLogin } = require('~/strategies');
 const { connectDb, indexSync } = require('~/lib/db');
 const { isEnabled } = require('~/server/utils');
@@ -15,10 +18,11 @@ const validateImageRequest = require('./middleware/validateImageRequest');
 const errorController = require('./controllers/ErrorController');
 const configureSocialLogins = require('./socialLogins');
 const AppService = require('./services/AppService');
+const staticCache = require('./utils/staticCache');
 const noIndex = require('./middleware/noIndex');
 const routes = require('./routes');
 
-const { PORT, HOST, ALLOW_SOCIAL_LOGIN } = process.env ?? {};
+const { PORT, HOST, ALLOW_SOCIAL_LOGIN, DISABLE_COMPRESSION } = process.env ?? {};
 
 const port = Number(PORT) || 3080;
 const host = HOST || 'localhost';
@@ -35,19 +39,27 @@ const startServer = async () => {
   app.disable('x-powered-by');
   await AppService(app);
 
+  const indexPath = path.join(app.locals.paths.dist, 'index.html');
+  const indexHTML = fs.readFileSync(indexPath, 'utf8');
+
   app.get('/health', (_req, res) => res.status(200).send('OK'));
 
-  // Middleware
+  /* Middleware */
   app.use(noIndex);
   app.use(errorController);
   app.use(express.json({ limit: '3mb' }));
   app.use(mongoSanitize());
   app.use(express.urlencoded({ extended: true, limit: '3mb' }));
-  app.use(express.static(app.locals.paths.dist));
-  app.use(express.static(app.locals.paths.fonts));
-  app.use(express.static(app.locals.paths.assets));
-  app.set('trust proxy', 1); // trust first proxy
+  app.use(staticCache(app.locals.paths.dist));
+  app.use(staticCache(app.locals.paths.fonts));
+  app.use(staticCache(app.locals.paths.assets));
+  app.set('trust proxy', 1); /* trust first proxy */
   app.use(cors());
+  app.use(cookieParser());
+
+  if (!isEnabled(DISABLE_COMPRESSION)) {
+    app.use(compression());
+  }
 
   if (!ALLOW_SOCIAL_LOGIN) {
     console.warn(
@@ -55,12 +67,12 @@ const startServer = async () => {
     );
   }
 
-  // OAUTH
+  /* OAUTH */
   app.use(passport.initialize());
   passport.use(await jwtLogin());
   passport.use(passportLogin());
 
-  // LDAP Auth
+  /* LDAP Auth */
   if (process.env.LDAP_URL && process.env.LDAP_USER_SEARCH_BASE) {
     passport.use(ldapLogin);
   }
@@ -70,7 +82,7 @@ const startServer = async () => {
   }
 
   app.use('/oauth', routes.oauth);
-  // API Endpoints
+  /* API Endpoints */
   app.use('/api/auth', routes.auth);
   app.use('/api/keys', routes.keys);
   app.use('/api/user', routes.user);
@@ -93,9 +105,16 @@ const startServer = async () => {
   app.use('/images/', validateImageRequest, routes.staticRoute);
   app.use('/api/share', routes.share);
   app.use('/api/roles', routes.roles);
+  app.use('/api/agents', routes.agents);
+  app.use('/api/bedrock', routes.bedrock);
+
+  app.use('/api/tags', routes.tags);
 
   app.use((req, res) => {
-    res.sendFile(path.join(app.locals.paths.dist, 'index.html'));
+    // Replace lang attribute in index.html with lang from cookies or accept-language header
+    const lang = req.cookies.lang || req.headers['accept-language']?.split(',')[0] || 'en-US';
+    const updatedIndexHtml = indexHTML.replace(/lang="en-US"/g, `lang="${lang}"`);
+    res.send(updatedIndexHtml);
   });
 
   app.listen(port, host, () => {

api/server/middleware/abortMiddleware.js

@@ -1,10 +1,10 @@
-const { isAssistantsEndpoint } = require('librechat-data-provider');
+const { isAssistantsEndpoint, ErrorTypes } = require('librechat-data-provider');
 const { sendMessage, sendError, countTokens, isEnabled } = require('~/server/utils');
 const { truncateText, smartTruncateText } = require('~/app/clients/prompts');
 const clearPendingReq = require('~/cache/clearPendingReq');
+const { spendTokens } = require('~/models/spendTokens');
 const abortControllers = require('./abortControllers');
 const { saveMessage, getConvo } = require('~/models');
-const spendTokens = require('~/models/spendTokens');
 const { abortRun } = require('./abortRun');
 const { logger } = require('~/config');
 
@@ -30,7 +30,10 @@ async function abortMessage(req, res) {
     return res.status(204).send({ message: 'Request not found' });
   }
   const finalEvent = await abortController.abortCompletion();
-  logger.info('[abortMessage] Aborted request', { abortKey });
+  logger.debug(
+    `[abortMessage] ID: ${req.user.id} | ${req.user.email} | Aborted request: ` +
+      JSON.stringify({ abortKey }),
+  );
   abortControllers.delete(abortKey);
 
   if (res.headersSent && finalEvent) {
@@ -104,7 +107,7 @@ const createAbortController = (req, res, getAbortData, getReqData) => {
       finish_reason: 'incomplete',
       endpoint: endpointOption.endpoint,
       iconURL: endpointOption.iconURL,
-      model: endpointOption.modelOptions.model,
+      model: endpointOption.modelOptions?.model ?? endpointOption.model_parameters?.model,
       unfinished: false,
       error: false,
       isCreatedByUser: false,
@@ -116,7 +119,11 @@ const createAbortController = (req, res, getAbortData, getReqData) => {
       { promptTokens, completionTokens },
     );
 
-    saveMessage({ ...responseMessage, user });
+    saveMessage(
+      req,
+      { ...responseMessage, user },
+      { context: 'api/server/middleware/abortMiddleware.js' },
+    );
 
     let conversation;
     if (userMessagePromise) {
@@ -158,10 +165,14 @@ const handleAbortError = async (res, req, error, data) => {
     );
   }
 
-  const errorText = error?.message?.includes('"type"')
+  let errorText = error?.message?.includes('"type"')
     ? error.message
     : 'An error occurred while processing your request. Please contact the Admin.';
 
+  if (error?.type === ErrorTypes.INVALID_REQUEST) {
+    errorText = `{"type":"${ErrorTypes.INVALID_REQUEST}"}`;
+  }
+
   const respondWithError = async (partialText) => {
     let options = {
       sender,
@@ -190,7 +201,7 @@ const handleAbortError = async (res, req, error, data) => {
       }
     };
 
-    await sendError(res, options, callback);
+    await sendError(req, res, options, callback);
   };
 
   if (partialText && partialText.length > 5) {

api/server/middleware/assistants/validate.js

@@ -19,7 +19,8 @@ const validateAssistant = async (req, res, next) => {
   }
 
   const { supportedIds, excludedIds } = assistantsConfig;
-  const error = { message: 'Assistant not supported' };
+  const error = { message: 'validateAssistant: Assistant not supported' };
+
   if (supportedIds?.length && !supportedIds.includes(assistant_id)) {
     return await handleAbortError(res, req, error, {
       sender: 'System',

api/server/middleware/buildEndpointOption.js

@@ -1,11 +1,13 @@
-const { parseConvo, EModelEndpoint } = require('librechat-data-provider');
+const { parseCompactConvo, EModelEndpoint, isAgentsEndpoint } = require('librechat-data-provider');
 const { getModelsConfig } = require('~/server/controllers/ModelController');
 const azureAssistants = require('~/server/services/Endpoints/azureAssistants');
 const assistants = require('~/server/services/Endpoints/assistants');
 const gptPlugins = require('~/server/services/Endpoints/gptPlugins');
 const { processFiles } = require('~/server/services/Files/process');
 const anthropic = require('~/server/services/Endpoints/anthropic');
+const bedrock = require('~/server/services/Endpoints/bedrock');
 const openAI = require('~/server/services/Endpoints/openAI');
+const agents = require('~/server/services/Endpoints/agents');
 const custom = require('~/server/services/Endpoints/custom');
 const google = require('~/server/services/Endpoints/google');
 const enforceModelSpec = require('./enforceModelSpec');
@@ -15,6 +17,8 @@ const buildFunction = {
   [EModelEndpoint.openAI]: openAI.buildOptions,
   [EModelEndpoint.google]: google.buildOptions,
   [EModelEndpoint.custom]: custom.buildOptions,
+  [EModelEndpoint.agents]: agents.buildOptions,
+  [EModelEndpoint.bedrock]: bedrock.buildOptions,
   [EModelEndpoint.azureOpenAI]: openAI.buildOptions,
   [EModelEndpoint.anthropic]: anthropic.buildOptions,
   [EModelEndpoint.gptPlugins]: gptPlugins.buildOptions,
@@ -24,7 +28,7 @@ const buildFunction = {
 
 async function buildEndpointOption(req, res, next) {
   const { endpoint, endpointType } = req.body;
-  const parsedBody = parseConvo({ endpoint, endpointType, conversation: req.body });
+  const parsedBody = parseCompactConvo({ endpoint, endpointType, conversation: req.body });
 
   if (req.app.locals.modelSpecs?.list && req.app.locals.modelSpecs?.enforce) {
     /** @type {{ list: TModelSpec[] }}*/
@@ -59,12 +63,13 @@ async function buildEndpointOption(req, res, next) {
     }
   }
 
-  req.body.endpointOption = buildFunction[endpointType ?? endpoint](
-    endpoint,
-    parsedBody,
-    endpointType,
-  );
+  const endpointFn = buildFunction[endpointType ?? endpoint];
+  const builder = isAgentsEndpoint(endpoint) ? (...args) => endpointFn(req, ...args) : endpointFn;
 
+  // TODO: use object params
+  req.body.endpointOption = builder(endpoint, parsedBody, endpointType);
+
+  // TODO: use `getModelsConfig` only when necessary
   const modelsConfig = await getModelsConfig(req);
   req.body.endpointOption.modelsConfig = modelsConfig;
 

api/server/middleware/checkInviteUser.js

@@ -0,0 +1,27 @@
+const { getInvite } = require('~/models/inviteUser');
+const { deleteTokens } = require('~/models/Token');
+
+async function checkInviteUser(req, res, next) {
+  const token = req.body.token;
+
+  if (!token || token === 'undefined') {
+    next();
+    return;
+  }
+
+  try {
+    const invite = await getInvite(token, req.body.email);
+
+    if (!invite || invite.error === true) {
+      return res.status(400).json({ message: 'Invalid invite token' });
+    }
+
+    await deleteTokens({ token: invite.token });
+    req.invite = invite;
+    next();
+  } catch (error) {
+    return res.status(429).json({ message: error.message });
+  }
+}
+
+module.exports = checkInviteUser;

api/server/middleware/concurrentLimiter.js

@@ -1,5 +1,7 @@
-const clearPendingReq = require('../../cache/clearPendingReq');
-const { logViolation, getLogStores } = require('../../cache');
+const { Time } = require('librechat-data-provider');
+const clearPendingReq = require('~/cache/clearPendingReq');
+const { logViolation, getLogStores } = require('~/cache');
+const { isEnabled } = require('~/server/utils');
 const denyRequest = require('./denyRequest');
 
 const {
@@ -7,7 +9,6 @@ const {
   CONCURRENT_MESSAGE_MAX = 1,
   CONCURRENT_VIOLATION_SCORE: score,
 } = process.env ?? {};
-const ttl = 1000 * 60 * 1;
 
 /**
  * Middleware to limit concurrent requests for a user.
@@ -38,7 +39,7 @@ const concurrentLimiter = async (req, res, next) => {
   const limit = Math.max(CONCURRENT_MESSAGE_MAX, 1);
   const type = 'concurrent';
 
-  const key = `${USE_REDIS ? namespace : ''}:${userId}`;
+  const key = `${isEnabled(USE_REDIS) ? namespace : ''}:${userId}`;
   const pendingRequests = +((await cache.get(key)) ?? 0);
 
   if (pendingRequests >= limit) {
@@ -51,7 +52,7 @@ const concurrentLimiter = async (req, res, next) => {
     await logViolation(req, res, type, errorMessage, score);
     return await denyRequest(req, res, errorMessage);
   } else {
-    await cache.set(key, pendingRequests + 1, ttl);
+    await cache.set(key, pendingRequests + 1, Time.ONE_MINUTE);
   }
 
   // Ensure the requests are removed from the store once the request is done

api/server/middleware/denyRequest.js

@@ -41,10 +41,14 @@ const denyRequest = async (req, res, errorMessage) => {
   const shouldSaveMessage = _convoId && parentMessageId && parentMessageId !== Constants.NO_PARENT;
 
   if (shouldSaveMessage) {
-    await saveMessage({ ...userMessage, user: req.user.id });
+    await saveMessage(
+      req,
+      { ...userMessage, user: req.user.id },
+      { context: `api/server/middleware/denyRequest.js - ${responseText}` },
+    );
   }
 
-  return await sendError(res, {
+  return await sendError(req, res, {
     sender: getResponseSender(req.body),
     messageId: crypto.randomUUID(),
     conversationId,