chore: revert removal of unused vars for versioning

- Add findPubliclyAccessibleResources function to PermissionService for bulk public resource queries
  - Add hasPublicPermission function to PermissionService for individual resource public permission checks
  - Update getAgentHandler to use hasPublicPermission for accurate individual agent public status
  - Replace instanceProjectId-based global checks with isPublic property from backend in client code
  - Add isPublic property to Agent type definition
  - Add NODE_TLS_REJECT_UNAUTHORIZED debug setting to VS Code launch config

.env.example

@@ -58,7 +58,7 @@ DEBUG_CONSOLE=false
 #                     Endpoints                     #
 #===================================================#
 
-# ENDPOINTS=openAI,assistants,azureOpenAI,google,anthropic
+# ENDPOINTS=openAI,assistants,azureOpenAI,google,gptPlugins,anthropic
 
 PROXY=
 
@@ -142,10 +142,10 @@ GOOGLE_KEY=user_provided
 # GOOGLE_AUTH_HEADER=true
 
 # Gemini API (AI Studio)
-# GOOGLE_MODELS=gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite-preview-06-17,gemini-2.0-flash,gemini-2.0-flash-lite
+# GOOGLE_MODELS=gemini-2.5-pro-preview-05-06,gemini-2.5-flash-preview-04-17,gemini-2.0-flash-001,gemini-2.0-flash-exp,gemini-2.0-flash-lite-001,gemini-1.5-pro-002,gemini-1.5-flash-002
 
 # Vertex AI
-# GOOGLE_MODELS=gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite-preview-06-17,gemini-2.0-flash-001,gemini-2.0-flash-lite-001
+# GOOGLE_MODELS=gemini-2.5-pro-preview-05-06,gemini-2.5-flash-preview-04-17,gemini-2.0-flash-001,gemini-2.0-flash-exp,gemini-2.0-flash-lite-001,gemini-1.5-pro-002,gemini-1.5-flash-002
 
 # GOOGLE_TITLE_MODEL=gemini-2.0-flash-lite-001
 
@@ -453,8 +453,8 @@ OPENID_REUSE_TOKENS=
 OPENID_JWKS_URL_CACHE_ENABLED=
 OPENID_JWKS_URL_CACHE_TIME= # 600000 ms eq to 10 minutes leave empty to disable caching
 #Set to true to trigger token exchange flow to acquire access token for the userinfo endpoint.
-OPENID_ON_BEHALF_FLOW_FOR_USERINFO_REQUIRED=
-OPENID_ON_BEHALF_FLOW_USERINFO_SCOPE="user.read" # example for Scope Needed for Microsoft Graph API
+OPENID_ON_BEHALF_FLOW_FOR_USERINFRO_REQUIRED=
+OPENID_ON_BEHALF_FLOW_USERINFRO_SCOPE = "user.read" # example for Scope Needed for Microsoft Graph API
 # Set to true to use the OpenID Connect end session endpoint for logout
 OPENID_USE_END_SESSION_ENDPOINT=
 
@@ -485,6 +485,21 @@ SAML_IMAGE_URL=
 # SAML_USE_AUTHN_RESPONSE_SIGNED=
 
 
+#===============================================#
+# Microsoft Graph API / Entra ID Integration  #
+#===============================================#
+
+# Enable Entra ID people search integration in permissions/sharing system
+# When enabled, the people picker will search both local database and Entra ID
+USE_ENTRA_ID_FOR_PEOPLE_SEARCH=false
+
+# When enabled, entra id groups owners will be considered as members of the group
+ENTRA_ID_INCLUDE_OWNERS_AS_MEMBERS=false
+
+# Microsoft Graph API scopes needed for people/group search
+# Default scopes provide access to user profiles and group memberships
+OPENID_GRAPH_SCOPES=User.Read,People.Read,GroupMember.Read.All
+
 # LDAP
 LDAP_URL=
 LDAP_BIND_DN=
@@ -657,4 +672,4 @@ OPENWEATHER_API_KEY=
 # Reranker (Required)
 # JINA_API_KEY=your_jina_api_key
 # or
-# COHERE_API_KEY=your_cohere_api_key
+# COHERE_API_KEY=your_cohere_api_key

.vscode/launch.json

@@ -8,7 +8,8 @@
       "skipFiles": ["<node_internals>/**"],
       "program": "${workspaceFolder}/api/server/index.js",
       "env": {
-        "NODE_ENV": "production"
+        "NODE_ENV": "production",
+         "NODE_TLS_REJECT_UNAUTHORIZED": "0"
       },
       "console": "integratedTerminal",
       "envFile": "${workspaceFolder}/.env"

api/app/clients/BaseClient.js

@@ -792,8 +792,7 @@ class BaseClient {
 
     userMessage.tokenCount = userMessageTokenCount;
     /*
-      Note: `AgentController` saves the user message if not saved here
-      (noted by `savedMessageIds`), so we update the count of its `userMessage` reference
+      Note: `AskController` saves the user message, so we update the count of its `userMessage` reference
     */
     if (typeof opts?.getReqData === 'function') {
       opts.getReqData({
@@ -802,8 +801,7 @@ class BaseClient {
     }
     /*
       Note: we update the user message to be sure it gets the calculated token count;
-      though `AgentController` saves the user message if not saved here
-      (noted by `savedMessageIds`), EditController does not
+      though `AskController` saves the user message, EditController does not
     */
     await userMessagePromise;
     await this.updateMessageInDatabase({

api/app/clients/ChatGPTClient.js

@@ -0,0 +1,804 @@
+const { Keyv } = require('keyv');
+const crypto = require('crypto');
+const { CohereClient } = require('cohere-ai');
+const { fetchEventSource } = require('@waylaidwanderer/fetch-event-source');
+const { constructAzureURL, genAzureChatCompletion } = require('@librechat/api');
+const { encoding_for_model: encodingForModel, get_encoding: getEncoding } = require('tiktoken');
+const {
+  ImageDetail,
+  EModelEndpoint,
+  resolveHeaders,
+  CohereConstants,
+  mapModelToAzureConfig,
+} = require('librechat-data-provider');
+const { createContextHandlers } = require('./prompts');
+const { createCoherePayload } = require('./llm');
+const { extractBaseURL } = require('~/utils');
+const BaseClient = require('./BaseClient');
+const { logger } = require('~/config');
+
+const CHATGPT_MODEL = 'gpt-3.5-turbo';
+const tokenizersCache = {};
+
+class ChatGPTClient extends BaseClient {
+  constructor(apiKey, options = {}, cacheOptions = {}) {
+    super(apiKey, options, cacheOptions);
+
+    cacheOptions.namespace = cacheOptions.namespace || 'chatgpt';
+    this.conversationsCache = new Keyv(cacheOptions);
+    this.setOptions(options);
+  }
+
+  setOptions(options) {
+    if (this.options && !this.options.replaceOptions) {
+      // nested options aren't spread properly, so we need to do this manually
+      this.options.modelOptions = {
+        ...this.options.modelOptions,
+        ...options.modelOptions,
+      };
+      delete options.modelOptions;
+      // now we can merge options
+      this.options = {
+        ...this.options,
+        ...options,
+      };
+    } else {
+      this.options = options;
+    }
+
+    if (this.options.openaiApiKey) {
+      this.apiKey = this.options.openaiApiKey;
+    }
+
+    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 || CHATGPT_MODEL,
+      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,
+    };
+
+    this.isChatGptModel = this.modelOptions.model.includes('gpt-');
+    const { isChatGptModel } = this;
+    this.isUnofficialChatGptModel =
+      this.modelOptions.model.startsWith('text-chat') ||
+      this.modelOptions.model.startsWith('text-davinci-002-render');
+    const { isUnofficialChatGptModel } = this;
+
+    // Davinci models have a max context length of 4097 tokens.
+    this.maxContextTokens = this.options.maxContextTokens || (isChatGptModel ? 4095 : 4097);
+    // I decided to reserve 1024 tokens for the response.
+    // The max prompt tokens is determined by the max context tokens minus the max response tokens.
+    // Earlier messages will be dropped until the prompt is within the limit.
+    this.maxResponseTokens = this.modelOptions.max_tokens || 1024;
+    this.maxPromptTokens =
+      this.options.maxPromptTokens || this.maxContextTokens - this.maxResponseTokens;
+
+    if (this.maxPromptTokens + this.maxResponseTokens > this.maxContextTokens) {
+      throw new Error(
+        `maxPromptTokens + max_tokens (${this.maxPromptTokens} + ${this.maxResponseTokens} = ${
+          this.maxPromptTokens + this.maxResponseTokens
+        }) must be less than or equal to maxContextTokens (${this.maxContextTokens})`,
+      );
+    }
+
+    this.userLabel = this.options.userLabel || 'User';
+    this.chatGptLabel = this.options.chatGptLabel || 'ChatGPT';
+
+    if (isChatGptModel) {
+      // Use these faux tokens to help the AI understand the context since we are building the chat log ourselves.
+      // Trying to use "<|im_start|>" causes the AI to still generate "<" or "<|" at the end sometimes for some reason,
+      // without tripping the stop sequences, so I'm using "||>" instead.
+      this.startToken = '||>';
+      this.endToken = '';
+      this.gptEncoder = this.constructor.getTokenizer('cl100k_base');
+    } else if (isUnofficialChatGptModel) {
+      this.startToken = '<|im_start|>';
+      this.endToken = '<|im_end|>';
+      this.gptEncoder = this.constructor.getTokenizer('text-davinci-003', true, {
+        '<|im_start|>': 100264,
+        '<|im_end|>': 100265,
+      });
+    } else {
+      // Previously I was trying to use "<|endoftext|>" but there seems to be some bug with OpenAI's token counting
+      // system that causes only the first "<|endoftext|>" to be counted as 1 token, and the rest are not treated
+      // as a single token. So we're using this instead.
+      this.startToken = '||>';
+      this.endToken = '';
+      try {
+        this.gptEncoder = this.constructor.getTokenizer(this.modelOptions.model, true);
+      } catch {
+        this.gptEncoder = this.constructor.getTokenizer('text-davinci-003', true);
+      }
+    }
+
+    if (!this.modelOptions.stop) {
+      const stopTokens = [this.startToken];
+      if (this.endToken && this.endToken !== this.startToken) {
+        stopTokens.push(this.endToken);
+      }
+      stopTokens.push(`\n${this.userLabel}:`);
+      stopTokens.push('<|diff_marker|>');
+      // I chose not to do one for `chatGptLabel` because I've never seen it happen
+      this.modelOptions.stop = stopTokens;
+    }
+
+    if (this.options.reverseProxyUrl) {
+      this.completionsUrl = this.options.reverseProxyUrl;
+    } else if (isChatGptModel) {
+      this.completionsUrl = 'https://api.openai.com/v1/chat/completions';
+    } else {
+      this.completionsUrl = 'https://api.openai.com/v1/completions';
+    }
+
+    return this;
+  }
+
+  static getTokenizer(encoding, isModelName = false, extendSpecialTokens = {}) {
+    if (tokenizersCache[encoding]) {
+      return tokenizersCache[encoding];
+    }
+    let tokenizer;
+    if (isModelName) {
+      tokenizer = encodingForModel(encoding, extendSpecialTokens);
+    } else {
+      tokenizer = getEncoding(encoding, extendSpecialTokens);
+    }
+    tokenizersCache[encoding] = tokenizer;
+    return tokenizer;
+  }
+
+  /** @type {getCompletion} */
+  async getCompletion(input, onProgress, onTokenProgress, abortController = null) {
+    if (!abortController) {
+      abortController = new AbortController();
+    }
+
+    let modelOptions = { ...this.modelOptions };
+    if (typeof onProgress === 'function') {
+      modelOptions.stream = true;
+    }
+    if (this.isChatGptModel) {
+      modelOptions.messages = input;
+    } else {
+      modelOptions.prompt = input;
+    }
+
+    if (this.useOpenRouter && modelOptions.prompt) {
+      delete modelOptions.stop;
+    }
+
+    const { debug } = this.options;
+    let baseURL = this.completionsUrl;
+    if (debug) {
+      console.debug();
+      console.debug(baseURL);
+      console.debug(modelOptions);
+      console.debug();
+    }
+
+    const opts = {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+      },
+    };
+
+    if (this.isVisionModel) {
+      modelOptions.max_tokens = 4000;
+    }
+
+    /** @type {TAzureConfig | undefined} */
+    const azureConfig = this.options?.req?.app?.locals?.[EModelEndpoint.azureOpenAI];
+
+    const isAzure = this.azure || this.options.azure;
+    if (
+      (isAzure && 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.headers = 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 (serverless === true) {
+        this.options.defaultQuery = azureOptions.azureOpenAIApiVersion
+          ? { 'api-version': azureOptions.azureOpenAIApiVersion }
+          : undefined;
+        this.options.headers['api-key'] = this.apiKey;
+      }
+    }
+
+    if (this.options.defaultQuery) {
+      opts.defaultQuery = this.options.defaultQuery;
+    }
+
+    if (this.options.headers) {
+      opts.headers = { ...opts.headers, ...this.options.headers };
+    }
+
+    if (isAzure) {
+      // Azure does not accept `model` in the body, so we need to remove it.
+      delete modelOptions.model;
+
+      baseURL = this.langchainProxy
+        ? constructAzureURL({
+            baseURL: this.langchainProxy,
+            azureOptions: this.azure,
+          })
+        : this.azureEndpoint.split(/(?<!\/)\/(chat|completion)\//)[0];
+
+      if (this.options.forcePrompt) {
+        baseURL += '/completions';
+      } else {
+        baseURL += '/chat/completions';
+      }
+
+      opts.defaultQuery = { 'api-version': this.azure.azureOpenAIApiVersion };
+      opts.headers = { ...opts.headers, 'api-key': this.apiKey };
+    } else if (this.apiKey) {
+      opts.headers.Authorization = `Bearer ${this.apiKey}`;
+    }
+
+    if (process.env.OPENAI_ORGANIZATION) {
+      opts.headers['OpenAI-Organization'] = process.env.OPENAI_ORGANIZATION;
+    }
+
+    if (this.useOpenRouter) {
+      opts.headers['HTTP-Referer'] = 'https://librechat.ai';
+      opts.headers['X-Title'] = 'LibreChat';
+    }
+
+    /* hacky fixes for Mistral AI API:
+      - Re-orders system message to the top of the messages payload, as not allowed anywhere else
+      - If there is only one message and it's a system message, change the role to user
+      */
+    if (baseURL.includes('https://api.mistral.ai/v1') && modelOptions.messages) {
+      const { messages } = modelOptions;
+
+      const systemMessageIndex = messages.findIndex((msg) => msg.role === 'system');
+
+      if (systemMessageIndex > 0) {
+        const [systemMessage] = messages.splice(systemMessageIndex, 1);
+        messages.unshift(systemMessage);
+      }
+
+      modelOptions.messages = messages;
+
+      if (messages.length === 1 && messages[0].role === 'system') {
+        modelOptions.messages[0].role = 'user';
+      }
+    }
+
+    if (this.options.addParams && typeof this.options.addParams === 'object') {
+      modelOptions = {
+        ...modelOptions,
+        ...this.options.addParams,
+      };
+      logger.debug('[ChatGPTClient] 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('[ChatGPTClient] chatCompletion: dropped params', {
+        dropParams: this.options.dropParams,
+        modelOptions,
+      });
+    }
+
+    if (baseURL.startsWith(CohereConstants.API_URL)) {
+      const payload = createCoherePayload({ modelOptions });
+      return await this.cohereChatCompletion({ payload, onTokenProgress });
+    }
+
+    if (baseURL.includes('v1') && !baseURL.includes('/completions') && !this.isChatCompletion) {
+      baseURL = baseURL.split('v1')[0] + 'v1/completions';
+    } else if (
+      baseURL.includes('v1') &&
+      !baseURL.includes('/chat/completions') &&
+      this.isChatCompletion
+    ) {
+      baseURL = baseURL.split('v1')[0] + 'v1/chat/completions';
+    }
+
+    const BASE_URL = new URL(baseURL);
+    if (opts.defaultQuery) {
+      Object.entries(opts.defaultQuery).forEach(([key, value]) => {
+        BASE_URL.searchParams.append(key, value);
+      });
+      delete opts.defaultQuery;
+    }
+
+    const completionsURL = BASE_URL.toString();
+    opts.body = JSON.stringify(modelOptions);
+
+    if (modelOptions.stream) {
+      return new Promise(async (resolve, reject) => {
+        try {
+          let done = false;
+          await fetchEventSource(completionsURL, {
+            ...opts,
+            signal: abortController.signal,
+            async onopen(response) {
+              if (response.status === 200) {
+                return;
+              }
+              if (debug) {
+                console.debug(response);
+              }
+              let error;
+              try {
+                const body = await response.text();
+                error = new Error(`Failed to send message. HTTP ${response.status} - ${body}`);
+                error.status = response.status;
+                error.json = JSON.parse(body);
+              } catch {
+                error = error || new Error(`Failed to send message. HTTP ${response.status}`);
+              }
+              throw error;
+            },
+            onclose() {
+              if (debug) {
+                console.debug('Server closed the connection unexpectedly, returning...');
+              }
+              // workaround for private API not sending [DONE] event
+              if (!done) {
+                onProgress('[DONE]');
+                resolve();
+              }
+            },
+            onerror(err) {
+              if (debug) {
+                console.debug(err);
+              }
+              // rethrow to stop the operation
+              throw err;
+            },
+            onmessage(message) {
+              if (debug) {
+                console.debug(message);
+              }
+              if (!message.data || message.event === 'ping') {
+                return;
+              }
+              if (message.data === '[DONE]') {
+                onProgress('[DONE]');
+                resolve();
+                done = true;
+                return;
+              }
+              onProgress(JSON.parse(message.data));
+            },
+          });
+        } catch (err) {
+          reject(err);
+        }
+      });
+    }
+    const response = await fetch(completionsURL, {
+      ...opts,
+      signal: abortController.signal,
+    });
+    if (response.status !== 200) {
+      const body = await response.text();
+      const error = new Error(`Failed to send message. HTTP ${response.status} - ${body}`);
+      error.status = response.status;
+      try {
+        error.json = JSON.parse(body);
+      } catch {
+        error.body = body;
+      }
+      throw error;
+    }
+    return response.json();
+  }
+
+  /** @type {cohereChatCompletion} */
+  async cohereChatCompletion({ payload, onTokenProgress }) {
+    const cohere = new CohereClient({
+      token: this.apiKey,
+      environment: this.completionsUrl,
+    });
+
+    if (!payload.stream) {
+      const chatResponse = await cohere.chat(payload);
+      return chatResponse.text;
+    }
+
+    const chatStream = await cohere.chatStream(payload);
+    let reply = '';
+    for await (const message of chatStream) {
+      if (!message) {
+        continue;
+      }
+
+      if (message.eventType === 'text-generation' && message.text) {
+        onTokenProgress(message.text);
+        reply += message.text;
+      }
+      /*
+      Cohere API Chinese Unicode character replacement hotfix.
+      Should be un-commented when the following issue is resolved:
+      https://github.com/cohere-ai/cohere-typescript/issues/151
+
+      else if (message.eventType === 'stream-end' && message.response) {
+        reply = message.response.text;
+      }
+      */
+    }
+
+    return reply;
+  }
+
+  async generateTitle(userMessage, botMessage) {
+    const instructionsPayload = {
+      role: 'system',
+      content: `Write an extremely concise subtitle for this conversation with no more than a few words. All words should be capitalized. Exclude punctuation.
+
+||>Message:
+${userMessage.message}
+||>Response:
+${botMessage.message}
+
+||>Title:`,
+    };
+
+    const titleGenClientOptions = JSON.parse(JSON.stringify(this.options));
+    titleGenClientOptions.modelOptions = {
+      model: 'gpt-3.5-turbo',
+      temperature: 0,
+      presence_penalty: 0,
+      frequency_penalty: 0,
+    };
+    const titleGenClient = new ChatGPTClient(this.apiKey, titleGenClientOptions);
+    const result = await titleGenClient.getCompletion([instructionsPayload], null);
+    // remove any non-alphanumeric characters, replace multiple spaces with 1, and then trim
+    return result.choices[0].message.content
+      .replace(/[^a-zA-Z0-9' ]/g, '')
+      .replace(/\s+/g, ' ')
+      .trim();
+  }
+
+  async sendMessage(message, opts = {}) {
+    if (opts.clientOptions && typeof opts.clientOptions === 'object') {
+      this.setOptions(opts.clientOptions);
+    }
+
+    const conversationId = opts.conversationId || crypto.randomUUID();
+    const parentMessageId = opts.parentMessageId || crypto.randomUUID();
+
+    let conversation =
+      typeof opts.conversation === 'object'
+        ? opts.conversation
+        : await this.conversationsCache.get(conversationId);
+
+    let isNewConversation = false;
+    if (!conversation) {
+      conversation = {
+        messages: [],
+        createdAt: Date.now(),
+      };
+      isNewConversation = true;
+    }
+
+    const shouldGenerateTitle = opts.shouldGenerateTitle && isNewConversation;
+
+    const userMessage = {
+      id: crypto.randomUUID(),
+      parentMessageId,
+      role: 'User',
+      message,
+    };
+    conversation.messages.push(userMessage);
+
+    // Doing it this way instead of having each message be a separate element in the array seems to be more reliable,
+    // especially when it comes to keeping the AI in character. It also seems to improve coherency and context retention.
+    const { prompt: payload, context } = await this.buildPrompt(
+      conversation.messages,
+      userMessage.id,
+      {
+        isChatGptModel: this.isChatGptModel,
+        promptPrefix: opts.promptPrefix,
+      },
+    );
+
+    if (this.options.keepNecessaryMessagesOnly) {
+      conversation.messages = context;
+    }
+
+    let reply = '';
+    let result = null;
+    if (typeof opts.onProgress === 'function') {
+      await this.getCompletion(
+        payload,
+        (progressMessage) => {
+          if (progressMessage === '[DONE]') {
+            return;
+          }
+          const token = this.isChatGptModel
+            ? progressMessage.choices[0].delta.content
+            : progressMessage.choices[0].text;
+          // first event's delta content is always undefined
+          if (!token) {
+            return;
+          }
+          if (this.options.debug) {
+            console.debug(token);
+          }
+          if (token === this.endToken) {
+            return;
+          }
+          opts.onProgress(token);
+          reply += token;
+        },
+        opts.abortController || new AbortController(),
+      );
+    } else {
+      result = await this.getCompletion(
+        payload,
+        null,
+        opts.abortController || new AbortController(),
+      );
+      if (this.options.debug) {
+        console.debug(JSON.stringify(result));
+      }
+      if (this.isChatGptModel) {
+        reply = result.choices[0].message.content;
+      } else {
+        reply = result.choices[0].text.replace(this.endToken, '');
+      }
+    }
+
+    // avoids some rendering issues when using the CLI app
+    if (this.options.debug) {
+      console.debug();
+    }
+
+    reply = reply.trim();
+
+    const replyMessage = {
+      id: crypto.randomUUID(),
+      parentMessageId: userMessage.id,
+      role: 'ChatGPT',
+      message: reply,
+    };
+    conversation.messages.push(replyMessage);
+
+    const returnData = {
+      response: replyMessage.message,
+      conversationId,
+      parentMessageId: replyMessage.parentMessageId,
+      messageId: replyMessage.id,
+      details: result || {},
+    };
+
+    if (shouldGenerateTitle) {
+      conversation.title = await this.generateTitle(userMessage, replyMessage);
+      returnData.title = conversation.title;
+    }
+
+    await this.conversationsCache.set(conversationId, conversation);
+
+    if (this.options.returnConversation) {
+      returnData.conversation = conversation;
+    }
+
+    return returnData;
+  }
+
+  async buildPrompt(messages, { isChatGptModel = false, promptPrefix = null }) {
+    promptPrefix = (promptPrefix || this.options.promptPrefix || '').trim();
+
+    // Handle attachments and create augmentedPrompt
+    if (this.options.attachments) {
+      const attachments = await this.options.attachments;
+      const lastMessage = messages[messages.length - 1];
+
+      if (this.message_file_map) {
+        this.message_file_map[lastMessage.messageId] = attachments;
+      } else {
+        this.message_file_map = {
+          [lastMessage.messageId]: attachments,
+        };
+      }
+
+      const files = await this.addImageURLs(lastMessage, attachments);
+      this.options.attachments = files;
+
+      this.contextHandlers = createContextHandlers(this.options.req, lastMessage.text);
+    }
+
+    if (this.message_file_map) {
+      this.contextHandlers = createContextHandlers(
+        this.options.req,
+        messages[messages.length - 1].text,
+      );
+    }
+
+    // Calculate image token cost and process embedded files
+    messages.forEach((message, i) => {
+      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;
+          }
+
+          messages[i].tokenCount =
+            (messages[i].tokenCount || 0) +
+            this.calculateImageTokenCost({
+              width: file.width,
+              height: file.height,
+              detail: this.options.imageDetail ?? ImageDetail.auto,
+            });
+        }
+      }
+    });
+
+    if (this.contextHandlers) {
+      this.augmentedPrompt = await this.contextHandlers.createContext();
+      promptPrefix = this.augmentedPrompt + promptPrefix;
+    }
+
+    if (promptPrefix) {
+      // If the prompt prefix doesn't end with the end token, add it.
+      if (!promptPrefix.endsWith(`${this.endToken}`)) {
+        promptPrefix = `${promptPrefix.trim()}${this.endToken}\n\n`;
+      }
+      promptPrefix = `${this.startToken}Instructions:\n${promptPrefix}`;
+    }
+    const promptSuffix = `${this.startToken}${this.chatGptLabel}:\n`; // Prompt ChatGPT to respond.
+
+    const instructionsPayload = {
+      role: 'system',
+      content: promptPrefix,
+    };
+
+    const messagePayload = {
+      role: 'system',
+      content: promptSuffix,
+    };
+
+    let currentTokenCount;
+    if (isChatGptModel) {
+      currentTokenCount =
+        this.getTokenCountForMessage(instructionsPayload) +
+        this.getTokenCountForMessage(messagePayload);
+    } else {
+      currentTokenCount = this.getTokenCount(`${promptPrefix}${promptSuffix}`);
+    }
+    let promptBody = '';
+    const maxTokenCount = this.maxPromptTokens;
+
+    const context = [];
+
+    // Iterate backwards through the messages, adding them to the prompt until we reach the max token count.
+    // Do this within a recursive async function so that it doesn't block the event loop for too long.
+    const buildPromptBody = async () => {
+      if (currentTokenCount < maxTokenCount && messages.length > 0) {
+        const message = messages.pop();
+        const roleLabel =
+          message?.isCreatedByUser || message?.role?.toLowerCase() === 'user'
+            ? this.userLabel
+            : this.chatGptLabel;
+        const messageString = `${this.startToken}${roleLabel}:\n${
+          message?.text ?? message?.message
+        }${this.endToken}\n`;
+        let newPromptBody;
+        if (promptBody || isChatGptModel) {
+          newPromptBody = `${messageString}${promptBody}`;
+        } else {
+          // Always insert prompt prefix before the last user message, if not gpt-3.5-turbo.
+          // This makes the AI obey the prompt instructions better, which is important for custom instructions.
+          // After a bunch of testing, it doesn't seem to cause the AI any confusion, even if you ask it things
+          // like "what's the last thing I wrote?".
+          newPromptBody = `${promptPrefix}${messageString}${promptBody}`;
+        }
+
+        context.unshift(message);
+
+        const tokenCountForMessage = this.getTokenCount(messageString);
+        const newTokenCount = currentTokenCount + tokenCountForMessage;
+        if (newTokenCount > maxTokenCount) {
+          if (promptBody) {
+            // This message would put us over the token limit, so don't add it.
+            return false;
+          }
+          // This is the first message, so we can't add it. Just throw an error.
+          throw new Error(
+            `Prompt is too long. Max token count is ${maxTokenCount}, but prompt is ${newTokenCount} tokens long.`,
+          );
+        }
+        promptBody = newPromptBody;
+        currentTokenCount = newTokenCount;
+        // wait for next tick to avoid blocking the event loop
+        await new Promise((resolve) => setImmediate(resolve));
+        return buildPromptBody();
+      }
+      return true;
+    };
+
+    await buildPromptBody();
+
+    const prompt = `${promptBody}${promptSuffix}`;
+    if (isChatGptModel) {
+      messagePayload.content = prompt;
+      // Add 3 tokens for Assistant Label priming after all messages have been counted.
+      currentTokenCount += 3;
+    }
+
+    // Use up to `this.maxContextTokens` tokens (prompt + response), but try to leave `this.maxTokens` tokens for the response.
+    this.modelOptions.max_tokens = Math.min(
+      this.maxContextTokens - currentTokenCount,
+      this.maxResponseTokens,
+    );
+
+    if (isChatGptModel) {
+      return { prompt: [instructionsPayload, messagePayload], context };
+    }
+    return { prompt, context, promptTokens: currentTokenCount };
+  }
+
+  getTokenCount(text) {
+    return this.gptEncoder.encode(text, 'all').length;
+  }
+
+  /**
+   * Algorithm adapted from "6. Counting tokens for chat API calls" of
+   * https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
+   *
+   * An additional 3 tokens need to be added for assistant label priming after all messages have been counted.
+   *
+   * @param {Object} message
+   */
+  getTokenCountForMessage(message) {
+    // Note: gpt-3.5-turbo and gpt-4 may update over time. Use default for these as well as for unknown models
+    let tokensPerMessage = 3;
+    let tokensPerName = 1;
+
+    if (this.modelOptions.model === 'gpt-3.5-turbo-0301') {
+      tokensPerMessage = 4;
+      tokensPerName = -1;
+    }
+
+    let numTokens = tokensPerMessage;
+    for (let [key, value] of Object.entries(message)) {
+      numTokens += this.getTokenCount(value);
+      if (key === 'name') {
+        numTokens += tokensPerName;
+      }
+    }
+
+    return numTokens;
+  }
+}
+
+module.exports = ChatGPTClient;

api/app/clients/GoogleClient.js

@@ -1,7 +1,7 @@
 const { google } = require('googleapis');
+const { Tokenizer } = require('@librechat/api');
 const { concat } = require('@langchain/core/utils/stream');
 const { ChatVertexAI } = require('@langchain/google-vertexai');
-const { Tokenizer, getSafetySettings } = require('@librechat/api');
 const { ChatGoogleGenerativeAI } = require('@langchain/google-genai');
 const { GoogleGenerativeAI: GenAI } = require('@google/generative-ai');
 const { HumanMessage, SystemMessage } = require('@langchain/core/messages');
@@ -12,13 +12,13 @@ const {
   endpointSettings,
   parseTextParts,
   EModelEndpoint,
-  googleSettings,
   ContentTypes,
   VisionModes,
   ErrorTypes,
   Constants,
   AuthKeys,
 } = require('librechat-data-provider');
+const { getSafetySettings } = require('~/server/services/Endpoints/google/llm');
 const { encodeAndFormat } = require('~/server/services/Files/images');
 const { spendTokens } = require('~/models/spendTokens');
 const { getModelMaxTokens } = require('~/utils');
@@ -166,16 +166,6 @@ class GoogleClient extends BaseClient {
       );
     }
 
-    // Add thinking configuration
-    this.modelOptions.thinkingConfig = {
-      thinkingBudget:
-        (this.modelOptions.thinking ?? googleSettings.thinking.default)
-          ? this.modelOptions.thinkingBudget
-          : 0,
-    };
-    delete this.modelOptions.thinking;
-    delete this.modelOptions.thinkingBudget;
-
     this.sender =
       this.options.sender ??
       getResponseSender({

api/app/clients/OpenAIClient.js

@@ -5,7 +5,6 @@ const {
   isEnabled,
   Tokenizer,
   createFetch,
-  resolveHeaders,
   constructAzureURL,
   genAzureChatCompletion,
   createStreamEventHandlers,
@@ -16,6 +15,7 @@ const {
   ContentTypes,
   parseTextParts,
   EModelEndpoint,
+  resolveHeaders,
   KnownEndpoints,
   openAISettings,
   ImageDetailCost,
@@ -37,6 +37,7 @@ const { addSpaceIfNeeded, sleep } = require('~/server/utils');
 const { spendTokens } = require('~/models/spendTokens');
 const { handleOpenAIErrors } = require('./tools/util');
 const { createLLM, RunManager } = require('./llm');
+const ChatGPTClient = require('./ChatGPTClient');
 const { summaryBuffer } = require('./memory');
 const { runTitleChain } = require('./chains');
 const { tokenSplit } = require('./document');
@@ -46,6 +47,12 @@ const { logger } = require('~/config');
 class OpenAIClient extends BaseClient {
   constructor(apiKey, options = {}) {
     super(apiKey, options);
+    this.ChatGPTClient = new ChatGPTClient();
+    this.buildPrompt = this.ChatGPTClient.buildPrompt.bind(this);
+    /** @type {getCompletion} */
+    this.getCompletion = this.ChatGPTClient.getCompletion.bind(this);
+    /** @type {cohereChatCompletion} */
+    this.cohereChatCompletion = this.ChatGPTClient.cohereChatCompletion.bind(this);
     this.contextStrategy = options.contextStrategy
       ? options.contextStrategy.toLowerCase()
       : 'discard';
@@ -372,12 +379,23 @@ class OpenAIClient extends BaseClient {
     return files;
   }
 
-  async buildMessages(messages, parentMessageId, { promptPrefix = null }, opts) {
+  async buildMessages(
+    messages,
+    parentMessageId,
+    { isChatCompletion = false, promptPrefix = null },
+    opts,
+  ) {
     let orderedMessages = this.constructor.getMessagesForConversation({
       messages,
       parentMessageId,
       summary: this.shouldSummarize,
     });
+    if (!isChatCompletion) {
+      return await this.buildPrompt(orderedMessages, {
+        isChatGptModel: isChatCompletion,
+        promptPrefix,
+      });
+    }
 
     let payload;
     let instructions;

api/app/clients/PluginsClient.js

@@ -0,0 +1,542 @@
+const OpenAIClient = require('./OpenAIClient');
+const { CallbackManager } = require('@langchain/core/callbacks/manager');
+const { BufferMemory, ChatMessageHistory } = require('langchain/memory');
+const { addImages, buildErrorInput, buildPromptPrefix } = require('./output_parsers');
+const { initializeCustomAgent, initializeFunctionsAgent } = require('./agents');
+const { processFileURL } = require('~/server/services/Files/process');
+const { EModelEndpoint } = require('librechat-data-provider');
+const { checkBalance } = require('~/models/balanceMethods');
+const { formatLangChainMessages } = require('./prompts');
+const { extractBaseURL } = require('~/utils');
+const { loadTools } = require('./tools/util');
+const { logger } = require('~/config');
+
+class PluginsClient extends OpenAIClient {
+  constructor(apiKey, options = {}) {
+    super(apiKey, options);
+    this.sender = options.sender ?? 'Assistant';
+    this.tools = [];
+    this.actions = [];
+    this.setOptions(options);
+    this.openAIApiKey = this.apiKey;
+    this.executor = null;
+  }
+
+  setOptions(options) {
+    this.agentOptions = { ...options.agentOptions };
+    this.functionsAgent = this.agentOptions?.agent === 'functions';
+    this.agentIsGpt3 = this.agentOptions?.model?.includes('gpt-3');
+
+    super.setOptions(options);
+
+    this.isGpt3 = this.modelOptions?.model?.includes('gpt-3');
+
+    if (this.options.reverseProxyUrl) {
+      this.langchainProxy = extractBaseURL(this.options.reverseProxyUrl);
+    }
+  }
+
+  getSaveOptions() {
+    return {
+      artifacts: this.options.artifacts,
+      chatGptLabel: this.options.chatGptLabel,
+      modelLabel: this.options.modelLabel,
+      promptPrefix: this.options.promptPrefix,
+      tools: this.options.tools,
+      ...this.modelOptions,
+      agentOptions: this.agentOptions,
+      iconURL: this.options.iconURL,
+      greeting: this.options.greeting,
+      spec: this.options.spec,
+    };
+  }
+
+  saveLatestAction(action) {
+    this.actions.push(action);
+  }
+
+  getFunctionModelName(input) {
+    if (/-(?!0314)\d{4}/.test(input)) {
+      return input;
+    } else if (input.includes('gpt-3.5-turbo')) {
+      return 'gpt-3.5-turbo';
+    } else if (input.includes('gpt-4')) {
+      return 'gpt-4';
+    } else {
+      return 'gpt-3.5-turbo';
+    }
+  }
+
+  getBuildMessagesOptions(opts) {
+    return {
+      isChatCompletion: true,
+      promptPrefix: opts.promptPrefix,
+      abortController: opts.abortController,
+    };
+  }
+
+  async initialize({ user, message, onAgentAction, onChainEnd, signal }) {
+    const modelOptions = {
+      modelName: this.agentOptions.model,
+      temperature: this.agentOptions.temperature,
+    };
+
+    const model = this.initializeLLM({
+      ...modelOptions,
+      context: 'plugins',
+      initialMessageCount: this.currentMessages.length + 1,
+    });
+
+    logger.debug(
+      `[PluginsClient] Agent Model: ${model.modelName} | Temp: ${model.temperature} | Functions: ${this.functionsAgent}`,
+    );
+
+    // Map Messages to Langchain format
+    const pastMessages = formatLangChainMessages(this.currentMessages.slice(0, -1), {
+      userName: this.options?.name,
+    });
+    logger.debug('[PluginsClient] pastMessages: ' + pastMessages.length);
+
+    // TODO: use readOnly memory, TokenBufferMemory? (both unavailable in LangChainJS)
+    const memory = new BufferMemory({
+      llm: model,
+      chatHistory: new ChatMessageHistory(pastMessages),
+    });
+
+    const { loadedTools } = await loadTools({
+      user,
+      model,
+      tools: this.options.tools,
+      functions: this.functionsAgent,
+      options: {
+        memory,
+        signal: this.abortController.signal,
+        openAIApiKey: this.openAIApiKey,
+        conversationId: this.conversationId,
+        fileStrategy: this.options.req.app.locals.fileStrategy,
+        processFileURL,
+        message,
+      },
+      useSpecs: true,
+    });
+
+    if (loadedTools.length === 0) {
+      return;
+    }
+
+    this.tools = loadedTools;
+
+    logger.debug('[PluginsClient] Requested Tools', this.options.tools);
+    logger.debug(
+      '[PluginsClient] Loaded Tools',
+      this.tools.map((tool) => tool.name),
+    );
+
+    const handleAction = (action, runId, callback = null) => {
+      this.saveLatestAction(action);
+
+      logger.debug('[PluginsClient] Latest Agent Action ', this.actions[this.actions.length - 1]);
+
+      if (typeof callback === 'function') {
+        callback(action, runId);
+      }
+    };
+
+    // initialize agent
+    const initializer = this.functionsAgent ? initializeFunctionsAgent : initializeCustomAgent;
+
+    let customInstructions = (this.options.promptPrefix ?? '').trim();
+    if (typeof this.options.artifactsPrompt === 'string' && this.options.artifactsPrompt) {
+      customInstructions = `${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,
+      callbackManager: CallbackManager.fromHandlers({
+        async handleAgentAction(action, runId) {
+          handleAction(action, runId, onAgentAction);
+        },
+        async handleChainEnd(action) {
+          if (typeof onChainEnd === 'function') {
+            onChainEnd(action);
+          }
+        },
+      }),
+    });
+
+    logger.debug('[PluginsClient] Loaded agent.');
+  }
+
+  async executorCall(message, { signal, stream, onToolStart, onToolEnd }) {
+    let errorMessage = '';
+    const maxAttempts = 1;
+
+    for (let attempts = 1; attempts <= maxAttempts; attempts++) {
+      const errorInput = buildErrorInput({
+        message,
+        errorMessage,
+        actions: this.actions,
+        functionsAgent: this.functionsAgent,
+      });
+      const input = attempts > 1 ? errorInput : message;
+
+      logger.debug(`[PluginsClient] Attempt ${attempts} of ${maxAttempts}`);
+
+      if (errorMessage.length > 0) {
+        logger.debug('[PluginsClient] Caught error, input: ' + JSON.stringify(input));
+      }
+
+      try {
+        this.result = await this.executor.call({ input, signal }, [
+          {
+            async handleToolStart(...args) {
+              await onToolStart(...args);
+            },
+            async handleToolEnd(...args) {
+              await onToolEnd(...args);
+            },
+            async handleLLMEnd(output) {
+              const { generations } = output;
+              const { text } = generations[0][0];
+              if (text && typeof stream === 'function') {
+                await stream(text);
+              }
+            },
+          },
+        ]);
+        break; // Exit the loop if the function call is successful
+      } catch (err) {
+        logger.error('[PluginsClient] executorCall error:', err);
+        if (attempts === maxAttempts) {
+          const { run } = this.runManager.getRunByConversationId(this.conversationId);
+          const defaultOutput = `Encountered an error while attempting to respond: ${err.message}`;
+          this.result.output = run && run.error ? run.error : defaultOutput;
+          this.result.errorMessage = run && run.error ? run.error : err.message;
+          this.result.intermediateSteps = this.actions;
+          break;
+        }
+      }
+    }
+  }
+
+  /**
+   *
+   * @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:', {
+      output,
+      errorMessage,
+      ...result,
+    });
+    const { error } = responseMessage;
+    if (!error) {
+      responseMessage.tokenCount = this.getTokenCountForResponse(responseMessage);
+      responseMessage.completionTokens = this.getTokenCount(responseMessage.text);
+    }
+
+    // Record usage only when completion is skipped as it is already recorded in the agent phase.
+    if (!this.agentOptions.skipCompletion && !error) {
+      await this.recordTokenUsage(responseMessage);
+    }
+
+    const databasePromise = this.saveMessageToDatabase(responseMessage, saveOptions, user);
+    delete responseMessage.tokenCount;
+    return { ...responseMessage, ...result, databasePromise };
+  }
+
+  async sendMessage(message, opts = {}) {
+    /** @type {Promise<TMessage>} */
+    let userMessagePromise;
+    /** @type {{ filteredTools: string[], includedTools: string[] }} */
+    const { filteredTools = [], includedTools = [] } = this.options.req.app.locals;
+
+    if (includedTools.length > 0) {
+      const tools = this.options.tools.filter((plugin) => includedTools.includes(plugin));
+      this.options.tools = tools;
+    } else {
+      const tools = this.options.tools.filter((plugin) => !filteredTools.includes(plugin));
+      this.options.tools = tools;
+    }
+
+    // If a message is edited, no tools can be used.
+    const completionMode = this.options.tools.length === 0 || opts.isEdited;
+    if (completionMode) {
+      this.setOptions(opts);
+      return super.sendMessage(message, opts);
+    }
+
+    logger.debug('[PluginsClient] sendMessage', { userMessageText: message, opts });
+    const {
+      user,
+      conversationId,
+      responseMessageId,
+      saveOptions,
+      userMessage,
+      onAgentAction,
+      onChainEnd,
+      onToolStart,
+      onToolEnd,
+    } = await this.handleStartMethods(message, opts);
+
+    if (opts.progressCallback) {
+      opts.onProgress = opts.progressCallback.call(null, {
+        ...(opts.progressOptions ?? {}),
+        parentMessageId: userMessage.messageId,
+        messageId: responseMessageId,
+      });
+    }
+
+    this.currentMessages.push(userMessage);
+
+    let {
+      prompt: payload,
+      tokenCountMap,
+      promptTokens,
+    } = await this.buildMessages(
+      this.currentMessages,
+      userMessage.messageId,
+      this.getBuildMessagesOptions({
+        promptPrefix: null,
+        abortController: this.abortController,
+      }),
+    );
+
+    if (tokenCountMap) {
+      logger.debug('[PluginsClient] tokenCountMap', { tokenCountMap });
+      if (tokenCountMap[userMessage.messageId]) {
+        userMessage.tokenCount = tokenCountMap[userMessage.messageId];
+        logger.debug('[PluginsClient] userMessage.tokenCount', userMessage.tokenCount);
+      }
+      this.handleTokenCountMap(tokenCountMap);
+    }
+
+    this.result = {};
+    if (payload) {
+      this.currentMessages = payload;
+    }
+
+    if (!this.skipSaveUserMessage) {
+      userMessagePromise = this.saveMessageToDatabase(userMessage, saveOptions, user);
+      if (typeof opts?.getReqData === 'function') {
+        opts.getReqData({
+          userMessagePromise,
+        });
+      }
+    }
+
+    const balance = this.options.req?.app?.locals?.balance;
+    if (balance?.enabled) {
+      await checkBalance({
+        req: this.options.req,
+        res: this.options.res,
+        txData: {
+          user: this.user,
+          tokenType: 'prompt',
+          amount: promptTokens,
+          debug: this.options.debug,
+          model: this.modelOptions.model,
+          endpoint: EModelEndpoint.openAI,
+        },
+      });
+    }
+
+    const responseMessage = {
+      endpoint: EModelEndpoint.gptPlugins,
+      iconURL: this.options.iconURL,
+      messageId: responseMessageId,
+      conversationId,
+      parentMessageId: userMessage.messageId,
+      isCreatedByUser: false,
+      model: this.modelOptions.model,
+      sender: this.sender,
+      promptTokens,
+    };
+
+    await this.initialize({
+      user,
+      message,
+      onAgentAction,
+      onChainEnd,
+      signal: this.abortController.signal,
+      onProgress: opts.onProgress,
+    });
+
+    // const stream = async (text) => {
+    //   await this.generateTextStream.call(this, text, opts.onProgress, { delay: 1 });
+    // };
+    await this.executorCall(message, {
+      signal: this.abortController.signal,
+      // stream,
+      onToolStart,
+      onToolEnd,
+    });
+
+    // If message was aborted mid-generation
+    if (this.result?.errorMessage?.length > 0 && this.result?.errorMessage?.includes('cancel')) {
+      responseMessage.text = 'Cancelled.';
+      return await this.handleResponseMessage(responseMessage, saveOptions, user);
+    }
+
+    // If error occurred during generation (likely token_balance)
+    if (this.result?.errorMessage?.length > 0) {
+      responseMessage.error = true;
+      responseMessage.text = this.result.output;
+      return await this.handleResponseMessage(responseMessage, saveOptions, user);
+    }
+
+    if (this.agentOptions.skipCompletion && this.result.output && this.functionsAgent) {
+      const partialText = opts.getPartialText();
+      const trimmedPartial = opts.getPartialText().replaceAll(':::plugin:::\n', '');
+      responseMessage.text =
+        trimmedPartial.length === 0 ? `${partialText}${this.result.output}` : partialText;
+      addImages(this.result.intermediateSteps, responseMessage);
+      await this.generateTextStream(this.result.output, opts.onProgress, { delay: 5 });
+      return await this.handleResponseMessage(responseMessage, saveOptions, user);
+    }
+
+    if (this.agentOptions.skipCompletion && this.result.output) {
+      responseMessage.text = this.result.output;
+      addImages(this.result.intermediateSteps, responseMessage);
+      await this.generateTextStream(this.result.output, opts.onProgress, { delay: 5 });
+      return await this.handleResponseMessage(responseMessage, saveOptions, user);
+    }
+
+    logger.debug('[PluginsClient] Completion phase: this.result', this.result);
+
+    const promptPrefix = buildPromptPrefix({
+      result: this.result,
+      message,
+      functionsAgent: this.functionsAgent,
+    });
+
+    logger.debug('[PluginsClient]', { promptPrefix });
+
+    payload = await this.buildCompletionPrompt({
+      messages: this.currentMessages,
+      promptPrefix,
+    });
+
+    logger.debug('[PluginsClient] buildCompletionPrompt Payload', payload);
+    responseMessage.text = await this.sendCompletion(payload, opts);
+    return await this.handleResponseMessage(responseMessage, saveOptions, user);
+  }
+
+  async buildCompletionPrompt({ messages, promptPrefix: _promptPrefix }) {
+    logger.debug('[PluginsClient] buildCompletionPrompt messages', messages);
+
+    const orderedMessages = messages;
+    let promptPrefix = _promptPrefix.trim();
+    // If the prompt prefix doesn't end with the end token, add it.
+    if (!promptPrefix.endsWith(`${this.endToken}`)) {
+      promptPrefix = `${promptPrefix.trim()}${this.endToken}\n\n`;
+    }
+    promptPrefix = `${this.startToken}Instructions:\n${promptPrefix}`;
+    const promptSuffix = `${this.startToken}${this.chatGptLabel ?? 'Assistant'}:\n`;
+
+    const instructionsPayload = {
+      role: 'system',
+      content: promptPrefix,
+    };
+
+    const messagePayload = {
+      role: 'system',
+      content: promptSuffix,
+    };
+
+    if (this.isGpt3) {
+      instructionsPayload.role = 'user';
+      messagePayload.role = 'user';
+      instructionsPayload.content += `\n${promptSuffix}`;
+    }
+
+    // testing if this works with browser endpoint
+    if (!this.isGpt3 && this.options.reverseProxyUrl) {
+      instructionsPayload.role = 'user';
+    }
+
+    let currentTokenCount =
+      this.getTokenCountForMessage(instructionsPayload) +
+      this.getTokenCountForMessage(messagePayload);
+
+    let promptBody = '';
+    const maxTokenCount = this.maxPromptTokens;
+    // Iterate backwards through the messages, adding them to the prompt until we reach the max token count.
+    // Do this within a recursive async function so that it doesn't block the event loop for too long.
+    const buildPromptBody = async () => {
+      if (currentTokenCount < maxTokenCount && orderedMessages.length > 0) {
+        const message = orderedMessages.pop();
+        const isCreatedByUser = message.isCreatedByUser || message.role?.toLowerCase() === 'user';
+        const roleLabel = isCreatedByUser ? this.userLabel : this.chatGptLabel;
+        let messageString = `${this.startToken}${roleLabel}:\n${
+          message.text ?? message.content ?? ''
+        }${this.endToken}\n`;
+        let newPromptBody = `${messageString}${promptBody}`;
+
+        const tokenCountForMessage = this.getTokenCount(messageString);
+        const newTokenCount = currentTokenCount + tokenCountForMessage;
+        if (newTokenCount > maxTokenCount) {
+          if (promptBody) {
+            // This message would put us over the token limit, so don't add it.
+            return false;
+          }
+          // This is the first message, so we can't add it. Just throw an error.
+          throw new Error(
+            `Prompt is too long. Max token count is ${maxTokenCount}, but prompt is ${newTokenCount} tokens long.`,
+          );
+        }
+        promptBody = newPromptBody;
+        currentTokenCount = newTokenCount;
+        // wait for next tick to avoid blocking the event loop
+        await new Promise((resolve) => setTimeout(resolve, 0));
+        return buildPromptBody();
+      }
+      return true;
+    };
+
+    await buildPromptBody();
+    const prompt = promptBody;
+    messagePayload.content = prompt;
+    // Add 2 tokens for metadata after all messages have been counted.
+    currentTokenCount += 2;
+
+    if (this.isGpt3 && messagePayload.content.length > 0) {
+      const context = 'Chat History:\n';
+      messagePayload.content = `${context}${prompt}`;
+      currentTokenCount += this.getTokenCount(context);
+    }
+
+    // Use up to `this.maxContextTokens` tokens (prompt + response), but try to leave `this.maxTokens` tokens for the response.
+    this.modelOptions.max_tokens = Math.min(
+      this.maxContextTokens - currentTokenCount,
+      this.maxResponseTokens,
+    );
+
+    if (this.isGpt3) {
+      messagePayload.content += promptSuffix;
+      return [instructionsPayload, messagePayload];
+    }
+
+    const result = [messagePayload, instructionsPayload];
+
+    if (this.functionsAgent && !this.isGpt3) {
+      result[1].content = `${result[1].content}\n${this.startToken}${this.chatGptLabel}:\nSure thing! Here is the output you requested:\n`;
+    }
+
+    return result.filter((message) => message.content.length > 0);
+  }
+}
+
+module.exports = PluginsClient;

api/app/clients/index.js

@@ -1,11 +1,15 @@
+const ChatGPTClient = require('./ChatGPTClient');
 const OpenAIClient = require('./OpenAIClient');
+const PluginsClient = require('./PluginsClient');
 const GoogleClient = require('./GoogleClient');
 const TextStream = require('./TextStream');
 const AnthropicClient = require('./AnthropicClient');
 const toolUtils = require('./tools/util');
 
 module.exports = {
+  ChatGPTClient,
   OpenAIClient,
+  PluginsClient,
   GoogleClient,
   TextStream,
   AnthropicClient,

api/app/clients/prompts/createContextHandlers.js

@@ -1,7 +1,6 @@
 const axios = require('axios');
-const { isEnabled } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
-const { generateShortLivedToken } = require('~/server/services/AuthService');
+const { isEnabled } = require('~/server/utils');
+const { logger } = require('~/config');
 
 const footer = `Use the context as your learned knowledge to better answer the user.
 
@@ -19,7 +18,7 @@ function createContextHandlers(req, userMessageContent) {
   const queryPromises = [];
   const processedFiles = [];
   const processedIds = new Set();
-  const jwtToken = generateShortLivedToken(req.user.id);
+  const jwtToken = req.headers.authorization.split(' ')[1];
   const useFullContext = isEnabled(process.env.RAG_USE_FULL_CONTEXT);
 
   const query = async (file) => {
@@ -97,35 +96,35 @@ function createContextHandlers(req, userMessageContent) {
         resolvedQueries.length === 0
           ? '\n\tThe semantic search did not return any results.'
           : resolvedQueries
-              .map((queryResult, index) => {
-                const file = processedFiles[index];
-                let contextItems = queryResult.data;
+            .map((queryResult, index) => {
+              const file = processedFiles[index];
+              let contextItems = queryResult.data;
 
-                const generateContext = (currentContext) =>
-                  `
+              const generateContext = (currentContext) =>
+                `
           <file>
             <filename>${file.filename}</filename>
             <context>${currentContext}
             </context>
           </file>`;
 
-                if (useFullContext) {
-                  return generateContext(`\n${contextItems}`);
-                }
+              if (useFullContext) {
+                return generateContext(`\n${contextItems}`);
+              }
 
-                contextItems = queryResult.data
-                  .map((item) => {
-                    const pageContent = item[0].page_content;
-                    return `
+              contextItems = queryResult.data
+                .map((item) => {
+                  const pageContent = item[0].page_content;
+                  return `
             <contextItem>
               <![CDATA[${pageContent?.trim()}]]>
             </contextItem>`;
-                  })
-                  .join('');
+                })
+                .join('');
 
-                return generateContext(contextItems);
-              })
-              .join('');
+              return generateContext(contextItems);
+            })
+            .join('');
 
       if (useFullContext) {
         const prompt = `${header}

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

@@ -531,6 +531,44 @@ describe('OpenAIClient', () => {
     });
   });
 
+  describe('sendMessage/getCompletion/chatCompletion', () => {
+    afterEach(() => {
+      delete process.env.AZURE_OPENAI_DEFAULT_MODEL;
+      delete process.env.AZURE_USE_MODEL_AS_DEPLOYMENT_NAME;
+    });
+
+    it('should call getCompletion and fetchEventSource when using a text/instruct model', async () => {
+      const model = 'text-davinci-003';
+      const onProgress = jest.fn().mockImplementation(() => ({}));
+
+      const testClient = new OpenAIClient('test-api-key', {
+        ...defaultOptions,
+        modelOptions: { model },
+      });
+
+      const getCompletion = jest.spyOn(testClient, 'getCompletion');
+      await testClient.sendMessage('Hi mom!', { onProgress });
+
+      expect(getCompletion).toHaveBeenCalled();
+      expect(getCompletion.mock.calls.length).toBe(1);
+
+      expect(getCompletion.mock.calls[0][0]).toBe('||>User:\nHi mom!\n||>Assistant:\n');
+
+      expect(fetchEventSource).toHaveBeenCalled();
+      expect(fetchEventSource.mock.calls.length).toBe(1);
+
+      // Check if the first argument (url) is correct
+      const firstCallArgs = fetchEventSource.mock.calls[0];
+
+      const expectedURL = 'https://api.openai.com/v1/completions';
+      expect(firstCallArgs[0]).toBe(expectedURL);
+
+      const requestBody = JSON.parse(firstCallArgs[1].body);
+      expect(requestBody).toHaveProperty('model');
+      expect(requestBody.model).toBe(model);
+    });
+  });
+
   describe('checkVisionRequest functionality', () => {
     let client;
     const attachments = [{ type: 'image/png' }];

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

@@ -0,0 +1,314 @@
+const crypto = require('crypto');
+const { Constants } = require('librechat-data-provider');
+const { HumanMessage, AIMessage } = require('@langchain/core/messages');
+const PluginsClient = require('../PluginsClient');
+
+jest.mock('~/db/connect');
+jest.mock('~/models/Conversation', () => {
+  return function () {
+    return {
+      save: jest.fn(),
+      deleteConvos: jest.fn(),
+    };
+  };
+});
+
+const defaultAzureOptions = {
+  azureOpenAIApiInstanceName: 'your-instance-name',
+  azureOpenAIApiDeploymentName: 'your-deployment-name',
+  azureOpenAIApiVersion: '2020-07-01-preview',
+};
+
+describe('PluginsClient', () => {
+  let TestAgent;
+  let options = {
+    tools: [],
+    modelOptions: {
+      model: 'gpt-3.5-turbo',
+      temperature: 0,
+      max_tokens: 2,
+    },
+    agentOptions: {
+      model: 'gpt-3.5-turbo',
+    },
+  };
+  let parentMessageId;
+  let conversationId;
+  const fakeMessages = [];
+  const userMessage = 'Hello, ChatGPT!';
+  const apiKey = 'fake-api-key';
+
+  beforeEach(() => {
+    TestAgent = new PluginsClient(apiKey, options);
+    TestAgent.loadHistory = jest
+      .fn()
+      .mockImplementation((conversationId, parentMessageId = null) => {
+        if (!conversationId) {
+          TestAgent.currentMessages = [];
+          return Promise.resolve([]);
+        }
+
+        const orderedMessages = TestAgent.constructor.getMessagesForConversation({
+          messages: fakeMessages,
+          parentMessageId,
+        });
+
+        const chatMessages = orderedMessages.map((msg) =>
+          msg?.isCreatedByUser || msg?.role?.toLowerCase() === 'user'
+            ? new HumanMessage(msg.text)
+            : new AIMessage(msg.text),
+        );
+
+        TestAgent.currentMessages = orderedMessages;
+        return Promise.resolve(chatMessages);
+      });
+    TestAgent.sendMessage = jest.fn().mockImplementation(async (message, opts = {}) => {
+      if (opts && typeof opts === 'object') {
+        TestAgent.setOptions(opts);
+      }
+      const conversationId = opts.conversationId || crypto.randomUUID();
+      const parentMessageId = opts.parentMessageId || Constants.NO_PARENT;
+      const userMessageId = opts.overrideParentMessageId || crypto.randomUUID();
+      this.pastMessages = await TestAgent.loadHistory(
+        conversationId,
+        TestAgent.options?.parentMessageId,
+      );
+
+      const userMessage = {
+        text: message,
+        sender: 'ChatGPT',
+        isCreatedByUser: true,
+        messageId: userMessageId,
+        parentMessageId,
+        conversationId,
+      };
+
+      const response = {
+        sender: 'ChatGPT',
+        text: 'Hello, User!',
+        isCreatedByUser: false,
+        messageId: crypto.randomUUID(),
+        parentMessageId: userMessage.messageId,
+        conversationId,
+      };
+
+      fakeMessages.push(userMessage);
+      fakeMessages.push(response);
+      return response;
+    });
+  });
+
+  test('initializes PluginsClient without crashing', () => {
+    expect(TestAgent).toBeInstanceOf(PluginsClient);
+  });
+
+  test('check setOptions function', () => {
+    expect(TestAgent.agentIsGpt3).toBe(true);
+  });
+
+  describe('sendMessage', () => {
+    test('sendMessage should return a response message', async () => {
+      const expectedResult = expect.objectContaining({
+        sender: 'ChatGPT',
+        text: expect.any(String),
+        isCreatedByUser: false,
+        messageId: expect.any(String),
+        parentMessageId: expect.any(String),
+        conversationId: expect.any(String),
+      });
+
+      const response = await TestAgent.sendMessage(userMessage);
+      parentMessageId = response.messageId;
+      conversationId = response.conversationId;
+      expect(response).toEqual(expectedResult);
+    });
+
+    test('sendMessage should work with provided conversationId and parentMessageId', async () => {
+      const userMessage = 'Second message in the conversation';
+      const opts = {
+        conversationId,
+        parentMessageId,
+      };
+
+      const expectedResult = expect.objectContaining({
+        sender: 'ChatGPT',
+        text: expect.any(String),
+        isCreatedByUser: false,
+        messageId: expect.any(String),
+        parentMessageId: expect.any(String),
+        conversationId: opts.conversationId,
+      });
+
+      const response = await TestAgent.sendMessage(userMessage, opts);
+      parentMessageId = response.messageId;
+      expect(response.conversationId).toEqual(conversationId);
+      expect(response).toEqual(expectedResult);
+    });
+
+    test('should return chat history', async () => {
+      const chatMessages = await TestAgent.loadHistory(conversationId, parentMessageId);
+      expect(TestAgent.currentMessages).toHaveLength(4);
+      expect(chatMessages[0].text).toEqual(userMessage);
+    });
+  });
+
+  describe('getFunctionModelName', () => {
+    let client;
+
+    beforeEach(() => {
+      client = new PluginsClient('dummy_api_key');
+    });
+
+    test('should return the input when it includes a dash followed by four digits', () => {
+      expect(client.getFunctionModelName('-1234')).toBe('-1234');
+      expect(client.getFunctionModelName('gpt-4-5678-preview')).toBe('gpt-4-5678-preview');
+    });
+
+    test('should return the input for all function-capable models (`0613` models and above)', () => {
+      expect(client.getFunctionModelName('gpt-4-0613')).toBe('gpt-4-0613');
+      expect(client.getFunctionModelName('gpt-4-32k-0613')).toBe('gpt-4-32k-0613');
+      expect(client.getFunctionModelName('gpt-3.5-turbo-0613')).toBe('gpt-3.5-turbo-0613');
+      expect(client.getFunctionModelName('gpt-3.5-turbo-16k-0613')).toBe('gpt-3.5-turbo-16k-0613');
+      expect(client.getFunctionModelName('gpt-3.5-turbo-1106')).toBe('gpt-3.5-turbo-1106');
+      expect(client.getFunctionModelName('gpt-4-1106-preview')).toBe('gpt-4-1106-preview');
+      expect(client.getFunctionModelName('gpt-4-1106')).toBe('gpt-4-1106');
+    });
+
+    test('should return the corresponding model if input is non-function capable (`0314` models)', () => {
+      expect(client.getFunctionModelName('gpt-4-0314')).toBe('gpt-4');
+      expect(client.getFunctionModelName('gpt-4-32k-0314')).toBe('gpt-4');
+      expect(client.getFunctionModelName('gpt-3.5-turbo-0314')).toBe('gpt-3.5-turbo');
+      expect(client.getFunctionModelName('gpt-3.5-turbo-16k-0314')).toBe('gpt-3.5-turbo');
+    });
+
+    test('should return "gpt-3.5-turbo" when the input includes "gpt-3.5-turbo"', () => {
+      expect(client.getFunctionModelName('test gpt-3.5-turbo model')).toBe('gpt-3.5-turbo');
+    });
+
+    test('should return "gpt-4" when the input includes "gpt-4"', () => {
+      expect(client.getFunctionModelName('testing gpt-4')).toBe('gpt-4');
+    });
+
+    test('should return "gpt-3.5-turbo" for input that does not meet any specific condition', () => {
+      expect(client.getFunctionModelName('random string')).toBe('gpt-3.5-turbo');
+      expect(client.getFunctionModelName('')).toBe('gpt-3.5-turbo');
+    });
+  });
+
+  describe('Azure OpenAI tests specific to Plugins', () => {
+    // TODO: add more tests for Azure OpenAI integration with Plugins
+    // let client;
+    // beforeEach(() => {
+    //   client = new PluginsClient('dummy_api_key');
+    // });
+
+    test('should not call getFunctionModelName when azure options are set', () => {
+      const spy = jest.spyOn(PluginsClient.prototype, 'getFunctionModelName');
+      const model = 'gpt-4-turbo';
+
+      // note, without the azure change in PR #1766, `getFunctionModelName` is called twice
+      const testClient = new PluginsClient('dummy_api_key', {
+        agentOptions: {
+          model,
+          agent: 'functions',
+        },
+        azure: defaultAzureOptions,
+      });
+
+      expect(spy).not.toHaveBeenCalled();
+      expect(testClient.agentOptions.model).toBe(model);
+
+      spy.mockRestore();
+    });
+  });
+
+  describe('sendMessage with filtered tools', () => {
+    let TestAgent;
+    const apiKey = 'fake-api-key';
+    const mockTools = [{ name: 'tool1' }, { name: 'tool2' }, { name: 'tool3' }, { name: 'tool4' }];
+
+    beforeEach(() => {
+      TestAgent = new PluginsClient(apiKey, {
+        tools: mockTools,
+        modelOptions: {
+          model: 'gpt-3.5-turbo',
+          temperature: 0,
+          max_tokens: 2,
+        },
+        agentOptions: {
+          model: 'gpt-3.5-turbo',
+        },
+      });
+
+      TestAgent.options.req = {
+        app: {
+          locals: {},
+        },
+      };
+
+      TestAgent.sendMessage = jest.fn().mockImplementation(async () => {
+        const { filteredTools = [], includedTools = [] } = TestAgent.options.req.app.locals;
+
+        if (includedTools.length > 0) {
+          const tools = TestAgent.options.tools.filter((plugin) =>
+            includedTools.includes(plugin.name),
+          );
+          TestAgent.options.tools = tools;
+        } else {
+          const tools = TestAgent.options.tools.filter(
+            (plugin) => !filteredTools.includes(plugin.name),
+          );
+          TestAgent.options.tools = tools;
+        }
+
+        return {
+          text: 'Mocked response',
+          tools: TestAgent.options.tools,
+        };
+      });
+    });
+
+    test('should filter out tools when filteredTools is provided', async () => {
+      TestAgent.options.req.app.locals.filteredTools = ['tool1', 'tool3'];
+      const response = await TestAgent.sendMessage('Test message');
+      expect(response.tools).toHaveLength(2);
+      expect(response.tools).toEqual(
+        expect.arrayContaining([
+          expect.objectContaining({ name: 'tool2' }),
+          expect.objectContaining({ name: 'tool4' }),
+        ]),
+      );
+    });
+
+    test('should only include specified tools when includedTools is provided', async () => {
+      TestAgent.options.req.app.locals.includedTools = ['tool2', 'tool4'];
+      const response = await TestAgent.sendMessage('Test message');
+      expect(response.tools).toHaveLength(2);
+      expect(response.tools).toEqual(
+        expect.arrayContaining([
+          expect.objectContaining({ name: 'tool2' }),
+          expect.objectContaining({ name: 'tool4' }),
+        ]),
+      );
+    });
+
+    test('should prioritize includedTools over filteredTools', async () => {
+      TestAgent.options.req.app.locals.filteredTools = ['tool1', 'tool3'];
+      TestAgent.options.req.app.locals.includedTools = ['tool1', 'tool2'];
+      const response = await TestAgent.sendMessage('Test message');
+      expect(response.tools).toHaveLength(2);
+      expect(response.tools).toEqual(
+        expect.arrayContaining([
+          expect.objectContaining({ name: 'tool1' }),
+          expect.objectContaining({ name: 'tool2' }),
+        ]),
+      );
+    });
+
+    test('should not modify tools when no filters are provided', async () => {
+      const response = await TestAgent.sendMessage('Test message');
+      expect(response.tools).toHaveLength(4);
+      expect(response.tools).toEqual(expect.arrayContaining(mockTools));
+    });
+  });
+});

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

@@ -107,12 +107,6 @@ const getImageEditPromptDescription = () => {
   return process.env.IMAGE_EDIT_OAI_PROMPT_DESCRIPTION || DEFAULT_IMAGE_EDIT_PROMPT_DESCRIPTION;
 };
 
-function createAbortHandler() {
-  return function () {
-    logger.debug('[ImageGenOAI] Image generation aborted');
-  };
-}
-
 /**
  * Creates OpenAI Image tools (generation and editing)
  * @param {Object} fields - Configuration fields
@@ -207,18 +201,10 @@ function createOpenAIImageTools(fields = {}) {
       }
 
       let resp;
-      /** @type {AbortSignal} */
-      let derivedSignal = null;
-      /** @type {() => void} */
-      let abortHandler = null;
-
       try {
-        if (runnableConfig?.signal) {
-          derivedSignal = AbortSignal.any([runnableConfig.signal]);
-          abortHandler = createAbortHandler();
-          derivedSignal.addEventListener('abort', abortHandler, { once: true });
-        }
-
+        const derivedSignal = runnableConfig?.signal
+          ? AbortSignal.any([runnableConfig.signal])
+          : undefined;
         resp = await openai.images.generate(
           {
             model: 'gpt-image-1',
@@ -242,10 +228,6 @@ function createOpenAIImageTools(fields = {}) {
         logAxiosError({ error, message });
         return returnValue(`Something went wrong when trying to generate the image. The OpenAI API may be unavailable:
 Error Message: ${error.message}`);
-      } finally {
-        if (abortHandler && derivedSignal) {
-          derivedSignal.removeEventListener('abort', abortHandler);
-        }
       }
 
       if (!resp) {
@@ -427,17 +409,10 @@ Error Message: ${error.message}`);
         headers['Authorization'] = `Bearer ${apiKey}`;
       }
 
-      /** @type {AbortSignal} */
-      let derivedSignal = null;
-      /** @type {() => void} */
-      let abortHandler = null;
-
       try {
-        if (runnableConfig?.signal) {
-          derivedSignal = AbortSignal.any([runnableConfig.signal]);
-          abortHandler = createAbortHandler();
-          derivedSignal.addEventListener('abort', abortHandler, { once: true });
-        }
+        const derivedSignal = runnableConfig?.signal
+          ? AbortSignal.any([runnableConfig.signal])
+          : undefined;
 
         /** @type {import('axios').AxiosRequestConfig} */
         const axiosConfig = {
@@ -492,10 +467,6 @@ Error Message: ${error.message}`);
         logAxiosError({ error, message });
         return returnValue(`Something went wrong when trying to edit the image. The OpenAI API may be unavailable:
 Error Message: ${error.message || 'Unknown error'}`);
-      } finally {
-        if (abortHandler && derivedSignal) {
-          derivedSignal.removeEventListener('abort', abortHandler);
-        }
       }
     },
     {

api/app/clients/tools/util/fileSearch.js

@@ -1,10 +1,9 @@
 const { z } = require('zod');
 const axios = require('axios');
 const { tool } = require('@langchain/core/tools');
-const { logger } = require('@librechat/data-schemas');
 const { Tools, EToolResources } = require('librechat-data-provider');
-const { generateShortLivedToken } = require('~/server/services/AuthService');
 const { getFiles } = require('~/models/File');
+const { logger } = require('~/config');
 
 /**
  *
@@ -60,7 +59,7 @@ const createFileSearchTool = async ({ req, files, entity_id }) => {
       if (files.length === 0) {
         return 'No files to search. Instruct the user to add files for the search.';
       }
-      const jwtToken = generateShortLivedToken(req.user.id);
+      const jwtToken = req.headers.authorization.split(' ')[1];
       if (!jwtToken) {
         return 'There was an error authenticating the file search request.';
       }

api/cache/banViolation.js

@@ -1,8 +1,7 @@
 const { logger } = require('@librechat/data-schemas');
-const { isEnabled, math } = require('@librechat/api');
 const { ViolationTypes } = require('librechat-data-provider');
+const { isEnabled, math, removePorts } = require('~/server/utils');
 const { deleteAllUserSessions } = require('~/models');
-const { removePorts } = require('~/server/utils');
 const getLogStores = require('./getLogStores');
 
 const { BAN_VIOLATIONS, BAN_INTERVAL } = process.env ?? {};

api/cache/getLogStores.js

@@ -1,7 +1,7 @@
 const { Keyv } = require('keyv');
-const { isEnabled, math } = require('@librechat/api');
 const { CacheKeys, ViolationTypes, Time } = require('librechat-data-provider');
 const { logFile, violationFile } = require('./keyvFiles');
+const { isEnabled, math } = require('~/server/utils');
 const keyvRedis = require('./keyvRedis');
 const keyvMongo = require('./keyvMongo');
 

api/db/indexSync.js

@@ -1,11 +1,8 @@
 const mongoose = require('mongoose');
 const { MeiliSearch } = require('meilisearch');
 const { logger } = require('@librechat/data-schemas');
-const { FlowStateManager } = require('@librechat/api');
-const { CacheKeys } = require('librechat-data-provider');
 
 const { isEnabled } = require('~/server/utils');
-const { getLogStores } = require('~/cache');
 
 const Conversation = mongoose.models.Conversation;
 const Message = mongoose.models.Message;
@@ -31,123 +28,43 @@ class MeiliSearchClient {
   }
 }
 
-/**
- * Performs the actual sync operations for messages and conversations
- */
-async function performSync() {
-  const client = MeiliSearchClient.getInstance();
-
-  const { status } = await client.health();
-  if (status !== 'available') {
-    throw new Error('Meilisearch not available');
-  }
-
-  if (indexingDisabled === true) {
-    logger.info('[indexSync] Indexing is disabled, skipping...');
-    return { messagesSync: false, convosSync: false };
-  }
-
-  let messagesSync = false;
-  let convosSync = false;
-
-  // Check if we need to sync messages
-  const messageProgress = await Message.getSyncProgress();
-  if (!messageProgress.isComplete) {
-    logger.info(
-      `[indexSync] Messages need syncing: ${messageProgress.totalProcessed}/${messageProgress.totalDocuments} indexed`,
-    );
-
-    // Check if we should do a full sync or incremental
-    const messageCount = await Message.countDocuments();
-    const messagesIndexed = messageProgress.totalProcessed;
-    const syncThreshold = parseInt(process.env.MEILI_SYNC_THRESHOLD || '1000', 10);
-
-    if (messageCount - messagesIndexed > syncThreshold) {
-      logger.info('[indexSync] Starting full message sync due to large difference');
-      await Message.syncWithMeili();
-      messagesSync = true;
-    } else if (messageCount !== messagesIndexed) {
-      logger.warn('[indexSync] Messages out of sync, performing incremental sync');
-      await Message.syncWithMeili();
-      messagesSync = true;
-    }
-  } else {
-    logger.info(
-      `[indexSync] Messages are fully synced: ${messageProgress.totalProcessed}/${messageProgress.totalDocuments}`,
-    );
-  }
-
-  // Check if we need to sync conversations
-  const convoProgress = await Conversation.getSyncProgress();
-  if (!convoProgress.isComplete) {
-    logger.info(
-      `[indexSync] Conversations need syncing: ${convoProgress.totalProcessed}/${convoProgress.totalDocuments} indexed`,
-    );
-
-    const convoCount = await Conversation.countDocuments();
-    const convosIndexed = convoProgress.totalProcessed;
-    const syncThreshold = parseInt(process.env.MEILI_SYNC_THRESHOLD || '1000', 10);
-
-    if (convoCount - convosIndexed > syncThreshold) {
-      logger.info('[indexSync] Starting full conversation sync due to large difference');
-      await Conversation.syncWithMeili();
-      convosSync = true;
-    } else if (convoCount !== convosIndexed) {
-      logger.warn('[indexSync] Convos out of sync, performing incremental sync');
-      await Conversation.syncWithMeili();
-      convosSync = true;
-    }
-  } else {
-    logger.info(
-      `[indexSync] Conversations are fully synced: ${convoProgress.totalProcessed}/${convoProgress.totalDocuments}`,
-    );
-  }
-
-  return { messagesSync, convosSync };
-}
-
-/**
- * Main index sync function that uses FlowStateManager to prevent concurrent execution
- */
 async function indexSync() {
   if (!searchEnabled) {
     return;
   }
-
-  logger.info('[indexSync] Starting index synchronization check...');
-
   try {
-    // Get or create FlowStateManager instance
-    const flowsCache = getLogStores(CacheKeys.FLOWS);
-    if (!flowsCache) {
-      logger.warn('[indexSync] Flows cache not available, falling back to direct sync');
-      return await performSync();
+    const client = MeiliSearchClient.getInstance();
+
+    const { status } = await client.health();
+    if (status !== 'available') {
+      throw new Error('Meilisearch not available');
     }
 
-    const flowManager = new FlowStateManager(flowsCache, {
-      ttl: 60000 * 10, // 10 minutes TTL for sync operations
-    });
-
-    // Use a unique flow ID for the sync operation
-    const flowId = 'meili-index-sync';
-    const flowType = 'MEILI_SYNC';
-
-    // This will only execute the handler if no other instance is running the sync
-    const result = await flowManager.createFlowWithHandler(flowId, flowType, performSync);
-
-    if (result.messagesSync || result.convosSync) {
-      logger.info('[indexSync] Sync completed successfully');
-    } else {
-      logger.debug('[indexSync] No sync was needed');
-    }
-
-    return result;
-  } catch (err) {
-    if (err.message.includes('flow already exists')) {
-      logger.info('[indexSync] Sync already running on another instance');
+    if (indexingDisabled === true) {
+      logger.info('[indexSync] Indexing is disabled, skipping...');
       return;
     }
 
+    const messageCount = await Message.countDocuments();
+    const convoCount = await Conversation.countDocuments();
+    const messages = await client.index('messages').getStats();
+    const convos = await client.index('convos').getStats();
+    const messagesIndexed = messages.numberOfDocuments;
+    const convosIndexed = convos.numberOfDocuments;
+
+    logger.debug(`[indexSync] There are ${messageCount} messages and ${messagesIndexed} indexed`);
+    logger.debug(`[indexSync] There are ${convoCount} convos and ${convosIndexed} indexed`);
+
+    if (messageCount !== messagesIndexed) {
+      logger.debug('[indexSync] Messages out of sync, indexing');
+      Message.syncWithMeili();
+    }
+
+    if (convoCount !== convosIndexed) {
+      logger.debug('[indexSync] Convos out of sync, indexing');
+      Conversation.syncWithMeili();
+    }
+  } catch (err) {
     if (err.message.includes('not found')) {
       logger.debug('[indexSync] Creating indices...');
       currentTimeout = setTimeout(async () => {

api/models/Agent.js

@@ -4,7 +4,6 @@ const { logger } = require('@librechat/data-schemas');
 const { SystemRoles, Tools, actionDelimiter } = require('librechat-data-provider');
 const { GLOBAL_PROJECT_NAME, EPHEMERAL_AGENT_ID, mcp_delimiter } =
   require('librechat-data-provider').Constants;
-const { CONFIG_STORE, STARTUP_CONFIG } = require('librechat-data-provider').CacheKeys;
 const {
   getProjectByName,
   addAgentIdsToProject,
@@ -12,7 +11,6 @@ const {
   removeAgentFromAllProjects,
 } = require('./Project');
 const { getCachedTools } = require('~/server/services/Config');
-const getLogStores = require('~/cache/getLogStores');
 const { getActions } = require('./Action');
 const { Agent } = require('~/db/models');
 
@@ -23,7 +21,7 @@ const { Agent } = require('~/db/models');
  * @throws {Error} If the agent creation fails.
  */
 const createAgent = async (agentData) => {
-  const { author, ...versionData } = agentData;
+  const { author: _author, ...versionData } = agentData;
   const timestamp = new Date();
   const initialAgentData = {
     ...agentData,
@@ -70,9 +68,6 @@ const loadEphemeralAgent = async ({ req, agent_id, endpoint, model_parameters: _
   if (ephemeralAgent?.execute_code === true) {
     tools.push(Tools.execute_code);
   }
-  if (ephemeralAgent?.file_search === true) {
-    tools.push(Tools.file_search);
-  }
   if (ephemeralAgent?.web_search === true) {
     tools.push(Tools.web_search);
   }
@@ -126,29 +121,7 @@ const loadAgent = async ({ req, agent_id, endpoint, model_parameters }) => {
   }
 
   agent.version = agent.versions ? agent.versions.length : 0;
-
-  if (agent.author.toString() === req.user.id) {
-    return agent;
-  }
-
-  if (!agent.projectIds) {
-    return null;
-  }
-
-  const cache = getLogStores(CONFIG_STORE);
-  /** @type {TStartupConfig} */
-  const cachedStartupConfig = await cache.get(STARTUP_CONFIG);
-  let { instanceProjectId } = cachedStartupConfig ?? {};
-  if (!instanceProjectId) {
-    instanceProjectId = (await getProjectByName(GLOBAL_PROJECT_NAME, '_id'))._id.toString();
-  }
-
-  for (const projectObjectId of agent.projectIds) {
-    const projectId = projectObjectId.toString();
-    if (projectId === instanceProjectId) {
-      return agent;
-    }
-  }
+  return agent;
 };
 
 /**
@@ -178,7 +151,7 @@ const isDuplicateVersion = (updateData, currentData, versions, actionsHash = nul
     'actionsHash', // Exclude actionsHash from direct comparison
   ];
 
-  const { $push, $pull, $addToSet, ...directUpdates } = updateData;
+  const { $push: _$push, $pull: _$pull, $addToSet: _$addToSet, ...directUpdates } = updateData;
 
   if (Object.keys(directUpdates).length === 0 && !actionsHash) {
     return null;
@@ -273,7 +246,14 @@ const updateAgent = async (searchParameter, updateData, options = {}) => {
 
   const currentAgent = await Agent.findOne(searchParameter);
   if (currentAgent) {
-    const { __v, _id, id, versions, author, ...versionData } = currentAgent.toObject();
+    const {
+      __v,
+      _id,
+      id: __id,
+      versions,
+      author: _author,
+      ...versionData
+    } = currentAgent.toObject();
     const { $push, $pull, $addToSet, ...directUpdates } = updateData;
 
     let actionsHash = null;
@@ -464,8 +444,110 @@ const deleteAgent = async (searchParameter) => {
   return agent;
 };
 
+/**
+ * Get agents by accessible IDs with optional cursor-based pagination.
+ * @param {Object} params - The parameters for getting accessible agents.
+ * @param {Array} [params.accessibleIds] - Array of agent ObjectIds the user has ACL access to.
+ * @param {Object} [params.otherParams] - Additional query parameters (including author filter).
+ * @param {number} [params.limit] - Number of agents to return (max 100). If not provided, returns all agents.
+ * @param {string} [params.after] - Cursor for pagination - get agents after this cursor. // base64 encoded JSON string with updatedAt and _id.
+ * @returns {Promise<Object>} A promise that resolves to an object containing the agents data and pagination info.
+ */
+const getListAgentsByAccess = async ({
+  accessibleIds = [],
+  otherParams = {},
+  limit = null,
+  after = null,
+}) => {
+  const isPaginated = limit !== null && limit !== undefined;
+  const normalizedLimit = isPaginated ? Math.min(Math.max(1, parseInt(limit) || 20), 100) : null;
+
+  // Build base query combining ACL accessible agents with other filters
+  const baseQuery = { ...otherParams };
+
+  if (accessibleIds.length > 0) {
+    baseQuery._id = { $in: accessibleIds };
+  }
+
+  // Add cursor condition
+  if (after) {
+    try {
+      const cursor = JSON.parse(Buffer.from(after, 'base64').toString('utf8'));
+      const { updatedAt, _id } = cursor;
+
+      const cursorCondition = {
+        $or: [
+          { updatedAt: { $lt: new Date(updatedAt) } },
+          { updatedAt: new Date(updatedAt), _id: { $gt: mongoose.Types.ObjectId(_id) } },
+        ],
+      };
+
+      // Merge cursor condition with base query
+      if (Object.keys(baseQuery).length > 0) {
+        baseQuery.$and = [{ ...baseQuery }, cursorCondition];
+        // Remove the original conditions from baseQuery to avoid duplication
+        Object.keys(baseQuery).forEach((key) => {
+          if (key !== '$and') delete baseQuery[key];
+        });
+      } else {
+        Object.assign(baseQuery, cursorCondition);
+      }
+    } catch (error) {
+      logger.warn('Invalid cursor:', error.message);
+    }
+  }
+
+  let query = Agent.find(baseQuery, {
+    id: 1,
+    _id: 1,
+    name: 1,
+    avatar: 1,
+    author: 1,
+    projectIds: 1,
+    description: 1,
+    updatedAt: 1,
+  }).sort({ updatedAt: -1, _id: 1 });
+
+  // Only apply limit if pagination is requested
+  if (isPaginated) {
+    query = query.limit(normalizedLimit + 1);
+  }
+
+  const agents = await query.lean();
+
+  const hasMore = isPaginated ? agents.length > normalizedLimit : false;
+  const data = (isPaginated ? agents.slice(0, normalizedLimit) : agents).map((agent) => {
+    if (agent.author) {
+      agent.author = agent.author.toString();
+    }
+    return agent;
+  });
+
+  // Generate next cursor only if paginated
+  let nextCursor = null;
+  if (isPaginated && hasMore && data.length > 0) {
+    const lastAgent = agents[normalizedLimit - 1];
+    nextCursor = Buffer.from(
+      JSON.stringify({
+        updatedAt: lastAgent.updatedAt.toISOString(),
+        _id: lastAgent._id.toString(),
+      }),
+    ).toString('base64');
+  }
+
+  return {
+    object: 'list',
+    data,
+    first_id: data.length > 0 ? data[0].id : null,
+    last_id: data.length > 0 ? data[data.length - 1].id : null,
+    has_more: hasMore,
+    after: nextCursor,
+  };
+};
+
 /**
  * Get all agents.
+ * @deprecated Use getListAgentsByAccess for ACL-aware agent listing
  * @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.
@@ -484,12 +566,13 @@ const getListAgents = async (searchParameter) => {
   const agents = (
     await Agent.find(query, {
       id: 1,
-      _id: 0,
+      _id: 1,
       name: 1,
       avatar: 1,
       author: 1,
       projectIds: 1,
       description: 1,
+      // @deprecated - isCollaborative replaced by ACL permissions
       isCollaborative: 1,
     }).lean()
   ).map((agent) => {
@@ -673,6 +756,7 @@ module.exports = {
   revertAgentVersion,
   updateAgentProjects,
   addAgentResourceFile,
+  getListAgentsByAccess,
   removeAgentResourceFiles,
   generateActionMetadataHash,
 };

api/models/Agent.spec.js

@@ -43,7 +43,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -413,7 +413,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -670,7 +670,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -1332,7 +1332,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -1514,7 +1514,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -1633,7 +1633,7 @@ describe('models/Agent', () => {
       expect(result.version).toBe(1);
     });
 
-    test('should return null when user is not author and agent has no projectIds', async () => {
+    test('should return agent even when user is not author (permissions checked at route level)', async () => {
       const authorId = new mongoose.Types.ObjectId();
       const userId = new mongoose.Types.ObjectId();
       const agentId = `agent_${uuidv4()}`;
@@ -1654,7 +1654,11 @@ describe('models/Agent', () => {
         model_parameters: { model: 'gpt-4' },
       });
 
-      expect(result).toBeFalsy();
+      // With the new permission system, loadAgent returns the agent regardless of permissions
+      // Permission checks are handled at the route level via middleware
+      expect(result).toBeTruthy();
+      expect(result.id).toBe(agentId);
+      expect(result.name).toBe('Test Agent');
     });
 
     test('should handle ephemeral agent with no MCP servers', async () => {
@@ -1762,7 +1766,7 @@ describe('models/Agent', () => {
         }
       });
 
-      test('should handle loadAgent with agent from different project', async () => {
+      test('should return agent from different project (permissions checked at route level)', async () => {
         const authorId = new mongoose.Types.ObjectId();
         const userId = new mongoose.Types.ObjectId();
         const agentId = `agent_${uuidv4()}`;
@@ -1785,7 +1789,11 @@ describe('models/Agent', () => {
           model_parameters: { model: 'gpt-4' },
         });
 
-        expect(result).toBeFalsy();
+        // With the new permission system, loadAgent returns the agent regardless of permissions
+        // Permission checks are handled at the route level via middleware
+        expect(result).toBeTruthy();
+        expect(result.id).toBe(agentId);
+        expect(result.name).toBe('Project Agent');
       });
     });
   });
@@ -1798,7 +1806,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();
@@ -2350,7 +2358,7 @@ describe('models/Agent', () => {
       const mongoUri = mongoServer.getUri();
       Agent = mongoose.models.Agent || mongoose.model('Agent', agentSchema);
       await mongoose.connect(mongoUri);
-    }, 20000);
+    });
 
     afterAll(async () => {
       await mongoose.disconnect();

api/models/Conversation.js

@@ -1,6 +1,4 @@
 const { logger } = require('@librechat/data-schemas');
-const { createTempChatExpirationDate } = require('@librechat/api');
-const getCustomConfig = require('~/server/services/Config/loadCustomConfig');
 const { getMessages, deleteMessages } = require('./Message');
 const { Conversation } = require('~/db/models');
 
@@ -100,15 +98,10 @@ module.exports = {
         update.conversationId = newConversationId;
       }
 
-      if (req?.body?.isTemporary) {
-        try {
-          const customConfig = await getCustomConfig();
-          update.expiredAt = createTempChatExpirationDate(customConfig);
-        } catch (err) {
-          logger.error('Error creating temporary chat expiration date:', err);
-          logger.info(`---\`saveConvo\` context: ${metadata?.context}`);
-          update.expiredAt = null;
-        }
+      if (req.body.isTemporary) {
+        const expiredAt = new Date();
+        expiredAt.setDate(expiredAt.getDate() + 30);
+        update.expiredAt = expiredAt;
       } else {
         update.expiredAt = null;
       }

api/models/File.js

@@ -1,5 +1,5 @@
 const { logger } = require('@librechat/data-schemas');
-const { EToolResources, FileContext } = require('librechat-data-provider');
+const { EToolResources } = require('librechat-data-provider');
 const { File } = require('~/db/models');
 
 /**
@@ -32,19 +32,19 @@ const getFiles = async (filter, _sortOptions, selectFields = { text: 0 }) => {
  * @returns {Promise<Array<MongoFile>>} Files that match the criteria
  */
 const getToolFilesByIds = async (fileIds, toolResourceSet) => {
-  if (!fileIds || !fileIds.length || !toolResourceSet?.size) {
+  if (!fileIds || !fileIds.length) {
     return [];
   }
 
   try {
     const filter = {
       file_id: { $in: fileIds },
-      $or: [],
     };
 
-    if (toolResourceSet.has(EToolResources.ocr)) {
-      filter.$or.push({ text: { $exists: true, $ne: null }, context: FileContext.agents });
+    if (toolResourceSet.size) {
+      filter.$or = [];
     }
+
     if (toolResourceSet.has(EToolResources.file_search)) {
       filter.$or.push({ embedded: true });
     }

api/models/Message.js

@@ -1,7 +1,5 @@
 const { z } = require('zod');
 const { logger } = require('@librechat/data-schemas');
-const { createTempChatExpirationDate } = require('@librechat/api');
-const getCustomConfig = require('~/server/services/Config/loadCustomConfig');
 const { Message } = require('~/db/models');
 
 const idSchema = z.string().uuid();
@@ -56,14 +54,9 @@ async function saveMessage(req, params, metadata) {
     };
 
     if (req?.body?.isTemporary) {
-      try {
-        const customConfig = await getCustomConfig();
-        update.expiredAt = createTempChatExpirationDate(customConfig);
-      } catch (err) {
-        logger.error('Error creating temporary chat expiration date:', err);
-        logger.info(`---\`saveMessage\` context: ${metadata?.context}`);
-        update.expiredAt = null;
-      }
+      const expiredAt = new Date();
+      expiredAt.setDate(expiredAt.getDate() + 30);
+      update.expiredAt = expiredAt;
     } else {
       update.expiredAt = null;
     }

api/package.json

@@ -48,13 +48,15 @@
     "@langchain/google-genai": "^0.2.13",
     "@langchain/google-vertexai": "^0.2.13",
     "@langchain/textsplitters": "^0.1.0",
-    "@librechat/agents": "^2.4.46",
+    "@librechat/agents": "^2.4.41",
     "@librechat/api": "*",
     "@librechat/data-schemas": "*",
     "@node-saml/passport-saml": "^5.0.0",
+    "@microsoft/microsoft-graph-client": "^3.0.7",
     "@waylaidwanderer/fetch-event-source": "^3.0.1",
     "axios": "^1.8.2",
     "bcryptjs": "^2.4.3",
+    "cohere-ai": "^7.9.1",
     "compression": "^1.7.4",
     "connect-redis": "^7.1.0",
     "cookie": "^0.7.2",

api/server/cleanup.js

@@ -169,6 +169,9 @@ function disposeClient(client) {
       client.isGenerativeModel = null;
     }
     // Properties specific to OpenAIClient
+    if (client.ChatGPTClient) {
+      client.ChatGPTClient = null;
+    }
     if (client.completionsUrl) {
       client.completionsUrl = null;
     }

api/server/controllers/AskController.js

@@ -0,0 +1,282 @@
+const { getResponseSender, Constants } = require('librechat-data-provider');
+const {
+  handleAbortError,
+  createAbortController,
+  cleanupAbortController,
+} = require('~/server/middleware');
+const {
+  disposeClient,
+  processReqData,
+  clientRegistry,
+  requestDataMap,
+} = require('~/server/cleanup');
+const { sendMessage, createOnProgress } = require('~/server/utils');
+const { saveMessage } = require('~/models');
+const { logger } = require('~/config');
+
+const AskController = async (req, res, next, initializeClient, addTitle) => {
+  let {
+    text,
+    endpointOption,
+    conversationId,
+    modelDisplayLabel,
+    parentMessageId = null,
+    overrideParentMessageId = null,
+  } = req.body;
+
+  let client = null;
+  let abortKey = null;
+  let cleanupHandlers = [];
+  let clientRef = null;
+
+  logger.debug('[AskController]', {
+    text,
+    conversationId,
+    ...endpointOption,
+    modelsConfig: endpointOption?.modelsConfig ? 'exists' : '',
+  });
+
+  let userMessage = null;
+  let userMessagePromise = null;
+  let promptTokens = null;
+  let userMessageId = null;
+  let responseMessageId = null;
+  let getAbortData = null;
+
+  const sender = getResponseSender({
+    ...endpointOption,
+    model: endpointOption.modelOptions.model,
+    modelDisplayLabel,
+  });
+  const initialConversationId = conversationId;
+  const newConvo = !initialConversationId;
+  const userId = req.user.id;
+
+  let reqDataContext = {
+    userMessage,
+    userMessagePromise,
+    responseMessageId,
+    promptTokens,
+    conversationId,
+    userMessageId,
+  };
+
+  const updateReqData = (data = {}) => {
+    reqDataContext = processReqData(data, reqDataContext);
+    abortKey = reqDataContext.abortKey;
+    userMessage = reqDataContext.userMessage;
+    userMessagePromise = reqDataContext.userMessagePromise;
+    responseMessageId = reqDataContext.responseMessageId;
+    promptTokens = reqDataContext.promptTokens;
+    conversationId = reqDataContext.conversationId;
+    userMessageId = reqDataContext.userMessageId;
+  };
+
+  let { onProgress: progressCallback, getPartialText } = createOnProgress();
+
+  const performCleanup = () => {
+    logger.debug('[AskController] Performing cleanup');
+    if (Array.isArray(cleanupHandlers)) {
+      for (const handler of cleanupHandlers) {
+        try {
+          if (typeof handler === 'function') {
+            handler();
+          }
+        } catch (e) {
+          // Ignore
+        }
+      }
+    }
+
+    if (abortKey) {
+      logger.debug('[AskController] Cleaning up abort controller');
+      cleanupAbortController(abortKey);
+      abortKey = null;
+    }
+
+    if (client) {
+      disposeClient(client);
+      client = null;
+    }
+
+    reqDataContext = null;
+    userMessage = null;
+    userMessagePromise = null;
+    promptTokens = null;
+    getAbortData = null;
+    progressCallback = null;
+    endpointOption = null;
+    cleanupHandlers = null;
+    addTitle = null;
+
+    if (requestDataMap.has(req)) {
+      requestDataMap.delete(req);
+    }
+    logger.debug('[AskController] Cleanup completed');
+  };
+
+  try {
+    ({ client } = await initializeClient({ req, res, endpointOption }));
+    if (clientRegistry && client) {
+      clientRegistry.register(client, { userId }, client);
+    }
+
+    if (client) {
+      requestDataMap.set(req, { client });
+    }
+
+    clientRef = new WeakRef(client);
+
+    getAbortData = () => {
+      const currentClient = clientRef?.deref();
+      const currentText =
+        currentClient?.getStreamText != null ? currentClient.getStreamText() : getPartialText();
+
+      return {
+        sender,
+        conversationId,
+        messageId: reqDataContext.responseMessageId,
+        parentMessageId: overrideParentMessageId ?? userMessageId,
+        text: currentText,
+        userMessage: userMessage,
+        userMessagePromise: userMessagePromise,
+        promptTokens: reqDataContext.promptTokens,
+      };
+    };
+
+    const { onStart, abortController } = createAbortController(
+      req,
+      res,
+      getAbortData,
+      updateReqData,
+    );
+
+    const closeHandler = () => {
+      logger.debug('[AskController] Request closed');
+      if (!abortController || abortController.signal.aborted || abortController.requestCompleted) {
+        return;
+      }
+      abortController.abort();
+      logger.debug('[AskController] Request aborted on close');
+    };
+
+    res.on('close', closeHandler);
+    cleanupHandlers.push(() => {
+      try {
+        res.removeListener('close', closeHandler);
+      } catch (e) {
+        // Ignore
+      }
+    });
+
+    const messageOptions = {
+      user: userId,
+      parentMessageId,
+      conversationId: reqDataContext.conversationId,
+      overrideParentMessageId,
+      getReqData: updateReqData,
+      onStart,
+      abortController,
+      progressCallback,
+      progressOptions: {
+        res,
+      },
+    };
+
+    /** @type {TMessage} */
+    let response = await client.sendMessage(text, messageOptions);
+    response.endpoint = endpointOption.endpoint;
+
+    const databasePromise = response.databasePromise;
+    delete response.databasePromise;
+
+    const { conversation: convoData = {} } = await databasePromise;
+    const conversation = { ...convoData };
+    conversation.title =
+      conversation && !conversation.title ? null : conversation?.title || 'New Chat';
+
+    const latestUserMessage = reqDataContext.userMessage;
+
+    if (client?.options?.attachments && latestUserMessage) {
+      latestUserMessage.files = client.options.attachments;
+      if (endpointOption?.modelOptions?.model) {
+        conversation.model = endpointOption.modelOptions.model;
+      }
+      delete latestUserMessage.image_urls;
+    }
+
+    if (!abortController.signal.aborted) {
+      const finalResponseMessage = { ...response };
+
+      sendMessage(res, {
+        final: true,
+        conversation,
+        title: conversation.title,
+        requestMessage: latestUserMessage,
+        responseMessage: finalResponseMessage,
+      });
+      res.end();
+
+      if (client?.savedMessageIds && !client.savedMessageIds.has(response.messageId)) {
+        await saveMessage(
+          req,
+          { ...finalResponseMessage, user: userId },
+          { context: 'api/server/controllers/AskController.js - response end' },
+        );
+      }
+    }
+
+    if (!client?.skipSaveUserMessage && latestUserMessage) {
+      await saveMessage(req, latestUserMessage, {
+        context: "api/server/controllers/AskController.js - don't skip saving user message",
+      });
+    }
+
+    if (typeof addTitle === 'function' && parentMessageId === Constants.NO_PARENT && newConvo) {
+      addTitle(req, {
+        text,
+        response: { ...response },
+        client,
+      })
+        .then(() => {
+          logger.debug('[AskController] Title generation started');
+        })
+        .catch((err) => {
+          logger.error('[AskController] Error in title generation', err);
+        })
+        .finally(() => {
+          logger.debug('[AskController] Title generation completed');
+          performCleanup();
+        });
+    } else {
+      performCleanup();
+    }
+  } catch (error) {
+    logger.error('[AskController] Error handling request', error);
+    let partialText = '';
+    try {
+      const currentClient = clientRef?.deref();
+      partialText =
+        currentClient?.getStreamText != null ? currentClient.getStreamText() : getPartialText();
+    } catch (getTextError) {
+      logger.error('[AskController] Error calling getText() during error handling', getTextError);
+    }
+
+    handleAbortError(res, req, error, {
+      sender,
+      partialText,
+      conversationId: reqDataContext.conversationId,
+      messageId: reqDataContext.responseMessageId,
+      parentMessageId: overrideParentMessageId ?? reqDataContext.userMessageId ?? parentMessageId,
+      userMessageId: reqDataContext.userMessageId,
+    })
+      .catch((err) => {
+        logger.error('[AskController] Error in `handleAbortError` during catch block', err);
+      })
+      .finally(() => {
+        performCleanup();
+      });
+  }
+};
+
+module.exports = AskController;

api/server/controllers/AuthController.js

@@ -1,17 +1,17 @@
 const cookies = require('cookie');
 const jwt = require('jsonwebtoken');
 const openIdClient = require('openid-client');
-const { isEnabled } = require('@librechat/api');
 const { logger } = require('@librechat/data-schemas');
 const {
-  requestPasswordReset,
-  setOpenIDAuthTokens,
+  registerUser,
   resetPassword,
   setAuthTokens,
-  registerUser,
+  requestPasswordReset,
+  setOpenIDAuthTokens,
 } = require('~/server/services/AuthService');
 const { findUser, getUserById, deleteAllUserSessions, findSession } = require('~/models');
 const { getOpenIdConfig } = require('~/strategies');
+const { isEnabled } = require('~/server/utils');
 
 const registrationController = async (req, res) => {
   try {

api/server/controllers/EditController.js

@@ -1,5 +1,3 @@
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const { getResponseSender } = require('librechat-data-provider');
 const {
   handleAbortError,
@@ -12,8 +10,9 @@ const {
   clientRegistry,
   requestDataMap,
 } = require('~/server/cleanup');
-const { createOnProgress } = require('~/server/utils');
+const { sendMessage, createOnProgress } = require('~/server/utils');
 const { saveMessage } = require('~/models');
+const { logger } = require('~/config');
 
 const EditController = async (req, res, next, initializeClient) => {
   let {
@@ -85,7 +84,7 @@ const EditController = async (req, res, next, initializeClient) => {
     }
 
     if (abortKey) {
-      logger.debug('[EditController] Cleaning up abort controller');
+      logger.debug('[AskController] Cleaning up abort controller');
       cleanupAbortController(abortKey);
       abortKey = null;
     }
@@ -199,7 +198,7 @@ const EditController = async (req, res, next, initializeClient) => {
       const finalUserMessage = reqDataContext.userMessage;
       const finalResponseMessage = { ...response };
 
-      sendEvent(res, {
+      sendMessage(res, {
         final: true,
         conversation,
         title: conversation.title,

api/server/controllers/PermissionsController.js

@@ -0,0 +1,437 @@
+/**
+ * @import { TUpdateResourcePermissionsRequest, TUpdateResourcePermissionsResponse } from 'librechat-data-provider'
+ */
+
+const mongoose = require('mongoose');
+const { logger } = require('@librechat/data-schemas');
+const {
+  getAvailableRoles,
+  ensurePrincipalExists,
+  getEffectivePermissions,
+  ensureGroupPrincipalExists,
+  bulkUpdateResourcePermissions,
+} = require('~/server/services/PermissionService');
+const { AclEntry } = require('~/db/models');
+const {
+  searchPrincipals: searchLocalPrincipals,
+  sortPrincipalsByRelevance,
+  calculateRelevanceScore,
+} = require('~/models');
+const {
+  searchEntraIdPrincipals,
+  entraIdPrincipalFeatureEnabled,
+} = require('~/server/services/GraphApiService');
+
+/**
+ * Generic controller for resource permission endpoints
+ * Delegates validation and logic to PermissionService
+ */
+
+/**
+ * Bulk update permissions for a resource (grant, update, remove)
+ * @route PUT /api/{resourceType}/{resourceId}/permissions
+ * @param {Object} req - Express request object
+ * @param {Object} req.params - Route parameters
+ * @param {string} req.params.resourceType - Resource type (e.g., 'agent')
+ * @param {string} req.params.resourceId - Resource ID
+ * @param {TUpdateResourcePermissionsRequest} req.body - Request body
+ * @param {Object} res - Express response object
+ * @returns {Promise<TUpdateResourcePermissionsResponse>} Updated permissions response
+ */
+const updateResourcePermissions = async (req, res) => {
+  try {
+    const { resourceType, resourceId } = req.params;
+    /** @type {TUpdateResourcePermissionsRequest} */
+    const { updated, removed, public: isPublic, publicAccessRoleId } = req.body;
+    const { id: userId } = req.user;
+
+    // Prepare principals for the service call
+    const updatedPrincipals = [];
+    const revokedPrincipals = [];
+
+    // Add updated principals
+    if (updated && Array.isArray(updated)) {
+      updatedPrincipals.push(...updated);
+    }
+
+    // Add public permission if enabled
+    if (isPublic && publicAccessRoleId) {
+      updatedPrincipals.push({
+        type: 'public',
+        id: null,
+        accessRoleId: publicAccessRoleId,
+      });
+    }
+
+    // Prepare authentication context for enhanced group member fetching
+    const useEntraId = entraIdPrincipalFeatureEnabled(req.user);
+    const authHeader = req.headers.authorization;
+    const accessToken =
+      authHeader && authHeader.startsWith('Bearer ') ? authHeader.substring(7) : null;
+    const authContext =
+      useEntraId && accessToken
+        ? {
+            accessToken,
+            sub: req.user.openidId,
+          }
+        : null;
+
+    // Ensure updated principals exist in the database before processing permissions
+    const validatedPrincipals = [];
+    for (const principal of updatedPrincipals) {
+      try {
+        let principalId;
+
+        if (principal.type === 'public') {
+          principalId = null; // Public principals don't need database records
+        } else if (principal.type === 'user') {
+          principalId = await ensurePrincipalExists(principal);
+        } else if (principal.type === 'group') {
+          // Pass authContext to enable member fetching for Entra ID groups when available
+          principalId = await ensureGroupPrincipalExists(principal, authContext);
+        } else {
+          logger.error(`Unsupported principal type: ${principal.type}`);
+          continue; // Skip invalid principal types
+        }
+
+        // Update the principal with the validated ID for ACL operations
+        validatedPrincipals.push({
+          ...principal,
+          id: principalId,
+        });
+      } catch (error) {
+        logger.error('Error ensuring principal exists:', {
+          principal: {
+            type: principal.type,
+            id: principal.id,
+            name: principal.name,
+            source: principal.source,
+          },
+          error: error.message,
+        });
+        // Continue with other principals instead of failing the entire operation
+        continue;
+      }
+    }
+
+    // Add removed principals
+    if (removed && Array.isArray(removed)) {
+      revokedPrincipals.push(...removed);
+    }
+
+    // If public is disabled, add public to revoked list
+    if (!isPublic) {
+      revokedPrincipals.push({
+        type: 'public',
+        id: null,
+      });
+    }
+
+    const results = await bulkUpdateResourcePermissions({
+      resourceType,
+      resourceId,
+      updatedPrincipals: validatedPrincipals,
+      revokedPrincipals,
+      grantedBy: userId,
+    });
+
+    /** @type {TUpdateResourcePermissionsResponse} */
+    const response = {
+      message: 'Permissions updated successfully',
+      results: {
+        principals: results.granted,
+        public: isPublic || false,
+        publicAccessRoleId: isPublic ? publicAccessRoleId : undefined,
+      },
+    };
+
+    res.status(200).json(response);
+  } catch (error) {
+    logger.error('Error updating resource permissions:', error);
+    res.status(400).json({
+      error: 'Failed to update permissions',
+      details: error.message,
+    });
+  }
+};
+
+/**
+ * Get principals with their permission roles for a resource (UI-friendly format)
+ * Uses efficient aggregation pipeline to join User/Group data in single query
+ * @route GET /api/permissions/{resourceType}/{resourceId}
+ */
+const getResourcePermissions = async (req, res) => {
+  try {
+    const { resourceType, resourceId } = req.params;
+
+    // Use aggregation pipeline for efficient single-query data retrieval
+    const results = await AclEntry.aggregate([
+      // Match ACL entries for this resource
+      {
+        $match: {
+          resourceType,
+          resourceId: mongoose.Types.ObjectId.isValid(resourceId)
+            ? mongoose.Types.ObjectId.createFromHexString(resourceId)
+            : resourceId,
+        },
+      },
+      // Lookup AccessRole information
+      {
+        $lookup: {
+          from: 'accessroles',
+          localField: 'roleId',
+          foreignField: '_id',
+          as: 'role',
+        },
+      },
+      // Lookup User information (for user principals)
+      {
+        $lookup: {
+          from: 'users',
+          localField: 'principalId',
+          foreignField: '_id',
+          as: 'userInfo',
+        },
+      },
+      // Lookup Group information (for group principals)
+      {
+        $lookup: {
+          from: 'groups',
+          localField: 'principalId',
+          foreignField: '_id',
+          as: 'groupInfo',
+        },
+      },
+      // Project final structure
+      {
+        $project: {
+          principalType: 1,
+          principalId: 1,
+          accessRoleId: { $arrayElemAt: ['$role.accessRoleId', 0] },
+          userInfo: { $arrayElemAt: ['$userInfo', 0] },
+          groupInfo: { $arrayElemAt: ['$groupInfo', 0] },
+        },
+      },
+    ]);
+
+    const principals = [];
+    let publicPermission = null;
+
+    // Process aggregation results
+    for (const result of results) {
+      if (result.principalType === 'public') {
+        publicPermission = {
+          public: true,
+          publicAccessRoleId: result.accessRoleId,
+        };
+      } else if (result.principalType === 'user' && result.userInfo) {
+        principals.push({
+          type: 'user',
+          id: result.userInfo._id.toString(),
+          name: result.userInfo.name || result.userInfo.username,
+          email: result.userInfo.email,
+          avatar: result.userInfo.avatar,
+          source: !result.userInfo._id ? 'entra' : 'local',
+          idOnTheSource: result.userInfo.idOnTheSource || result.userInfo._id.toString(),
+          accessRoleId: result.accessRoleId,
+        });
+      } else if (result.principalType === 'group' && result.groupInfo) {
+        principals.push({
+          type: 'group',
+          id: result.groupInfo._id.toString(),
+          name: result.groupInfo.name,
+          email: result.groupInfo.email,
+          description: result.groupInfo.description,
+          avatar: result.groupInfo.avatar,
+          source: result.groupInfo.source || 'local',
+          idOnTheSource: result.groupInfo.idOnTheSource || result.groupInfo._id.toString(),
+          accessRoleId: result.accessRoleId,
+        });
+      }
+    }
+
+    // Return response in format expected by frontend
+    const response = {
+      resourceType,
+      resourceId,
+      principals,
+      public: publicPermission?.public || false,
+      ...(publicPermission?.publicAccessRoleId && {
+        publicAccessRoleId: publicPermission.publicAccessRoleId,
+      }),
+    };
+
+    res.status(200).json(response);
+  } catch (error) {
+    logger.error('Error getting resource permissions principals:', error);
+    res.status(500).json({
+      error: 'Failed to get permissions principals',
+      details: error.message,
+    });
+  }
+};
+
+/**
+ * Get available roles for a resource type
+ * @route GET /api/{resourceType}/roles
+ */
+const getResourceRoles = async (req, res) => {
+  try {
+    const { resourceType } = req.params;
+
+    const roles = await getAvailableRoles({ resourceType });
+
+    res.status(200).json(
+      roles.map((role) => ({
+        accessRoleId: role.accessRoleId,
+        name: role.name,
+        description: role.description,
+        permBits: role.permBits,
+      })),
+    );
+  } catch (error) {
+    logger.error('Error getting resource roles:', error);
+    res.status(500).json({
+      error: 'Failed to get roles',
+      details: error.message,
+    });
+  }
+};
+
+/**
+ * Get user's effective permission bitmask for a resource
+ * @route GET /api/{resourceType}/{resourceId}/effective
+ */
+const getUserEffectivePermissions = async (req, res) => {
+  try {
+    const { resourceType, resourceId } = req.params;
+    const { id: userId } = req.user;
+
+    const permissionBits = await getEffectivePermissions({
+      userId,
+      resourceType,
+      resourceId,
+    });
+
+    res.status(200).json({
+      permissionBits,
+    });
+  } catch (error) {
+    logger.error('Error getting user effective permissions:', error);
+    res.status(500).json({
+      error: 'Failed to get effective permissions',
+      details: error.message,
+    });
+  }
+};
+
+/**
+ * Search for users and groups to grant permissions
+ * Supports hybrid local database + Entra ID search when configured
+ * @route GET /api/permissions/search-principals
+ */
+const searchPrincipals = async (req, res) => {
+  try {
+    const { q: query, limit = 20, type } = req.query;
+
+    if (!query || query.trim().length === 0) {
+      return res.status(400).json({
+        error: 'Query parameter "q" is required and must not be empty',
+      });
+    }
+
+    if (query.trim().length < 2) {
+      return res.status(400).json({
+        error: 'Query must be at least 2 characters long',
+      });
+    }
+
+    const searchLimit = Math.min(Math.max(1, parseInt(limit) || 10), 50);
+    const typeFilter = ['user', 'group'].includes(type) ? type : null;
+
+    const localResults = await searchLocalPrincipals(query.trim(), searchLimit, typeFilter);
+    let allPrincipals = [...localResults];
+
+    const useEntraId = entraIdPrincipalFeatureEnabled(req.user);
+
+    if (useEntraId && localResults.length < searchLimit) {
+      try {
+        const graphTypeMap = {
+          user: 'users',
+          group: 'groups',
+          null: 'all',
+        };
+
+        const authHeader = req.headers.authorization;
+        const accessToken =
+          authHeader && authHeader.startsWith('Bearer ') ? authHeader.substring(7) : null;
+
+        if (accessToken) {
+          const graphResults = await searchEntraIdPrincipals(
+            accessToken,
+            req.user.openidId,
+            query.trim(),
+            graphTypeMap[typeFilter],
+            searchLimit - localResults.length,
+          );
+
+          const localEmails = new Set(
+            localResults.map((p) => p.email?.toLowerCase()).filter(Boolean),
+          );
+          const localGroupSourceIds = new Set(
+            localResults.map((p) => p.idOnTheSource).filter(Boolean),
+          );
+
+          for (const principal of graphResults) {
+            const isDuplicateByEmail =
+              principal.email && localEmails.has(principal.email.toLowerCase());
+            const isDuplicateBySourceId =
+              principal.idOnTheSource && localGroupSourceIds.has(principal.idOnTheSource);
+
+            if (!isDuplicateByEmail && !isDuplicateBySourceId) {
+              allPrincipals.push(principal);
+            }
+          }
+        }
+      } catch (graphError) {
+        logger.warn('Graph API search failed, falling back to local results:', graphError.message);
+      }
+    }
+    const scoredResults = allPrincipals.map((item) => ({
+      ...item,
+      _searchScore: calculateRelevanceScore(item, query.trim()),
+    }));
+
+    allPrincipals = sortPrincipalsByRelevance(scoredResults)
+      .slice(0, searchLimit)
+      .map((result) => {
+        const { _searchScore, ...resultWithoutScore } = result;
+        return resultWithoutScore;
+      });
+    res.status(200).json({
+      query: query.trim(),
+      limit: searchLimit,
+      type: typeFilter,
+      results: allPrincipals,
+      count: allPrincipals.length,
+      sources: {
+        local: allPrincipals.filter((r) => r.source === 'local').length,
+        entra: allPrincipals.filter((r) => r.source === 'entra').length,
+      },
+    });
+  } catch (error) {
+    logger.error('Error searching principals:', error);
+    res.status(500).json({
+      error: 'Failed to search principals',
+      details: error.message,
+    });
+  }
+};
+
+module.exports = {
+  updateResourcePermissions,
+  getResourcePermissions,
+  getResourceRoles,
+  getUserEffectivePermissions,
+  searchPrincipals,
+};

api/server/controllers/agents/client.js

@@ -4,13 +4,11 @@ const {
   sendEvent,
   createRun,
   Tokenizer,
-  checkAccess,
   memoryInstructions,
   createMemoryProcessor,
 } = require('@librechat/api');
 const {
   Callback,
-  Providers,
   GraphEvents,
   formatMessage,
   formatAgentMessages,
@@ -33,29 +31,22 @@ const {
 } = require('librechat-data-provider');
 const { DynamicStructuredTool } = require('@langchain/core/tools');
 const { getBufferString, HumanMessage } = require('@langchain/core/messages');
-const { createGetMCPAuthMap, checkCapability } = require('~/server/services/Config');
+const {
+  getCustomEndpointConfig,
+  createGetMCPAuthMap,
+  checkCapability,
+} = require('~/server/services/Config');
 const { addCacheControl, createContextHandlers } = require('~/app/clients/prompts');
 const { initializeAgent } = require('~/server/services/Endpoints/agents/agent');
 const { spendTokens, spendStructuredTokens } = require('~/models/spendTokens');
 const { getFormattedMemories, deleteMemory, setMemory } = require('~/models');
 const { encodeAndFormat } = require('~/server/services/Files/images/encode');
-const { getProviderConfig } = require('~/server/services/Endpoints');
+const initOpenAI = require('~/server/services/Endpoints/openAI/initialize');
+const { checkAccess } = require('~/server/middleware/roles/access');
 const BaseClient = require('~/app/clients/BaseClient');
-const { getRoleByName } = require('~/models/Role');
 const { loadAgent } = require('~/models/Agent');
 const { getMCPManager } = require('~/config');
 
-const omitTitleOptions = new Set([
-  'stream',
-  'thinking',
-  'streaming',
-  'clientOptions',
-  'thinkingConfig',
-  'thinkingBudget',
-  'includeThoughts',
-  'maxOutputTokens',
-]);
-
 /**
  * @param {ServerRequest} req
  * @param {Agent} agent
@@ -402,12 +393,7 @@ class AgentClient extends BaseClient {
     if (user.personalization?.memories === false) {
       return;
     }
-    const hasAccess = await checkAccess({
-      user,
-      permissionType: PermissionTypes.MEMORIES,
-      permissions: [Permissions.USE],
-      getRoleByName,
-    });
+    const hasAccess = await checkAccess(user, PermissionTypes.MEMORIES, [Permissions.USE]);
 
     if (!hasAccess) {
       logger.debug(
@@ -691,7 +677,7 @@ class AgentClient extends BaseClient {
           hide_sequential_outputs: this.options.agent.hide_sequential_outputs,
           user: this.options.req.user,
         },
-        recursionLimit: agentsEConfig?.recursionLimit ?? 25,
+        recursionLimit: agentsEConfig?.recursionLimit,
         signal: abortController.signal,
         streamMode: 'values',
         version: 'v2',
@@ -997,26 +983,23 @@ class AgentClient extends BaseClient {
       throw new Error('Run not initialized');
     }
     const { handleLLMEnd, collected: collectedMetadata } = createMetadataAggregator();
-    const { req, res, agent } = this.options;
-    const endpoint = agent.endpoint;
-
+    const endpoint = this.options.agent.endpoint;
+    const { req, res } = this.options;
     /** @type {import('@librechat/agents').ClientOptions} */
     let clientOptions = {
       maxTokens: 75,
-      model: agent.model_parameters.model,
     };
-
-    const { getOptions, overrideProvider, customEndpointConfig } =
-      await getProviderConfig(endpoint);
-
-    /** @type {TEndpoint | undefined} */
-    const endpointConfig = req.app.locals[endpoint] ?? customEndpointConfig;
+    let endpointConfig = req.app.locals[endpoint];
     if (!endpointConfig) {
-      logger.warn(
-        '[api/server/controllers/agents/client.js #titleConvo] Error getting endpoint config',
-      );
+      try {
+        endpointConfig = await getCustomEndpointConfig(endpoint);
+      } catch (err) {
+        logger.error(
+          '[api/server/controllers/agents/client.js #titleConvo] Error getting custom endpoint config',
+          err,
+        );
+      }
     }
-
     if (
       endpointConfig &&
       endpointConfig.titleModel &&
@@ -1024,50 +1007,30 @@ class AgentClient extends BaseClient {
     ) {
       clientOptions.model = endpointConfig.titleModel;
     }
-
-    const options = await getOptions({
-      req,
-      res,
-      optionsOnly: true,
-      overrideEndpoint: endpoint,
-      overrideModel: clientOptions.model,
-      endpointOption: { model_parameters: clientOptions },
-    });
-
-    let provider = options.provider ?? overrideProvider ?? agent.provider;
     if (
       endpoint === EModelEndpoint.azureOpenAI &&
-      options.llmConfig?.azureOpenAIApiInstanceName == null
+      clientOptions.model &&
+      this.options.agent.model_parameters.model !== clientOptions.model
     ) {
-      provider = Providers.OPENAI;
+      clientOptions =
+        (
+          await initOpenAI({
+            req,
+            res,
+            optionsOnly: true,
+            overrideModel: clientOptions.model,
+            overrideEndpoint: endpoint,
+            endpointOption: {
+              model_parameters: clientOptions,
+            },
+          })
+        )?.llmConfig ?? clientOptions;
     }
-
-    /** @type {import('@librechat/agents').ClientOptions} */
-    clientOptions = { ...options.llmConfig };
-    if (options.configOptions) {
-      clientOptions.configuration = options.configOptions;
-    }
-
-    // Ensure maxTokens is set for non-o1 models
-    if (!/\b(o\d)\b/i.test(clientOptions.model) && !clientOptions.maxTokens) {
-      clientOptions.maxTokens = 75;
-    } else if (/\b(o\d)\b/i.test(clientOptions.model) && clientOptions.maxTokens != null) {
+    if (/\b(o\d)\b/i.test(clientOptions.model) && clientOptions.maxTokens != null) {
       delete clientOptions.maxTokens;
     }
-
-    clientOptions = Object.assign(
-      Object.fromEntries(
-        Object.entries(clientOptions).filter(([key]) => !omitTitleOptions.has(key)),
-      ),
-    );
-
-    if (provider === Providers.GOOGLE) {
-      clientOptions.json = true;
-    }
-
     try {
       const titleResult = await this.run.generateTitle({
-        provider,
         inputText: text,
         contentParts: this.contentParts,
         clientOptions,
@@ -1085,10 +1048,8 @@ class AgentClient extends BaseClient {
         let input_tokens, output_tokens;
 
         if (item.usage) {
-          input_tokens =
-            item.usage.prompt_tokens || item.usage.input_tokens || item.usage.inputTokens;
-          output_tokens =
-            item.usage.completion_tokens || item.usage.output_tokens || item.usage.outputTokens;
+          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;

api/server/controllers/agents/errors.js

@@ -1,10 +1,10 @@
 // errorHandler.js
-const { logger } = require('@librechat/data-schemas');
+const { logger } = require('~/config');
+const getLogStores = require('~/cache/getLogStores');
 const { CacheKeys, ViolationTypes } = require('librechat-data-provider');
-const { sendResponse } = require('~/server/middleware/error');
 const { recordUsage } = require('~/server/services/Threads');
 const { getConvo } = require('~/models/Conversation');
-const getLogStores = require('~/cache/getLogStores');
+const { sendResponse } = require('~/server/utils');
 
 /**
  * @typedef {Object} ErrorHandlerContext
@@ -75,7 +75,7 @@ const createErrorHandler = ({ req, res, getContext, originPath = '/assistants/ch
     } 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."
+          ? ' 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);

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

@@ -1,5 +1,3 @@
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const { Constants } = require('librechat-data-provider');
 const {
   handleAbortError,
@@ -7,7 +5,9 @@ const {
   cleanupAbortController,
 } = require('~/server/middleware');
 const { disposeClient, clientRegistry, requestDataMap } = require('~/server/cleanup');
+const { sendMessage } = require('~/server/utils');
 const { saveMessage } = require('~/models');
+const { logger } = require('~/config');
 
 const AgentController = async (req, res, next, initializeClient, addTitle) => {
   let {
@@ -206,7 +206,7 @@ const AgentController = async (req, res, next, initializeClient, addTitle) => {
       // Create a new response object with minimal copies
       const finalResponse = { ...response };
 
-      sendEvent(res, {
+      sendMessage(res, {
         final: true,
         conversation,
         title: conversation.title,

api/server/controllers/agents/v1.js

@@ -1,11 +1,10 @@
 const fs = require('fs').promises;
 const { nanoid } = require('nanoid');
-const { logger } = require('@librechat/data-schemas');
+const { logger, PermissionBits } = require('@librechat/data-schemas');
 const {
   Tools,
-  Constants,
-  FileSources,
   SystemRoles,
+  FileSources,
   EToolResources,
   actionDelimiter,
 } = require('librechat-data-provider');
@@ -14,16 +13,20 @@ const {
   createAgent,
   updateAgent,
   deleteAgent,
-  getListAgents,
+  getListAgentsByAccess,
 } = require('~/models/Agent');
+const {
+  grantPermission,
+  findAccessibleResources,
+  findPubliclyAccessibleResources,
+  hasPublicPermission,
+} = require('~/server/services/PermissionService');
 const { getStrategyFunctions } = require('~/server/services/Files/strategies');
 const { resizeAvatar } = require('~/server/services/Files/images/avatar');
 const { refreshS3Url } = require('~/server/services/Files/S3/crud');
 const { filterFile } = require('~/server/services/Files/process');
 const { updateAction, getActions } = require('~/models/Action');
 const { getCachedTools } = require('~/server/services/Config');
-const { updateAgentProjects } = require('~/models/Agent');
-const { getProjectByName } = require('~/models/Project');
 const { revertAgentVersion } = require('~/models/Agent');
 const { deleteFileByFilter } = require('~/models/File');
 
@@ -69,6 +72,27 @@ const createAgentHandler = async (req, res) => {
 
     agentData.id = `agent_${nanoid()}`;
     const agent = await createAgent(agentData);
+
+    // Automatically grant owner permissions to the creator
+    try {
+      await grantPermission({
+        principalType: 'user',
+        principalId: userId,
+        resourceType: 'agent',
+        resourceId: agent._id,
+        accessRoleId: 'agent_owner',
+        grantedBy: userId,
+      });
+      logger.debug(
+        `[createAgent] Granted owner permissions to user ${userId} for agent ${agent.id}`,
+      );
+    } catch (permissionError) {
+      logger.error(
+        `[createAgent] Failed to grant owner permissions for agent ${agent.id}:`,
+        permissionError,
+      );
+    }
+
     res.status(201).json(agent);
   } catch (error) {
     logger.error('[/Agents] Error creating agent', error);
@@ -87,21 +111,14 @@ const createAgentHandler = async (req, res) => {
  * @returns {Promise<Agent>} 200 - success response - application/json
  * @returns {Error} 404 - Agent not found
  */
-const getAgentHandler = async (req, res) => {
+const getAgentHandler = async (req, res, expandProperties = false) => {
   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);
+    // Permissions are validated by middleware before calling this function
+    // Simply load the agent by ID
+    const agent = await getAgent({ id });
 
     if (!agent) {
       return res.status(404).json({ error: 'Agent not found' });
@@ -118,23 +135,45 @@ const getAgentHandler = async (req, res) => {
     }
 
     agent.author = agent.author.toString();
+
+    // @deprecated - isCollaborative replaced by ACL permissions
     agent.isCollaborative = !!agent.isCollaborative;
 
+    // Check if agent is public
+    const isPublic = await hasPublicPermission({
+      resourceType: 'agent',
+      resourceId: agent._id,
+      requiredPermissions: PermissionBits.VIEW,
+    });
+    agent.isPublic = isPublic;
+
     if (agent.author !== author) {
       delete agent.author;
     }
 
-    if (!agent.isCollaborative && agent.author !== author && req.user.role !== SystemRoles.ADMIN) {
+    if (!expandProperties) {
+      // VIEW permission: Basic agent info only
       return res.status(200).json({
+        _id: agent._id,
         id: agent.id,
         name: agent.name,
+        description: agent.description,
         avatar: agent.avatar,
         author: agent.author,
+        provider: agent.provider,
+        model: agent.model,
         projectIds: agent.projectIds,
+        // @deprecated - isCollaborative replaced by ACL permissions
         isCollaborative: agent.isCollaborative,
+        isPublic: agent.isPublic,
         version: agent.version,
+        // Safe metadata
+        createdAt: agent.createdAt,
+        updatedAt: agent.updatedAt,
       });
     }
+
+    // EDIT permission: Full agent details including sensitive configuration
     return res.status(200).json(agent);
   } catch (error) {
     logger.error('[/Agents/:id] Error retrieving agent', error);
@@ -154,42 +193,20 @@ const getAgentHandler = async (req, res) => {
 const updateAgentHandler = async (req, res) => {
   try {
     const id = req.params.id;
-    const { projectIds, removeProjectIds, ...updateData } = req.body;
-    const isAdmin = req.user.role === SystemRoles.ADMIN;
+    const { _id, ...updateData } = req.body;
     const existingAgent = await getAgent({ id });
-    const isAuthor = existingAgent.author.toString() === req.user.id;
 
     if (!existingAgent) {
       return res.status(404).json({ error: 'Agent not found' });
     }
-    const hasEditPermission = existingAgent.isCollaborative || isAdmin || isAuthor;
-
-    if (!hasEditPermission) {
-      return res.status(403).json({
-        error: 'You do not have permission to modify this non-collaborative agent',
-      });
-    }
-
-    /** @type {boolean} */
-    const isProjectUpdate = (projectIds?.length ?? 0) > 0 || (removeProjectIds?.length ?? 0) > 0;
 
     let updatedAgent =
       Object.keys(updateData).length > 0
         ? await updateAgent({ id }, updateData, {
             updatingUserId: req.user.id,
-            skipVersioning: isProjectUpdate,
           })
         : existingAgent;
 
-    if (isProjectUpdate) {
-      updatedAgent = await updateAgentProjects({
-        user: req.user,
-        agentId: id,
-        projectIds,
-        removeProjectIds,
-      });
-    }
-
     if (updatedAgent.author) {
       updatedAgent.author = updatedAgent.author.toString();
     }
@@ -307,6 +324,26 @@ const duplicateAgentHandler = async (req, res) => {
     newAgentData.actions = agentActions;
     const newAgent = await createAgent(newAgentData);
 
+    // Automatically grant owner permissions to the duplicator
+    try {
+      await grantPermission({
+        principalType: 'user',
+        principalId: userId,
+        resourceType: 'agent',
+        resourceId: newAgent._id,
+        accessRoleId: 'agent_owner',
+        grantedBy: userId,
+      });
+      logger.debug(
+        `[duplicateAgent] Granted owner permissions to user ${userId} for duplicated agent ${newAgent.id}`,
+      );
+    } catch (permissionError) {
+      logger.error(
+        `[duplicateAgent] Failed to grant owner permissions for duplicated agent ${newAgent.id}:`,
+        permissionError,
+      );
+    }
+
     return res.status(201).json({
       agent: newAgent,
       actions: newActionsList,
@@ -333,7 +370,7 @@ const deleteAgentHandler = async (req, res) => {
     if (!agent) {
       return res.status(404).json({ error: 'Agent not found' });
     }
-    await deleteAgent({ id, author: req.user.id });
+    await deleteAgent({ id });
     return res.json({ message: 'Agent deleted' });
   } catch (error) {
     logger.error('[/Agents/:id] Error deleting Agent', error);
@@ -342,7 +379,7 @@ const deleteAgentHandler = async (req, res) => {
 };
 
 /**
- *
+ * Lists agents using ACL-aware permissions (ownership + explicit shares).
  * @route GET /Agents
  * @param {object} req - Express Request
  * @param {object} req.query - Request query
@@ -351,9 +388,31 @@ const deleteAgentHandler = async (req, res) => {
  */
 const getListAgentsHandler = async (req, res) => {
   try {
-    const data = await getListAgents({
-      author: req.user.id,
+    const userId = req.user.id;
+
+    // Get agent IDs the user has VIEW access to via ACL
+    const accessibleIds = await findAccessibleResources({
+      userId,
+      resourceType: 'agent',
+      requiredPermissions: PermissionBits.VIEW,
     });
+    const publiclyAccessibleIds = await findPubliclyAccessibleResources({
+      resourceType: 'agent',
+      requiredPermissions: PermissionBits.VIEW,
+    });
+    // Use the new ACL-aware function
+    const data = await getListAgentsByAccess({
+      accessibleIds,
+      otherParams: {}, // Can add query params here if needed
+    });
+    if (data?.data?.length) {
+      data.data = data.data.map((agent) => {
+        if (publiclyAccessibleIds.some((id) => id.equals(agent._id))) {
+          agent.isPublic = true;
+        }
+        return agent;
+      });
+    }
     return res.json(data);
   } catch (error) {
     logger.error('[/Agents] Error listing Agents', error);
@@ -431,7 +490,7 @@ const uploadAgentAvatarHandler = async (req, res) => {
     };
 
     promises.push(
-      await updateAgent({ id: agent_id, author: req.user.id }, data, {
+      await updateAgent({ id: agent_id }, data, {
         updatingUserId: req.user.id,
       }),
     );

api/server/controllers/assistants/chatV1.js

@@ -1,7 +1,4 @@
 const { v4 } = require('uuid');
-const { sleep } = require('@librechat/agents');
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const {
   Time,
   Constants,
@@ -22,20 +19,20 @@ const {
   addThreadMetadata,
   saveAssistantMessage,
 } = require('~/server/services/Threads');
+const { sendResponse, sendMessage, sleep, countTokens } = require('~/server/utils');
 const { runAssistant, createOnTextProgress } = require('~/server/services/AssistantService');
 const validateAuthor = require('~/server/middleware/assistants/validateAuthor');
 const { formatMessage, createVisionPrompt } = require('~/app/clients/prompts');
 const { createRun, StreamRunManager } = require('~/server/services/Runs');
 const { addTitle } = require('~/server/services/Endpoints/assistants');
 const { createRunBody } = require('~/server/services/createRunBody');
-const { sendResponse } = require('~/server/middleware/error');
 const { getTransactions } = require('~/models/Transaction');
 const { checkBalance } = require('~/models/balanceMethods');
 const { getConvo } = require('~/models/Conversation');
 const getLogStores = require('~/cache/getLogStores');
-const { countTokens } = require('~/server/utils');
 const { getModelMaxTokens } = require('~/utils');
 const { getOpenAIClient } = require('./helpers');
+const { logger } = require('~/config');
 
 /**
  * @route POST /
@@ -474,7 +471,7 @@ const chatV1 = async (req, res) => {
     await Promise.all(promises);
 
     const sendInitialResponse = () => {
-      sendEvent(res, {
+      sendMessage(res, {
         sync: true,
         conversationId,
         // messages: previousMessages,
@@ -590,7 +587,7 @@ const chatV1 = async (req, res) => {
       iconURL: endpointOption.iconURL,
     };
 
-    sendEvent(res, {
+    sendMessage(res, {
       final: true,
       conversation,
       requestMessage: {

api/server/controllers/assistants/chatV2.js

@@ -1,7 +1,4 @@
 const { v4 } = require('uuid');
-const { sleep } = require('@librechat/agents');
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const {
   Time,
   Constants,
@@ -25,14 +22,15 @@ 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');
+const { sendMessage, sleep, countTokens } = require('~/server/utils');
 const { createRunBody } = require('~/server/services/createRunBody');
 const { getTransactions } = require('~/models/Transaction');
 const { checkBalance } = require('~/models/balanceMethods');
 const { getConvo } = require('~/models/Conversation');
 const getLogStores = require('~/cache/getLogStores');
-const { countTokens } = require('~/server/utils');
 const { getModelMaxTokens } = require('~/utils');
 const { getOpenAIClient } = require('./helpers');
+const { logger } = require('~/config');
 
 /**
  * @route POST /
@@ -311,7 +309,7 @@ const chatV2 = async (req, res) => {
     await Promise.all(promises);
 
     const sendInitialResponse = () => {
-      sendEvent(res, {
+      sendMessage(res, {
         sync: true,
         conversationId,
         // messages: previousMessages,
@@ -434,7 +432,7 @@ const chatV2 = async (req, res) => {
       iconURL: endpointOption.iconURL,
     };
 
-    sendEvent(res, {
+    sendMessage(res, {
       final: true,
       conversation,
       requestMessage: {

api/server/controllers/assistants/errors.js

@@ -1,10 +1,10 @@
 // errorHandler.js
-const { logger } = require('@librechat/data-schemas');
-const { CacheKeys, ViolationTypes, ContentTypes } = require('librechat-data-provider');
-const { recordUsage, checkMessageGaps } = require('~/server/services/Threads');
-const { sendResponse } = require('~/server/middleware/error');
-const { getConvo } = require('~/models/Conversation');
+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
@@ -78,7 +78,7 @@ const createErrorHandler = ({ req, res, getContext, originPath = '/assistants/ch
     } 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."
+          ? ' 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);

api/server/controllers/tools.js

@@ -1,7 +1,5 @@
 const { nanoid } = require('nanoid');
 const { EnvVar } = require('@librechat/agents');
-const { checkAccess } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const {
   Tools,
   AuthType,
@@ -15,8 +13,9 @@ const { processCodeOutput } = require('~/server/services/Files/Code/process');
 const { createToolCall, getToolCallsByConvo } = require('~/models/ToolCall');
 const { loadAuthValues } = require('~/server/services/Tools/credentials');
 const { loadTools } = require('~/app/clients/tools/util');
-const { getRoleByName } = require('~/models/Role');
+const { checkAccess } = require('~/server/middleware');
 const { getMessage } = require('~/models/Message');
+const { logger } = require('~/config');
 
 const fieldsMap = {
   [Tools.execute_code]: [EnvVar.CODE_API_KEY],
@@ -80,7 +79,6 @@ const verifyToolAuth = async (req, res) => {
         throwError: false,
       });
     } catch (error) {
-      logger.error('Error loading auth values', error);
       res.status(200).json({ authenticated: false, message: AuthType.USER_PROVIDED });
       return;
     }
@@ -134,12 +132,7 @@ const callTool = async (req, res) => {
     logger.debug(`[${toolId}/call] User: ${req.user.id}`);
     let hasAccess = true;
     if (toolAccessPermType[toolId]) {
-      hasAccess = await checkAccess({
-        user: req.user,
-        permissionType: toolAccessPermType[toolId],
-        permissions: [Permissions.USE],
-        getRoleByName,
-      });
+      hasAccess = await checkAccess(req.user, toolAccessPermType[toolId], [Permissions.USE]);
     }
     if (!hasAccess) {
       logger.warn(

api/server/index.js

@@ -39,9 +39,7 @@ const startServer = async () => {
   await connectDb();
 
   logger.info('Connected to MongoDB');
-  indexSync().catch((err) => {
-    logger.error('[indexSync] Background sync failed:', err);
-  });
+  await indexSync();
 
   app.disable('x-powered-by');
   app.set('trust proxy', trusted_proxy);
@@ -97,6 +95,7 @@ const startServer = async () => {
   app.use('/api/actions', routes.actions);
   app.use('/api/keys', routes.keys);
   app.use('/api/user', routes.user);
+  app.use('/api/ask', routes.ask);
   app.use('/api/search', routes.search);
   app.use('/api/edit', routes.edit);
   app.use('/api/messages', routes.messages);
@@ -117,7 +116,10 @@ const startServer = async () => {
   app.use('/api/roles', routes.roles);
   app.use('/api/agents', routes.agents);
   app.use('/api/banner', routes.banner);
+  app.use('/api/bedrock', routes.bedrock);
   app.use('/api/memories', routes.memories);
+  app.use('/api/permissions', routes.accessPermissions);
+
   app.use('/api/tags', routes.tags);
   app.use('/api/mcp', routes.mcp);
 

api/server/middleware/abortMiddleware.js

@@ -1,13 +1,13 @@
-const { logger } = require('@librechat/data-schemas');
-const { countTokens, isEnabled, sendEvent } = require('@librechat/api');
+// abortMiddleware.js
 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 { sendError } = require('~/server/middleware/error');
 const { spendTokens } = require('~/models/spendTokens');
 const abortControllers = require('./abortControllers');
 const { saveMessage, getConvo } = require('~/models');
 const { abortRun } = require('./abortRun');
+const { logger } = require('~/config');
 
 const abortDataMap = new WeakMap();
 
@@ -101,7 +101,7 @@ async function abortMessage(req, res) {
   cleanupAbortController(abortKey);
 
   if (res.headersSent && finalEvent) {
-    return sendEvent(res, finalEvent);
+    return sendMessage(res, finalEvent);
   }
 
   res.setHeader('Content-Type', 'application/json');
@@ -174,7 +174,7 @@ const createAbortController = (req, res, getAbortData, getReqData) => {
    * @param {string} responseMessageId
    */
   const onStart = (userMessage, responseMessageId) => {
-    sendEvent(res, { message: userMessage, created: true });
+    sendMessage(res, { message: userMessage, created: true });
 
     const abortKey = userMessage?.conversationId ?? req.user.id;
     getReqData({ abortKey });

api/server/middleware/abortRun.js

@@ -1,11 +1,11 @@
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const { CacheKeys, RunStatus, isUUID } = require('librechat-data-provider');
 const { initializeClient } = require('~/server/services/Endpoints/assistants');
 const { checkMessageGaps, recordUsage } = require('~/server/services/Threads');
 const { deleteMessages } = require('~/models/Message');
 const { getConvo } = require('~/models/Conversation');
 const getLogStores = require('~/cache/getLogStores');
+const { sendMessage } = require('~/server/utils');
+const { logger } = require('~/config');
 
 const three_minutes = 1000 * 60 * 3;
 
@@ -34,7 +34,7 @@ async function abortRun(req, res) {
   const [thread_id, run_id] = runValues.split(':');
 
   if (!run_id) {
-    logger.warn("[abortRun] Couldn't find run for cancel request", { thread_id });
+    logger.warn('[abortRun] Couldn\'t find run for cancel request', { thread_id });
     return res.status(204).send({ message: 'Run not found' });
   } else if (run_id === 'cancelled') {
     logger.warn('[abortRun] Run already cancelled', { thread_id });
@@ -93,7 +93,7 @@ async function abortRun(req, res) {
   };
 
   if (res.headersSent && finalEvent) {
-    return sendEvent(res, finalEvent);
+    return sendMessage(res, finalEvent);
   }
 
   res.json(finalEvent);

api/server/middleware/accessResources/canAccessAgentFromBody.js

@@ -0,0 +1,97 @@
+const { logger } = require('@librechat/data-schemas');
+const { Constants, isAgentsEndpoint } = require('librechat-data-provider');
+const { canAccessResource } = require('./canAccessResource');
+const { getAgent } = require('~/models/Agent');
+
+/**
+ * Agent ID resolver function for agent_id from request body
+ * Resolves custom agent ID (e.g., "agent_abc123") to MongoDB ObjectId
+ * This is used specifically for chat routes where agent_id comes from request body
+ *
+ * @param {string} agentCustomId - Custom agent ID from request body
+ * @returns {Promise<Object|null>} Agent document with _id field, or null if not found
+ */
+const resolveAgentIdFromBody = async (agentCustomId) => {
+  // Handle ephemeral agents - they don't need permission checks
+  if (agentCustomId === Constants.EPHEMERAL_AGENT_ID) {
+    return null; // No permission check needed for ephemeral agents
+  }
+
+  return await getAgent({ id: agentCustomId });
+};
+
+/**
+ * Middleware factory that creates middleware to check agent access permissions from request body.
+ * This middleware is specifically designed for chat routes where the agent_id comes from req.body
+ * instead of route parameters.
+ *
+ * @param {Object} options - Configuration options
+ * @param {number} options.requiredPermission - The permission bit required (1=view, 2=edit, 4=delete, 8=share)
+ * @returns {Function} Express middleware function
+ *
+ * @example
+ * // Basic usage for agent chat (requires VIEW permission)
+ * router.post('/chat',
+ *   canAccessAgentFromBody({ requiredPermission: PermissionBits.VIEW }),
+ *   buildEndpointOption,
+ *   chatController
+ * );
+ */
+const canAccessAgentFromBody = (options) => {
+  const { requiredPermission } = options;
+
+  // Validate required options
+  if (!requiredPermission || typeof requiredPermission !== 'number') {
+    throw new Error('canAccessAgentFromBody: requiredPermission is required and must be a number');
+  }
+
+  return async (req, res, next) => {
+    try {
+      const { endpoint, agent_id } = req.body;
+      let agentId = agent_id;
+
+      if (!isAgentsEndpoint(endpoint)) {
+        agentId = Constants.EPHEMERAL_AGENT_ID;
+      }
+
+      if (!agentId) {
+        return res.status(400).json({
+          error: 'Bad Request',
+          message: 'agent_id is required in request body',
+        });
+      }
+
+      // Skip permission checks for ephemeral agents
+      if (agentId === Constants.EPHEMERAL_AGENT_ID) {
+        return next();
+      }
+
+      const agentAccessMiddleware = canAccessResource({
+        resourceType: 'agent',
+        requiredPermission,
+        resourceIdParam: 'agent_id', // This will be ignored since we use custom resolver
+        idResolver: () => resolveAgentIdFromBody(agentId),
+      });
+
+      const tempReq = {
+        ...req,
+        params: {
+          ...req.params,
+          agent_id: agentId,
+        },
+      };
+
+      return agentAccessMiddleware(tempReq, res, next);
+    } catch (error) {
+      logger.error('Failed to validate agent access permissions', error);
+      return res.status(500).json({
+        error: 'Internal Server Error',
+        message: 'Failed to validate agent access permissions',
+      });
+    }
+  };
+};
+
+module.exports = {
+  canAccessAgentFromBody,
+};

api/server/middleware/accessResources/canAccessAgentResource.js

@@ -0,0 +1,58 @@
+const { getAgent } = require('~/models/Agent');
+const { canAccessResource } = require('./canAccessResource');
+
+/**
+ * Agent ID resolver function
+ * Resolves custom agent ID (e.g., "agent_abc123") to MongoDB ObjectId
+ *
+ * @param {string} agentCustomId - Custom agent ID from route parameter
+ * @returns {Promise<Object|null>} Agent document with _id field, or null if not found
+ */
+const resolveAgentId = async (agentCustomId) => {
+  return await getAgent({ id: agentCustomId });
+};
+
+/**
+ * Agent-specific middleware factory that creates middleware to check agent access permissions.
+ * This middleware extends the generic canAccessResource to handle agent custom ID resolution.
+ *
+ * @param {Object} options - Configuration options
+ * @param {number} options.requiredPermission - The permission bit required (1=view, 2=edit, 4=delete, 8=share)
+ * @param {string} [options.resourceIdParam='id'] - The name of the route parameter containing the agent custom ID
+ * @returns {Function} Express middleware function
+ *
+ * @example
+ * // Basic usage for viewing agents
+ * router.get('/agents/:id',
+ *   canAccessAgentResource({ requiredPermission: 1 }),
+ *   getAgent
+ * );
+ *
+ * @example
+ * // Custom resource ID parameter and edit permission
+ * router.patch('/agents/:agent_id',
+ *   canAccessAgentResource({
+ *     requiredPermission: 2,
+ *     resourceIdParam: 'agent_id'
+ *   }),
+ *   updateAgent
+ * );
+ */
+const canAccessAgentResource = (options) => {
+  const { requiredPermission, resourceIdParam = 'id' } = options;
+
+  if (!requiredPermission || typeof requiredPermission !== 'number') {
+    throw new Error('canAccessAgentResource: requiredPermission is required and must be a number');
+  }
+
+  return canAccessResource({
+    resourceType: 'agent',
+    requiredPermission,
+    resourceIdParam,
+    idResolver: resolveAgentId,
+  });
+};
+
+module.exports = {
+  canAccessAgentResource,
+};

api/server/middleware/accessResources/canAccessAgentResource.spec.js

@@ -0,0 +1,384 @@
+const mongoose = require('mongoose');
+const { MongoMemoryServer } = require('mongodb-memory-server');
+const { canAccessAgentResource } = require('./canAccessAgentResource');
+const { User, Role, AclEntry } = require('~/db/models');
+const { createAgent } = require('~/models/Agent');
+
+describe('canAccessAgentResource middleware', () => {
+  let mongoServer;
+  let req, res, next;
+  let testUser;
+
+  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();
+    await Role.create({
+      name: 'test-role',
+      permissions: {
+        AGENTS: {
+          USE: true,
+          CREATE: true,
+          SHARED_GLOBAL: false,
+        },
+      },
+    });
+
+    // Create a test user
+    testUser = await User.create({
+      email: 'test@example.com',
+      name: 'Test User',
+      username: 'testuser',
+      role: 'test-role',
+    });
+
+    req = {
+      user: { id: testUser._id.toString(), role: 'test-role' },
+      params: {},
+    };
+    res = {
+      status: jest.fn().mockReturnThis(),
+      json: jest.fn(),
+    };
+    next = jest.fn();
+
+    jest.clearAllMocks();
+  });
+
+  describe('middleware factory', () => {
+    test('should throw error if requiredPermission is not provided', () => {
+      expect(() => canAccessAgentResource({})).toThrow(
+        'canAccessAgentResource: requiredPermission is required and must be a number',
+      );
+    });
+
+    test('should throw error if requiredPermission is not a number', () => {
+      expect(() => canAccessAgentResource({ requiredPermission: '1' })).toThrow(
+        'canAccessAgentResource: requiredPermission is required and must be a number',
+      );
+    });
+
+    test('should create middleware with default resourceIdParam', () => {
+      const middleware = canAccessAgentResource({ requiredPermission: 1 });
+      expect(typeof middleware).toBe('function');
+      expect(middleware.length).toBe(3); // Express middleware signature
+    });
+
+    test('should create middleware with custom resourceIdParam', () => {
+      const middleware = canAccessAgentResource({
+        requiredPermission: 2,
+        resourceIdParam: 'agent_id',
+      });
+      expect(typeof middleware).toBe('function');
+      expect(middleware.length).toBe(3);
+    });
+  });
+
+  describe('permission checking with real agents', () => {
+    test('should allow access when user is the agent author', async () => {
+      // Create an agent owned by the test user
+      const agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Test Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: testUser._id,
+      });
+
+      // Create ACL entry for the author (owner permissions)
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 15, // All permissions (1+2+4+8)
+        grantedBy: testUser._id,
+      });
+
+      req.params.id = agent.id;
+
+      const middleware = canAccessAgentResource({ requiredPermission: 1 }); // VIEW permission
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+      expect(res.status).not.toHaveBeenCalled();
+    });
+
+    test('should deny access when user is not the author and has no ACL entry', async () => {
+      // Create an agent owned by a different user
+      const otherUser = await User.create({
+        email: 'other@example.com',
+        name: 'Other User',
+        username: 'otheruser',
+        role: 'test-role',
+      });
+
+      const agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Other User Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: otherUser._id,
+      });
+
+      // Create ACL entry for the other user (owner)
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: otherUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 15, // All permissions
+        grantedBy: otherUser._id,
+      });
+
+      req.params.id = agent.id;
+
+      const middleware = canAccessAgentResource({ requiredPermission: 1 }); // VIEW permission
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(403);
+      expect(res.json).toHaveBeenCalledWith({
+        error: 'Forbidden',
+        message: 'Insufficient permissions to access this agent',
+      });
+    });
+
+    test('should allow access when user has ACL entry with sufficient permissions', async () => {
+      // Create an agent owned by a different user
+      const otherUser = await User.create({
+        email: 'other2@example.com',
+        name: 'Other User 2',
+        username: 'otheruser2',
+        role: 'test-role',
+      });
+
+      const agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Shared Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: otherUser._id,
+      });
+
+      // Create ACL entry granting view permission to test user
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 1, // VIEW permission
+        grantedBy: otherUser._id,
+      });
+
+      req.params.id = agent.id;
+
+      const middleware = canAccessAgentResource({ requiredPermission: 1 }); // VIEW permission
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+      expect(res.status).not.toHaveBeenCalled();
+    });
+
+    test('should deny access when ACL permissions are insufficient', async () => {
+      // Create an agent owned by a different user
+      const otherUser = await User.create({
+        email: 'other3@example.com',
+        name: 'Other User 3',
+        username: 'otheruser3',
+        role: 'test-role',
+      });
+
+      const agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Limited Access Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: otherUser._id,
+      });
+
+      // Create ACL entry granting only view permission
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 1, // VIEW permission only
+        grantedBy: otherUser._id,
+      });
+
+      req.params.id = agent.id;
+
+      const middleware = canAccessAgentResource({ requiredPermission: 2 }); // EDIT permission required
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(403);
+      expect(res.json).toHaveBeenCalledWith({
+        error: 'Forbidden',
+        message: 'Insufficient permissions to access this agent',
+      });
+    });
+
+    test('should handle non-existent agent', async () => {
+      req.params.id = 'agent_nonexistent';
+
+      const middleware = canAccessAgentResource({ requiredPermission: 1 });
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(404);
+      expect(res.json).toHaveBeenCalledWith({
+        error: 'Not Found',
+        message: 'agent not found',
+      });
+    });
+
+    test('should use custom resourceIdParam', async () => {
+      const agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Custom Param Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: testUser._id,
+      });
+
+      // Create ACL entry for the author
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 15, // All permissions
+        grantedBy: testUser._id,
+      });
+
+      req.params.agent_id = agent.id; // Using custom param name
+
+      const middleware = canAccessAgentResource({
+        requiredPermission: 1,
+        resourceIdParam: 'agent_id',
+      });
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+      expect(res.status).not.toHaveBeenCalled();
+    });
+  });
+
+  describe('permission levels', () => {
+    let agent;
+
+    beforeEach(async () => {
+      agent = await createAgent({
+        id: `agent_${Date.now()}`,
+        name: 'Permission Test Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: testUser._id,
+      });
+
+      // Create ACL entry with all permissions for the owner
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 15, // All permissions (1+2+4+8)
+        grantedBy: testUser._id,
+      });
+
+      req.params.id = agent.id;
+    });
+
+    test('should support view permission (1)', async () => {
+      const middleware = canAccessAgentResource({ requiredPermission: 1 });
+      await middleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+
+    test('should support edit permission (2)', async () => {
+      const middleware = canAccessAgentResource({ requiredPermission: 2 });
+      await middleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+
+    test('should support delete permission (4)', async () => {
+      const middleware = canAccessAgentResource({ requiredPermission: 4 });
+      await middleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+
+    test('should support share permission (8)', async () => {
+      const middleware = canAccessAgentResource({ requiredPermission: 8 });
+      await middleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+
+    test('should support combined permissions', async () => {
+      const viewAndEdit = 1 | 2; // 3
+      const middleware = canAccessAgentResource({ requiredPermission: viewAndEdit });
+      await middleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+  });
+
+  describe('integration with agent operations', () => {
+    test('should work with agent CRUD operations', async () => {
+      const agentId = `agent_${Date.now()}`;
+
+      // Create agent
+      const agent = await createAgent({
+        id: agentId,
+        name: 'Integration Test Agent',
+        provider: 'openai',
+        model: 'gpt-4',
+        author: testUser._id,
+        description: 'Testing integration',
+      });
+
+      // Create ACL entry for the author
+      await AclEntry.create({
+        principalType: 'user',
+        principalId: testUser._id,
+        principalModel: 'User',
+        resourceType: 'agent',
+        resourceId: agent._id,
+        permBits: 15, // All permissions
+        grantedBy: testUser._id,
+      });
+
+      req.params.id = agentId;
+
+      // Test view access
+      const viewMiddleware = canAccessAgentResource({ requiredPermission: 1 });
+      await viewMiddleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+      jest.clearAllMocks();
+
+      // Update the agent
+      const { updateAgent } = require('~/models/Agent');
+      await updateAgent({ id: agentId }, { description: 'Updated description' });
+
+      // Test edit access
+      const editMiddleware = canAccessAgentResource({ requiredPermission: 2 });
+      await editMiddleware(req, res, next);
+      expect(next).toHaveBeenCalled();
+    });
+  });
+});

api/server/middleware/accessResources/canAccessResource.js

@@ -0,0 +1,157 @@
+const { logger } = require('@librechat/data-schemas');
+const { SystemRoles } = require('librechat-data-provider');
+const { checkPermission } = require('~/server/services/PermissionService');
+
+/**
+ * Generic base middleware factory that creates middleware to check resource access permissions.
+ * This middleware expects MongoDB ObjectIds as resource identifiers for ACL permission checks.
+ *
+ * @param {Object} options - Configuration options
+ * @param {string} options.resourceType - The type of resource (e.g., 'agent', 'file', 'project')
+ * @param {number} options.requiredPermission - The permission bit required (1=view, 2=edit, 4=delete, 8=share)
+ * @param {string} [options.resourceIdParam='resourceId'] - The name of the route parameter containing the resource ID
+ * @param {Function} [options.idResolver] - Optional function to resolve custom IDs to ObjectIds
+ * @returns {Function} Express middleware function
+ *
+ * @example
+ * // Direct usage with ObjectId (for resources that use MongoDB ObjectId in routes)
+ * router.get('/prompts/:promptId',
+ *   canAccessResource({ resourceType: 'prompt', requiredPermission: 1 }),
+ *   getPrompt
+ * );
+ *
+ * @example
+ * // Usage with custom ID resolver (for resources that use custom string IDs)
+ * router.get('/agents/:id',
+ *   canAccessResource({
+ *     resourceType: 'agent',
+ *     requiredPermission: 1,
+ *     resourceIdParam: 'id',
+ *     idResolver: (customId) => resolveAgentId(customId)
+ *   }),
+ *   getAgent
+ * );
+ */
+const canAccessResource = (options) => {
+  const {
+    resourceType,
+    requiredPermission,
+    resourceIdParam = 'resourceId',
+    idResolver = null,
+  } = options;
+
+  if (!resourceType || typeof resourceType !== 'string') {
+    throw new Error('canAccessResource: resourceType is required and must be a string');
+  }
+
+  if (!requiredPermission || typeof requiredPermission !== 'number') {
+    throw new Error('canAccessResource: requiredPermission is required and must be a number');
+  }
+
+  return async (req, res, next) => {
+    try {
+      // Extract resource ID from route parameters
+      const rawResourceId = req.params[resourceIdParam];
+
+      if (!rawResourceId) {
+        logger.warn(`[canAccessResource] Missing ${resourceIdParam} in route parameters`);
+        return res.status(400).json({
+          error: 'Bad Request',
+          message: `${resourceIdParam} is required`,
+        });
+      }
+
+      // Check if user is authenticated
+      if (!req.user || !req.user.id) {
+        logger.warn(
+          `[canAccessResource] Unauthenticated request for ${resourceType} ${rawResourceId}`,
+        );
+        return res.status(401).json({
+          error: 'Unauthorized',
+          message: 'Authentication required',
+        });
+      }
+      // if system admin let through
+      if (req.user.role === SystemRoles.ADMIN) {
+        return next();
+      }
+      const userId = req.user.id;
+      let resourceId = rawResourceId;
+      let resourceInfo = null;
+
+      // Resolve custom ID to ObjectId if resolver is provided
+      if (idResolver) {
+        logger.debug(
+          `[canAccessResource] Resolving ${resourceType} custom ID ${rawResourceId} to ObjectId`,
+        );
+
+        const resolutionResult = await idResolver(rawResourceId);
+
+        if (!resolutionResult) {
+          logger.warn(`[canAccessResource] ${resourceType} not found: ${rawResourceId}`);
+          return res.status(404).json({
+            error: 'Not Found',
+            message: `${resourceType} not found`,
+          });
+        }
+
+        // Handle different resolver return formats
+        if (typeof resolutionResult === 'string' || resolutionResult._id) {
+          resourceId = resolutionResult._id || resolutionResult;
+          resourceInfo = typeof resolutionResult === 'object' ? resolutionResult : null;
+        } else {
+          resourceId = resolutionResult;
+        }
+
+        logger.debug(
+          `[canAccessResource] Resolved ${resourceType} ${rawResourceId} to ObjectId ${resourceId}`,
+        );
+      }
+
+      // Check permissions using PermissionService with ObjectId
+      const hasPermission = await checkPermission({
+        userId,
+        resourceType,
+        resourceId,
+        requiredPermission,
+      });
+
+      if (hasPermission) {
+        logger.debug(
+          `[canAccessResource] User ${userId} has permission ${requiredPermission} on ${resourceType} ${rawResourceId} (${resourceId})`,
+        );
+
+        req.resourceAccess = {
+          resourceType,
+          resourceId, // MongoDB ObjectId for ACL operations
+          customResourceId: rawResourceId, // Original ID from route params
+          permission: requiredPermission,
+          userId,
+          ...(resourceInfo && { resourceInfo }),
+        };
+
+        return next();
+      }
+
+      logger.warn(
+        `[canAccessResource] User ${userId} denied access to ${resourceType} ${rawResourceId} ` +
+          `(required permission: ${requiredPermission})`,
+      );
+
+      return res.status(403).json({
+        error: 'Forbidden',
+        message: `Insufficient permissions to access this ${resourceType}`,
+      });
+    } catch (error) {
+      logger.error(`[canAccessResource] Error checking access for ${resourceType}:`, error);
+      return res.status(500).json({
+        error: 'Internal Server Error',
+        message: 'Failed to check resource access permissions',
+      });
+    }
+  };
+};
+
+module.exports = {
+  canAccessResource,
+};

api/server/middleware/accessResources/index.js

@@ -0,0 +1,9 @@
+const { canAccessResource } = require('./canAccessResource');
+const { canAccessAgentResource } = require('./canAccessAgentResource');
+const { canAccessAgentFromBody } = require('./canAccessAgentFromBody');
+
+module.exports = {
+  canAccessResource,
+  canAccessAgentResource,
+  canAccessAgentFromBody,
+};

api/server/middleware/buildEndpointOption.js

@@ -1,12 +1,13 @@
-const { logger } = require('@librechat/data-schemas');
 const {
-  EndpointURLs,
+  parseCompactConvo,
   EModelEndpoint,
   isAgentsEndpoint,
-  parseCompactConvo,
+  EndpointURLs,
 } = require('librechat-data-provider');
 const azureAssistants = require('~/server/services/Endpoints/azureAssistants');
+const { getModelsConfig } = require('~/server/controllers/ModelController');
 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');
@@ -24,6 +25,7 @@ const buildFunction = {
   [EModelEndpoint.bedrock]: bedrock.buildOptions,
   [EModelEndpoint.azureOpenAI]: openAI.buildOptions,
   [EModelEndpoint.anthropic]: anthropic.buildOptions,
+  [EModelEndpoint.gptPlugins]: gptPlugins.buildOptions,
   [EModelEndpoint.assistants]: assistants.buildOptions,
   [EModelEndpoint.azureAssistants]: azureAssistants.buildOptions,
 };
@@ -34,9 +36,6 @@ async function buildEndpointOption(req, res, next) {
   try {
     parsedBody = parseCompactConvo({ endpoint, endpointType, conversation: req.body });
   } catch (error) {
-    logger.warn(
-      `Error parsing conversation for endpoint ${endpoint}${error?.message ? `: ${error.message}` : ''}`,
-    );
     return handleError(res, { text: 'Error parsing conversation' });
   }
 
@@ -58,6 +57,15 @@ async function buildEndpointOption(req, res, next) {
       return handleError(res, { text: 'Model spec mismatch' });
     }
 
+    if (
+      currentModelSpec.preset.endpoint !== EModelEndpoint.gptPlugins &&
+      currentModelSpec.preset.tools
+    ) {
+      return handleError(res, {
+        text: `Only the "${EModelEndpoint.gptPlugins}" endpoint can have tools defined in the preset`,
+      });
+    }
+
     try {
       currentModelSpec.preset.spec = spec;
       if (currentModelSpec.iconURL != null && currentModelSpec.iconURL !== '') {
@@ -69,7 +77,6 @@ async function buildEndpointOption(req, res, next) {
         conversation: currentModelSpec.preset,
       });
     } catch (error) {
-      logger.error(`Error parsing model spec for endpoint ${endpoint}`, error);
       return handleError(res, { text: 'Error parsing model spec' });
     }
   }
@@ -77,23 +84,20 @@ async function buildEndpointOption(req, res, next) {
   try {
     const isAgents =
       isAgentsEndpoint(endpoint) || req.baseUrl.startsWith(EndpointURLs[EModelEndpoint.agents]);
-    const builder = isAgents
-      ? (...args) => buildFunction[EModelEndpoint.agents](req, ...args)
-      : buildFunction[endpointType ?? endpoint];
+    const endpointFn = buildFunction[isAgents ? EModelEndpoint.agents : (endpointType ?? endpoint)];
+    const builder = isAgents ? (...args) => endpointFn(req, ...args) : endpointFn;
 
     // TODO: use object params
     req.body.endpointOption = await builder(endpoint, parsedBody, endpointType);
 
+    // TODO: use `getModelsConfig` only when necessary
+    const modelsConfig = await getModelsConfig(req);
+    req.body.endpointOption.modelsConfig = modelsConfig;
     if (req.body.files && !isAgents) {
       req.body.endpointOption.attachments = processFiles(req.body.files);
     }
-
     next();
   } catch (error) {
-    logger.error(
-      `Error building endpoint option for endpoint ${endpoint} with type ${endpointType}`,
-      error,
-    );
     return handleError(res, { text: 'Error building endpoint option' });
   }
 }

api/server/middleware/denyRequest.js

@@ -1,7 +1,6 @@
 const crypto = require('crypto');
-const { sendEvent } = require('@librechat/api');
 const { getResponseSender, Constants } = require('librechat-data-provider');
-const { sendError } = require('~/server/middleware/error');
+const { sendMessage, sendError } = require('~/server/utils');
 const { saveMessage } = require('~/models');
 
 /**
@@ -37,7 +36,7 @@ const denyRequest = async (req, res, errorMessage) => {
     isCreatedByUser: true,
     text,
   };
-  sendEvent(res, { message: userMessage, created: true });
+  sendMessage(res, { message: userMessage, created: true });
 
   const shouldSaveMessage = _convoId && parentMessageId && parentMessageId !== Constants.NO_PARENT;
 

api/server/middleware/index.js

@@ -8,6 +8,7 @@ const concurrentLimiter = require('./concurrentLimiter');
 const validateEndpoint = require('./validateEndpoint');
 const requireLocalAuth = require('./requireLocalAuth');
 const canDeleteAccount = require('./canDeleteAccount');
+const accessResources = require('./accessResources');
 const setBalanceConfig = require('./setBalanceConfig');
 const requireLdapAuth = require('./requireLdapAuth');
 const abortMiddleware = require('./abortMiddleware');
@@ -29,6 +30,7 @@ module.exports = {
   ...validate,
   ...limiters,
   ...roles,
+  ...accessResources,
   noIndex,
   checkBan,
   uaParser,

api/server/middleware/roles/access.js

@@ -0,0 +1,78 @@
+const { getRoleByName } = require('~/models/Role');
+const { logger } = require('~/config');
+
+/**
+ * Core function to check if a user has one or more required permissions
+ *
+ * @param {object} user - The user object
+ * @param {PermissionTypes} permissionType - The type of permission to check
+ * @param {Permissions[]} permissions - The list of specific permissions to check
+ * @param {Record<Permissions, string[]>} [bodyProps] - An optional object where keys are permissions and values are arrays of properties to check
+ * @param {object} [checkObject] - The object to check properties against
+ * @returns {Promise<boolean>} Whether the user has the required permissions
+ */
+const checkAccess = async (user, permissionType, permissions, bodyProps = {}, checkObject = {}) => {
+  if (!user) {
+    return false;
+  }
+
+  const role = await getRoleByName(user.role);
+  if (role && role.permissions && role.permissions[permissionType]) {
+    const hasAnyPermission = permissions.some((permission) => {
+      if (role.permissions[permissionType][permission]) {
+        return true;
+      }
+
+      if (bodyProps[permission] && checkObject) {
+        return bodyProps[permission].some((prop) =>
+          Object.prototype.hasOwnProperty.call(checkObject, prop),
+        );
+      }
+
+      return false;
+    });
+
+    return hasAnyPermission;
+  }
+
+  return false;
+};
+
+/**
+ * Middleware to check if a user has one or more required permissions, optionally based on `req.body` properties.
+ *
+ * @param {PermissionTypes} permissionType - The type of permission to check.
+ * @param {Permissions[]} permissions - The list of specific permissions to check.
+ * @param {Record<Permissions, string[]>} [bodyProps] - An optional object where keys are permissions and values are arrays of `req.body` properties to check.
+ * @returns {(req: ServerRequest, res: ServerResponse, next: NextFunction) => Promise<void>} Express middleware function.
+ */
+const generateCheckAccess = (permissionType, permissions, bodyProps = {}) => {
+  return async (req, res, next) => {
+    try {
+      const hasAccess = await checkAccess(
+        req.user,
+        permissionType,
+        permissions,
+        bodyProps,
+        req.body,
+      );
+
+      if (hasAccess) {
+        return next();
+      }
+
+      logger.warn(
+        `[${permissionType}] Forbidden: Insufficient permissions for User ${req.user.id}: ${permissions.join(', ')}`,
+      );
+      return res.status(403).json({ message: 'Forbidden: Insufficient permissions' });
+    } catch (error) {
+      logger.error(error);
+      return res.status(500).json({ message: `Server error: ${error.message}` });
+    }
+  };
+};
+
+module.exports = {
+  checkAccess,
+  generateCheckAccess,
+};

api/server/middleware/roles/access.spec.js

@@ -0,0 +1,251 @@
+const mongoose = require('mongoose');
+const { MongoMemoryServer } = require('mongodb-memory-server');
+const { checkAccess, generateCheckAccess } = require('./access');
+const { PermissionTypes, Permissions } = require('librechat-data-provider');
+const { Role } = require('~/db/models');
+
+// Mock only the logger
+jest.mock('~/config', () => ({
+  logger: {
+    warn: jest.fn(),
+    error: jest.fn(),
+  },
+}));
+
+describe('Access Middleware', () => {
+  let mongoServer;
+  let req, res, next;
+
+  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();
+
+    // Create test roles
+    await Role.create({
+      name: 'user',
+      permissions: {
+        [PermissionTypes.AGENTS]: {
+          [Permissions.USE]: true,
+          [Permissions.CREATE]: false,
+          [Permissions.SHARED_GLOBAL]: false,
+        },
+      },
+    });
+
+    await Role.create({
+      name: 'admin',
+      permissions: {
+        [PermissionTypes.AGENTS]: {
+          [Permissions.USE]: true,
+          [Permissions.CREATE]: true,
+          [Permissions.SHARED_GLOBAL]: true,
+        },
+      },
+    });
+
+    req = {
+      user: { id: 'user123', role: 'user' },
+      body: {},
+    };
+    res = {
+      status: jest.fn().mockReturnThis(),
+      json: jest.fn(),
+    };
+    next = jest.fn();
+    jest.clearAllMocks();
+  });
+
+  describe('checkAccess', () => {
+    test('should return false if user is not provided', async () => {
+      const result = await checkAccess(null, PermissionTypes.AGENTS, [Permissions.USE]);
+      expect(result).toBe(false);
+    });
+
+    test('should return true if user has required permission', async () => {
+      const result = await checkAccess(req.user, PermissionTypes.AGENTS, [Permissions.USE]);
+      expect(result).toBe(true);
+    });
+
+    test('should return false if user lacks required permission', async () => {
+      const result = await checkAccess(req.user, PermissionTypes.AGENTS, [Permissions.CREATE]);
+      expect(result).toBe(false);
+    });
+
+    test('should return true if user has any of multiple permissions', async () => {
+      const result = await checkAccess(req.user, PermissionTypes.AGENTS, [
+        Permissions.USE,
+        Permissions.CREATE,
+      ]);
+      expect(result).toBe(true);
+    });
+
+    test('should check body properties when permission is not directly granted', async () => {
+      // User role doesn't have CREATE permission, but bodyProps allows it
+      const bodyProps = {
+        [Permissions.CREATE]: ['agentId', 'name'],
+      };
+
+      const checkObject = { agentId: 'agent123' };
+
+      const result = await checkAccess(
+        req.user,
+        PermissionTypes.AGENTS,
+        [Permissions.CREATE],
+        bodyProps,
+        checkObject,
+      );
+      expect(result).toBe(true);
+    });
+
+    test('should return false if role is not found', async () => {
+      req.user.role = 'nonexistent';
+      const result = await checkAccess(req.user, PermissionTypes.AGENTS, [Permissions.USE]);
+      expect(result).toBe(false);
+    });
+
+    test('should return false if role has no permissions for the requested type', async () => {
+      await Role.create({
+        name: 'limited',
+        permissions: {
+          // Explicitly set AGENTS permissions to false
+          [PermissionTypes.AGENTS]: {
+            [Permissions.USE]: false,
+            [Permissions.CREATE]: false,
+            [Permissions.SHARED_GLOBAL]: false,
+          },
+          // Has permissions for other types
+          [PermissionTypes.PROMPTS]: {
+            [Permissions.USE]: true,
+          },
+        },
+      });
+      req.user.role = 'limited';
+
+      const result = await checkAccess(req.user, PermissionTypes.AGENTS, [Permissions.USE]);
+      expect(result).toBe(false);
+    });
+
+    test('should handle admin role with all permissions', async () => {
+      req.user.role = 'admin';
+
+      const createResult = await checkAccess(req.user, PermissionTypes.AGENTS, [
+        Permissions.CREATE,
+      ]);
+      expect(createResult).toBe(true);
+
+      const shareResult = await checkAccess(req.user, PermissionTypes.AGENTS, [
+        Permissions.SHARED_GLOBAL,
+      ]);
+      expect(shareResult).toBe(true);
+    });
+  });
+
+  describe('generateCheckAccess', () => {
+    test('should call next() when user has required permission', async () => {
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+      expect(res.status).not.toHaveBeenCalled();
+    });
+
+    test('should return 403 when user lacks permission', async () => {
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.CREATE]);
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(403);
+      expect(res.json).toHaveBeenCalledWith({ message: 'Forbidden: Insufficient permissions' });
+    });
+
+    test('should check body properties when configured', async () => {
+      req.body = { agentId: 'agent123', description: 'test' };
+
+      const bodyProps = {
+        [Permissions.CREATE]: ['agentId'],
+      };
+
+      const middleware = generateCheckAccess(
+        PermissionTypes.AGENTS,
+        [Permissions.CREATE],
+        bodyProps,
+      );
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+      expect(res.status).not.toHaveBeenCalled();
+    });
+
+    test('should handle database errors gracefully', async () => {
+      // Create a user with an invalid role that will cause getRoleByName to fail
+      req.user.role = { invalid: 'object' }; // This will cause an error when querying
+
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(500);
+      expect(res.json).toHaveBeenCalledWith({
+        message: expect.stringContaining('Server error:'),
+      });
+    });
+
+    test('should work with multiple permission types', async () => {
+      req.user.role = 'admin';
+
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [
+        Permissions.USE,
+        Permissions.CREATE,
+        Permissions.SHARED_GLOBAL,
+      ]);
+      await middleware(req, res, next);
+
+      expect(next).toHaveBeenCalled();
+    });
+
+    test('should handle missing user gracefully', async () => {
+      req.user = null;
+
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(500);
+      expect(res.json).toHaveBeenCalledWith({
+        message: expect.stringContaining('Server error:'),
+      });
+    });
+
+    test('should handle role with no AGENTS permissions', async () => {
+      await Role.create({
+        name: 'noaccess',
+        permissions: {
+          // Explicitly set AGENTS with all permissions false
+          [PermissionTypes.AGENTS]: {
+            [Permissions.USE]: false,
+            [Permissions.CREATE]: false,
+            [Permissions.SHARED_GLOBAL]: false,
+          },
+        },
+      });
+      req.user.role = 'noaccess';
+
+      const middleware = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+      await middleware(req, res, next);
+
+      expect(next).not.toHaveBeenCalled();
+      expect(res.status).toHaveBeenCalledWith(403);
+      expect(res.json).toHaveBeenCalledWith({ message: 'Forbidden: Insufficient permissions' });
+    });
+  });
+});

api/server/middleware/roles/index.js

@@ -1,5 +1,8 @@
 const checkAdmin = require('./admin');
+const { checkAccess, generateCheckAccess } = require('./access');
 
 module.exports = {
   checkAdmin,
+  checkAccess,
+  generateCheckAccess,
 };

api/server/routes/accessPermissions.js

@@ -0,0 +1,62 @@
+const express = require('express');
+const { PermissionBits } = require('@librechat/data-schemas');
+const {
+  getUserEffectivePermissions,
+  updateResourcePermissions,
+  getResourcePermissions,
+  getResourceRoles,
+  searchPrincipals,
+} = require('~/server/controllers/PermissionsController');
+const { requireJwtAuth, checkBan, uaParser, canAccessResource } = require('~/server/middleware');
+
+const router = express.Router();
+
+// Apply common middleware
+router.use(requireJwtAuth);
+router.use(checkBan);
+router.use(uaParser);
+
+/**
+ * Generic routes for resource permissions
+ * Pattern: /api/permissions/{resourceType}/{resourceId}
+ */
+
+/**
+ * GET /api/permissions/search-principals
+ * Search for users and groups to grant permissions
+ */
+router.get('/search-principals', searchPrincipals);
+
+/**
+ * GET /api/permissions/{resourceType}/roles
+ * Get available roles for a resource type
+ */
+router.get('/:resourceType/roles', getResourceRoles);
+
+/**
+ * GET /api/permissions/{resourceType}/{resourceId}
+ * Get all permissions for a specific resource
+ */
+router.get('/:resourceType/:resourceId', getResourcePermissions);
+
+/**
+ * PUT /api/permissions/{resourceType}/{resourceId}
+ * Bulk update permissions for a specific resource
+ */
+router.put(
+  '/:resourceType/:resourceId',
+  canAccessResource({
+    resourceType: 'agent',
+    requiredPermission: PermissionBits.SHARE,
+    resourceIdParam: 'resourceId',
+  }),
+  updateResourcePermissions,
+);
+
+/**
+ * GET /api/permissions/{resourceType}/{resourceId}/effective
+ * Get user's effective permissions for a specific resource
+ */
+router.get('/:resourceType/:resourceId/effective', getUserEffectivePermissions);
+
+module.exports = router;

api/server/routes/agents/actions.js

@@ -1,34 +1,15 @@
 const express = require('express');
 const { nanoid } = require('nanoid');
-const { logger } = require('@librechat/data-schemas');
-const { generateCheckAccess } = require('@librechat/api');
-const {
-  SystemRoles,
-  Permissions,
-  PermissionTypes,
-  actionDelimiter,
-  removeNullishValues,
-} = require('librechat-data-provider');
+const { logger, PermissionBits } = require('@librechat/data-schemas');
+const { actionDelimiter, removeNullishValues } = require('librechat-data-provider');
 const { encryptMetadata, domainParser } = require('~/server/services/ActionService');
 const { updateAction, getActions, deleteAction } = require('~/models/Action');
 const { isActionDomainAllowed } = require('~/server/services/domains');
+const { canAccessAgentResource } = require('~/server/middleware');
 const { getAgent, updateAgent } = require('~/models/Agent');
-const { getRoleByName } = require('~/models/Role');
 
 const router = express.Router();
 
-const checkAgentCreate = generateCheckAccess({
-  permissionType: PermissionTypes.AGENTS,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  getRoleByName,
-});
-
-// If the user has ADMIN role
-// then action edition is possible even if not owner of the assistant
-const isAdmin = (req) => {
-  return req.user.role === SystemRoles.ADMIN;
-};
-
 /**
  * Retrieves all user's actions
  * @route GET /actions/
@@ -37,9 +18,8 @@ const isAdmin = (req) => {
  */
 router.get('/', async (req, res) => {
   try {
-    const admin = isAdmin(req);
-    // If admin, get all actions, otherwise only user's actions
-    const searchParams = admin ? {} : { user: req.user.id };
+    // Get all actions for the user (admin permissions handled by middleware if needed)
+    const searchParams = { user: req.user.id };
     res.json(await getActions(searchParams));
   } catch (error) {
     res.status(500).json({ error: error.message });
@@ -55,106 +35,110 @@ router.get('/', async (req, res) => {
  * @param {ActionMetadata} req.body.metadata - Metadata for the action.
  * @returns {Object} 200 - success response - application/json
  */
-router.post('/:agent_id', checkAgentCreate, async (req, res) => {
-  try {
-    const { agent_id } = req.params;
+router.post(
+  '/:agent_id',
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.EDIT,
+    resourceIdParam: 'agent_id',
+  }),
+  async (req, res) => {
+    try {
+      const { agent_id } = req.params;
 
-    /** @type {{ functions: FunctionTool[], action_id: string, metadata: ActionMetadata }} */
-    const { functions, action_id: _action_id, metadata: _metadata } = req.body;
-    if (!functions.length) {
-      return res.status(400).json({ message: 'No functions provided' });
-    }
-
-    let metadata = await encryptMetadata(removeNullishValues(_metadata, true));
-    const isDomainAllowed = await isActionDomainAllowed(metadata.domain);
-    if (!isDomainAllowed) {
-      return res.status(400).json({ message: 'Domain not allowed' });
-    }
-
-    let { domain } = metadata;
-    domain = await domainParser(domain, true);
-
-    if (!domain) {
-      return res.status(400).json({ message: 'No domain provided' });
-    }
-
-    const action_id = _action_id ?? nanoid();
-    const initialPromises = [];
-    const admin = isAdmin(req);
-
-    // If admin, can edit any agent, otherwise only user's agents
-    const agentQuery = admin ? { id: agent_id } : { id: agent_id, author: req.user.id };
-    // TODO: share agents
-    initialPromises.push(getAgent(agentQuery));
-    if (_action_id) {
-      initialPromises.push(getActions({ action_id }, true));
-    }
-
-    /** @type {[Agent, [Action|undefined]]} */
-    const [agent, actions_result] = await Promise.all(initialPromises);
-    if (!agent) {
-      return res.status(404).json({ message: 'Agent not found for adding action' });
-    }
-
-    if (actions_result && actions_result.length) {
-      const action = actions_result[0];
-      metadata = { ...action.metadata, ...metadata };
-    }
-
-    const { actions: _actions = [], author: agent_author } = agent ?? {};
-    const actions = [];
-    for (const action of _actions) {
-      const [_action_domain, current_action_id] = action.split(actionDelimiter);
-      if (current_action_id === action_id) {
-        continue;
+      /** @type {{ functions: FunctionTool[], action_id: string, metadata: ActionMetadata }} */
+      const { functions, action_id: _action_id, metadata: _metadata } = req.body;
+      if (!functions.length) {
+        return res.status(400).json({ message: 'No functions provided' });
       }
 
-      actions.push(action);
-    }
-
-    actions.push(`${domain}${actionDelimiter}${action_id}`);
-
-    /** @type {string[]}} */
-    const { tools: _tools = [] } = agent;
-
-    const tools = _tools
-      .filter((tool) => !(tool && (tool.includes(domain) || tool.includes(action_id))))
-      .concat(functions.map((tool) => `${tool.function.name}${actionDelimiter}${domain}`));
-
-    // Force version update since actions are changing
-    const updatedAgent = await updateAgent(
-      agentQuery,
-      { tools, actions },
-      {
-        updatingUserId: req.user.id,
-        forceVersion: true,
-      },
-    );
-
-    // Only update user field for new actions
-    const actionUpdateData = { metadata, agent_id };
-    if (!actions_result || !actions_result.length) {
-      // For new actions, use the agent owner's user ID
-      actionUpdateData.user = agent_author || req.user.id;
-    }
-
-    /** @type {[Action]} */
-    const updatedAction = await updateAction({ action_id }, actionUpdateData);
-
-    const sensitiveFields = ['api_key', 'oauth_client_id', 'oauth_client_secret'];
-    for (let field of sensitiveFields) {
-      if (updatedAction.metadata[field]) {
-        delete updatedAction.metadata[field];
+      let metadata = await encryptMetadata(removeNullishValues(_metadata, true));
+      const isDomainAllowed = await isActionDomainAllowed(metadata.domain);
+      if (!isDomainAllowed) {
+        return res.status(400).json({ message: 'Domain not allowed' });
       }
-    }
 
-    res.json([updatedAgent, updatedAction]);
-  } catch (error) {
-    const message = 'Trouble updating the Agent Action';
-    logger.error(message, error);
-    res.status(500).json({ message });
-  }
-});
+      let { domain } = metadata;
+      domain = await domainParser(domain, true);
+
+      if (!domain) {
+        return res.status(400).json({ message: 'No domain provided' });
+      }
+
+      const action_id = _action_id ?? nanoid();
+      const initialPromises = [];
+
+      // Permissions already validated by middleware - load agent directly
+      initialPromises.push(getAgent({ id: agent_id }));
+      if (_action_id) {
+        initialPromises.push(getActions({ action_id }, true));
+      }
+
+      /** @type {[Agent, [Action|undefined]]} */
+      const [agent, actions_result] = await Promise.all(initialPromises);
+      if (!agent) {
+        return res.status(404).json({ message: 'Agent not found for adding action' });
+      }
+
+      if (actions_result && actions_result.length) {
+        const action = actions_result[0];
+        metadata = { ...action.metadata, ...metadata };
+      }
+
+      const { actions: _actions = [], author: agent_author } = agent ?? {};
+      const actions = [];
+      for (const action of _actions) {
+        const [_action_domain, current_action_id] = action.split(actionDelimiter);
+        if (current_action_id === action_id) {
+          continue;
+        }
+
+        actions.push(action);
+      }
+
+      actions.push(`${domain}${actionDelimiter}${action_id}`);
+
+      /** @type {string[]}} */
+      const { tools: _tools = [] } = agent;
+
+      const tools = _tools
+        .filter((tool) => !(tool && (tool.includes(domain) || tool.includes(action_id))))
+        .concat(functions.map((tool) => `${tool.function.name}${actionDelimiter}${domain}`));
+
+      // Force version update since actions are changing
+      const updatedAgent = await updateAgent(
+        { id: agent_id },
+        { tools, actions },
+        {
+          updatingUserId: req.user.id,
+          forceVersion: true,
+        },
+      );
+
+      // Only update user field for new actions
+      const actionUpdateData = { metadata, agent_id };
+      if (!actions_result || !actions_result.length) {
+        // For new actions, use the agent owner's user ID
+        actionUpdateData.user = agent_author || req.user.id;
+      }
+
+      /** @type {[Action]} */
+      const updatedAction = await updateAction({ action_id }, actionUpdateData);
+
+      const sensitiveFields = ['api_key', 'oauth_client_id', 'oauth_client_secret'];
+      for (let field of sensitiveFields) {
+        if (updatedAction.metadata[field]) {
+          delete updatedAction.metadata[field];
+        }
+      }
+
+      res.json([updatedAgent, updatedAction]);
+    } catch (error) {
+      const message = 'Trouble updating the Agent Action';
+      logger.error(message, error);
+      res.status(500).json({ message });
+    }
+  },
+);
 
 /**
  * Deletes an action for a specific agent.
@@ -163,52 +147,55 @@ router.post('/:agent_id', checkAgentCreate, async (req, res) => {
  * @param {string} req.params.action_id - The ID of the action to delete.
  * @returns {Object} 200 - success response - application/json
  */
-router.delete('/:agent_id/:action_id', checkAgentCreate, async (req, res) => {
-  try {
-    const { agent_id, action_id } = req.params;
-    const admin = isAdmin(req);
+router.delete(
+  '/:agent_id/:action_id',
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.EDIT,
+    resourceIdParam: 'agent_id',
+  }),
+  async (req, res) => {
+    try {
+      const { agent_id, action_id } = req.params;
 
-    // If admin, can delete any agent, otherwise only user's agents
-    const agentQuery = admin ? { id: agent_id } : { id: agent_id, author: req.user.id };
-    const agent = await getAgent(agentQuery);
-    if (!agent) {
-      return res.status(404).json({ message: 'Agent not found for deleting action' });
-    }
-
-    const { tools = [], actions = [] } = agent;
-
-    let domain = '';
-    const updatedActions = actions.filter((action) => {
-      if (action.includes(action_id)) {
-        [domain] = action.split(actionDelimiter);
-        return false;
+      // Permissions already validated by middleware - load agent directly
+      const agent = await getAgent({ id: agent_id });
+      if (!agent) {
+        return res.status(404).json({ message: 'Agent not found for deleting action' });
       }
-      return true;
-    });
 
-    domain = await domainParser(domain, true);
+      const { tools = [], actions = [] } = agent;
 
-    if (!domain) {
-      return res.status(400).json({ message: 'No domain provided' });
+      let domain = '';
+      const updatedActions = actions.filter((action) => {
+        if (action.includes(action_id)) {
+          [domain] = action.split(actionDelimiter);
+          return false;
+        }
+        return true;
+      });
+
+      domain = await domainParser(domain, true);
+
+      if (!domain) {
+        return res.status(400).json({ message: 'No domain provided' });
+      }
+
+      const updatedTools = tools.filter((tool) => !(tool && tool.includes(domain)));
+
+      // Force version update since actions are being removed
+      await updateAgent(
+        { id: agent_id },
+        { tools: updatedTools, actions: updatedActions },
+        { updatingUserId: req.user.id, forceVersion: true },
+      );
+      await deleteAction({ action_id });
+      res.status(200).json({ message: 'Action deleted successfully' });
+    } catch (error) {
+      const message = 'Trouble deleting the Agent Action';
+      logger.error(message, error);
+      res.status(500).json({ message });
     }
-
-    const updatedTools = tools.filter((tool) => !(tool && tool.includes(domain)));
-
-    // Force version update since actions are being removed
-    await updateAgent(
-      agentQuery,
-      { tools: updatedTools, actions: updatedActions },
-      { updatingUserId: req.user.id, forceVersion: true },
-    );
-    // If admin, can delete any action, otherwise only user's actions
-    const actionQuery = admin ? { action_id } : { action_id, user: req.user.id };
-    await deleteAction(actionQuery);
-    res.status(200).json({ message: 'Action deleted successfully' });
-  } catch (error) {
-    const message = 'Trouble deleting the Agent Action';
-    logger.error(message, error);
-    res.status(500).json({ message });
-  }
-});
+  },
+);
 
 module.exports = router;

api/server/routes/agents/chat.js

@@ -1,30 +1,30 @@
 const express = require('express');
-const { generateCheckAccess, skipAgentCheck } = require('@librechat/api');
+const { PermissionBits } = require('@librechat/data-schemas');
 const { PermissionTypes, Permissions } = require('librechat-data-provider');
 const {
   setHeaders,
   moderateText,
   // validateModel,
+  generateCheckAccess,
   validateConvoAccess,
   buildEndpointOption,
+  canAccessAgentFromBody,
 } = require('~/server/middleware');
 const { initializeClient } = require('~/server/services/Endpoints/agents');
 const AgentController = require('~/server/controllers/agents/request');
 const addTitle = require('~/server/services/Endpoints/agents/title');
-const { getRoleByName } = require('~/models/Role');
 
 const router = express.Router();
 
 router.use(moderateText);
 
-const checkAgentAccess = generateCheckAccess({
-  permissionType: PermissionTypes.AGENTS,
-  permissions: [Permissions.USE],
-  skipCheck: skipAgentCheck,
-  getRoleByName,
+const checkAgentAccess = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+const checkAgentResourceAccess = canAccessAgentFromBody({
+  requiredPermission: PermissionBits.VIEW,
 });
 
 router.use(checkAgentAccess);
+router.use(checkAgentResourceAccess);
 router.use(validateConvoAccess);
 router.use(buildEndpointOption);
 router.use(setHeaders);

api/server/routes/agents/tools.js

@@ -1,5 +1,4 @@
 const express = require('express');
-const { addTool, updateTool, deleteTool } = require('@librechat/api');
 const { callTool, verifyToolAuth, getToolCalls } = require('~/server/controllers/tools');
 const { getAvailableTools } = require('~/server/controllers/PluginController');
 const { toolCallLimiter } = require('~/server/middleware/limiters');
@@ -37,29 +36,4 @@ router.get('/:toolId/auth', verifyToolAuth);
  */
 router.post('/:toolId/call', toolCallLimiter, callTool);
 
-/**
- * Add a new tool/MCP to the system
- * @route POST /agents/tools/add
- * @param {object} req.body - Request body containing tool/MCP data
- * @returns {object} Created tool/MCP object
- */
-router.post('/add', addTool);
-
-/**
- * Update an existing tool/MCP in the system
- * @route PUT /agents/tools/:mcp_id
- * @param {string} mcp_id - The ID of the MCP to update
- * @param {object} req.body - Request body containing updated tool/MCP data
- * @returns {object} Updated tool/MCP object
- */
-router.put('/:mcp_id', updateTool);
-
-/**
- * Delete a tool/MCP from the system
- * @route DELETE /agents/tools/:mcp_id
- * @param {string} mcp_id - The ID of the MCP to delete
- * @returns {object} Deletion confirmation
- */
-router.delete('/:mcp_id', deleteTool);
-
 module.exports = router;

api/server/routes/agents/v1.js

@@ -1,36 +1,34 @@
 const express = require('express');
-const { generateCheckAccess } = require('@librechat/api');
+const { PermissionBits } = require('@librechat/data-schemas');
 const { PermissionTypes, Permissions } = require('librechat-data-provider');
-const { requireJwtAuth } = require('~/server/middleware');
+const {
+  requireJwtAuth,
+  generateCheckAccess,
+  canAccessAgentResource,
+} = require('~/server/middleware');
 const v1 = require('~/server/controllers/agents/v1');
-const { getRoleByName } = require('~/models/Role');
 const actions = require('./actions');
 const tools = require('./tools');
 
 const router = express.Router();
 const avatar = express.Router();
 
-const checkAgentAccess = generateCheckAccess({
-  permissionType: PermissionTypes.AGENTS,
-  permissions: [Permissions.USE],
-  getRoleByName,
-});
-const checkAgentCreate = generateCheckAccess({
-  permissionType: PermissionTypes.AGENTS,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  getRoleByName,
-});
+const checkAgentAccess = generateCheckAccess(PermissionTypes.AGENTS, [Permissions.USE]);
+const checkAgentCreate = generateCheckAccess(PermissionTypes.AGENTS, [
+  Permissions.USE,
+  Permissions.CREATE,
+]);
 
-const checkGlobalAgentShare = generateCheckAccess({
-  permissionType: PermissionTypes.AGENTS,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  bodyProps: {
+const checkGlobalAgentShare = generateCheckAccess(
+  PermissionTypes.AGENTS,
+  [Permissions.USE, Permissions.CREATE],
+  {
     [Permissions.SHARED_GLOBAL]: ['projectIds', 'removeProjectIds'],
   },
-  getRoleByName,
-});
+);
 
 router.use(requireJwtAuth);
+router.use(checkAgentAccess);
 
 /**
  * Agent actions route.
@@ -53,13 +51,38 @@ router.use('/tools', tools);
 router.post('/', checkAgentCreate, v1.createAgent);
 
 /**
- * Retrieves an agent.
+ * Retrieves basic agent information (VIEW permission required).
+ * Returns safe, non-sensitive agent data for viewing purposes.
  * @route GET /agents/:id
  * @param {string} req.params.id - Agent identifier.
- * @returns {Agent} 200 - Success response - application/json
+ * @returns {Agent} 200 - Basic agent info - application/json
  */
-router.get('/:id', checkAgentAccess, v1.getAgent);
+router.get(
+  '/:id',
+  checkAgentAccess,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.VIEW,
+    resourceIdParam: 'id',
+  }),
+  v1.getAgent,
+);
 
+/**
+ * Retrieves full agent details including sensitive configuration (EDIT permission required).
+ * Returns complete agent data for editing/configuration purposes.
+ * @route GET /agents/:id/expanded
+ * @param {string} req.params.id - Agent identifier.
+ * @returns {Agent} 200 - Full agent details - application/json
+ */
+router.get(
+  '/:id/expanded',
+  checkAgentAccess,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.EDIT,
+    resourceIdParam: 'id',
+  }),
+  (req, res) => v1.getAgent(req, res, true), // Expanded version
+);
 /**
  * Updates an agent.
  * @route PATCH /agents/:id
@@ -67,7 +90,15 @@ router.get('/:id', checkAgentAccess, v1.getAgent);
  * @param {AgentUpdateParams} req.body - The agent update parameters.
  * @returns {Agent} 200 - Success response - application/json
  */
-router.patch('/:id', checkGlobalAgentShare, v1.updateAgent);
+router.patch(
+  '/:id',
+  checkGlobalAgentShare,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.EDIT,
+    resourceIdParam: 'id',
+  }),
+  v1.updateAgent,
+);
 
 /**
  * Duplicates an agent.
@@ -75,7 +106,15 @@ router.patch('/:id', checkGlobalAgentShare, v1.updateAgent);
  * @param {string} req.params.id - Agent identifier.
  * @returns {Agent} 201 - Success response - application/json
  */
-router.post('/:id/duplicate', checkAgentCreate, v1.duplicateAgent);
+router.post(
+  '/:id/duplicate',
+  checkAgentCreate,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.VIEW,
+    resourceIdParam: 'id',
+  }),
+  v1.duplicateAgent,
+);
 
 /**
  * Deletes an agent.
@@ -83,7 +122,15 @@ router.post('/:id/duplicate', checkAgentCreate, v1.duplicateAgent);
  * @param {string} req.params.id - Agent identifier.
  * @returns {Agent} 200 - success response - application/json
  */
-router.delete('/:id', checkAgentCreate, v1.deleteAgent);
+router.delete(
+  '/:id',
+  checkAgentCreate,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.DELETE,
+    resourceIdParam: 'id',
+  }),
+  v1.deleteAgent,
+);
 
 /**
  * Reverts an agent to a previous version.
@@ -110,6 +157,14 @@ router.get('/', checkAgentAccess, v1.getListAgents);
  * @param {string} [req.body.metadata] - Optional metadata for the agent's avatar.
  * @returns {Object} 200 - success response - application/json
  */
-avatar.post('/:agent_id/avatar/', checkAgentAccess, v1.uploadAgentAvatar);
+avatar.post(
+  '/:agent_id/avatar/',
+  checkAgentAccess,
+  canAccessAgentResource({
+    requiredPermission: PermissionBits.EDIT,
+    resourceIdParam: 'agent_id',
+  }),
+  v1.uploadAgentAvatar,
+);
 
 module.exports = { v1: router, avatar };

api/server/routes/ask/addToCache.js

@@ -0,0 +1,63 @@
+const { Keyv } = require('keyv');
+const { KeyvFile } = require('keyv-file');
+const { logger } = require('~/config');
+
+const addToCache = async ({ endpoint, endpointOption, userMessage, responseMessage }) => {
+  try {
+    const conversationsCache = new Keyv({
+      store: new KeyvFile({ filename: './data/cache.json' }),
+      namespace: 'chatgpt', // should be 'bing' for bing/sydney
+    });
+
+    const {
+      conversationId,
+      messageId: userMessageId,
+      parentMessageId: userParentMessageId,
+      text: userText,
+    } = userMessage;
+    const {
+      messageId: responseMessageId,
+      parentMessageId: responseParentMessageId,
+      text: responseText,
+    } = responseMessage;
+
+    let conversation = await conversationsCache.get(conversationId);
+    // used to generate a title for the conversation if none exists
+    // let isNewConversation = false;
+    if (!conversation) {
+      conversation = {
+        messages: [],
+        createdAt: Date.now(),
+      };
+      // isNewConversation = true;
+    }
+
+    const roles = (options) => {
+      if (endpoint === 'openAI') {
+        return options?.chatGptLabel || 'ChatGPT';
+      }
+    };
+
+    let _userMessage = {
+      id: userMessageId,
+      parentMessageId: userParentMessageId,
+      role: 'User',
+      message: userText,
+    };
+
+    let _responseMessage = {
+      id: responseMessageId,
+      parentMessageId: responseParentMessageId,
+      role: roles(endpointOption),
+      message: responseText,
+    };
+
+    conversation.messages.push(_userMessage, _responseMessage);
+
+    await conversationsCache.set(conversationId, conversation);
+  } catch (error) {
+    logger.error('[addToCache] Error adding conversation to cache', error);
+  }
+};
+
+module.exports = addToCache;

api/server/routes/ask/anthropic.js

@@ -0,0 +1,25 @@
+const express = require('express');
+const AskController = require('~/server/controllers/AskController');
+const { addTitle, initializeClient } = require('~/server/services/Endpoints/anthropic');
+const {
+  setHeaders,
+  handleAbort,
+  validateModel,
+  validateEndpoint,
+  buildEndpointOption,
+} = require('~/server/middleware');
+
+const router = express.Router();
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res, next) => {
+    await AskController(req, res, next, initializeClient, addTitle);
+  },
+);
+
+module.exports = router;

api/server/routes/ask/custom.js

@@ -0,0 +1,25 @@
+const express = require('express');
+const AskController = require('~/server/controllers/AskController');
+const { initializeClient } = require('~/server/services/Endpoints/custom');
+const { addTitle } = require('~/server/services/Endpoints/openAI');
+const {
+  setHeaders,
+  validateModel,
+  validateEndpoint,
+  buildEndpointOption,
+} = require('~/server/middleware');
+
+const router = express.Router();
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res, next) => {
+    await AskController(req, res, next, initializeClient, addTitle);
+  },
+);
+
+module.exports = router;

api/server/routes/ask/google.js

@@ -0,0 +1,24 @@
+const express = require('express');
+const AskController = require('~/server/controllers/AskController');
+const { initializeClient, addTitle } = require('~/server/services/Endpoints/google');
+const {
+  setHeaders,
+  validateModel,
+  validateEndpoint,
+  buildEndpointOption,
+} = require('~/server/middleware');
+
+const router = express.Router();
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res, next) => {
+    await AskController(req, res, next, initializeClient, addTitle);
+  },
+);
+
+module.exports = router;

api/server/routes/ask/gptPlugins.js

@@ -0,0 +1,241 @@
+const express = require('express');
+const { getResponseSender, Constants } = require('librechat-data-provider');
+const { initializeClient } = require('~/server/services/Endpoints/gptPlugins');
+const { sendMessage, createOnProgress } = require('~/server/utils');
+const { addTitle } = require('~/server/services/Endpoints/openAI');
+const { saveMessage, updateMessage } = require('~/models');
+const {
+  handleAbort,
+  createAbortController,
+  handleAbortError,
+  setHeaders,
+  validateModel,
+  validateEndpoint,
+  buildEndpointOption,
+  moderateText,
+} = require('~/server/middleware');
+const { validateTools } = require('~/app');
+const { logger } = require('~/config');
+
+const router = express.Router();
+
+router.use(moderateText);
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res) => {
+    let {
+      text,
+      endpointOption,
+      conversationId,
+      parentMessageId = null,
+      overrideParentMessageId = null,
+    } = req.body;
+
+    logger.debug('[/ask/gptPlugins]', { text, conversationId, ...endpointOption });
+
+    let userMessage;
+    let userMessagePromise;
+    let promptTokens;
+    let userMessageId;
+    let responseMessageId;
+    const sender = getResponseSender({
+      ...endpointOption,
+      model: endpointOption.modelOptions.model,
+    });
+    const newConvo = !conversationId;
+    const user = req.user.id;
+
+    const plugins = [];
+
+    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 (!conversationId && key === 'conversationId') {
+          conversationId = data[key];
+        }
+      }
+    };
+
+    let streaming = null;
+    let timer = null;
+
+    const {
+      onProgress: progressCallback,
+      sendIntermediateMessage,
+      getPartialText,
+    } = createOnProgress({
+      onProgress: () => {
+        if (timer) {
+          clearTimeout(timer);
+        }
+
+        streaming = new Promise((resolve) => {
+          timer = setTimeout(() => {
+            resolve();
+          }, 250);
+        });
+      },
+    });
+
+    const pluginMap = new Map();
+    const onAgentAction = async (action, runId) => {
+      pluginMap.set(runId, action.tool);
+      sendIntermediateMessage(res, {
+        plugins,
+        parentMessageId: userMessage.messageId,
+        messageId: responseMessageId,
+      });
+    };
+
+    const onToolStart = async (tool, input, runId, parentRunId) => {
+      const pluginName = pluginMap.get(parentRunId);
+      const latestPlugin = {
+        runId,
+        loading: true,
+        inputs: [input],
+        latest: pluginName,
+        outputs: null,
+      };
+
+      if (streaming) {
+        await streaming;
+      }
+      const extraTokens = ':::plugin:::\n';
+      plugins.push(latestPlugin);
+      sendIntermediateMessage(
+        res,
+        { plugins, parentMessageId: userMessage.messageId, messageId: responseMessageId },
+        extraTokens,
+      );
+    };
+
+    const onToolEnd = async (output, runId) => {
+      if (streaming) {
+        await streaming;
+      }
+
+      const pluginIndex = plugins.findIndex((plugin) => plugin.runId === runId);
+
+      if (pluginIndex !== -1) {
+        plugins[pluginIndex].loading = false;
+        plugins[pluginIndex].outputs = output;
+      }
+    };
+
+    const getAbortData = () => ({
+      sender,
+      conversationId,
+      userMessagePromise,
+      messageId: responseMessageId,
+      parentMessageId: overrideParentMessageId ?? userMessageId,
+      text: getPartialText(),
+      plugins: plugins.map((p) => ({ ...p, loading: false })),
+      userMessage,
+      promptTokens,
+    });
+    const { abortController, onStart } = createAbortController(req, res, getAbortData, getReqData);
+
+    try {
+      endpointOption.tools = await validateTools(user, endpointOption.tools);
+      const { client } = await initializeClient({ req, res, endpointOption });
+
+      const onChainEnd = () => {
+        if (!client.skipSaveUserMessage) {
+          saveMessage(
+            req,
+            { ...userMessage, user },
+            { context: 'api/server/routes/ask/gptPlugins.js - onChainEnd' },
+          );
+        }
+        sendIntermediateMessage(res, {
+          plugins,
+          parentMessageId: userMessage.messageId,
+          messageId: responseMessageId,
+        });
+      };
+
+      let response = await client.sendMessage(text, {
+        user,
+        conversationId,
+        parentMessageId,
+        overrideParentMessageId,
+        getReqData,
+        onAgentAction,
+        onChainEnd,
+        onToolStart,
+        onToolEnd,
+        onStart,
+        getPartialText,
+        ...endpointOption,
+        progressCallback,
+        progressOptions: {
+          res,
+          // parentMessageId: overrideParentMessageId || userMessageId,
+          plugins,
+        },
+        abortController,
+      });
+
+      if (overrideParentMessageId) {
+        response.parentMessageId = overrideParentMessageId;
+      }
+
+      logger.debug('[/ask/gptPlugins]', response);
+
+      const { conversation = {} } = await response.databasePromise;
+      delete response.databasePromise;
+      conversation.title =
+        conversation && !conversation.title ? null : conversation?.title || 'New Chat';
+
+      sendMessage(res, {
+        title: conversation.title,
+        final: true,
+        conversation,
+        requestMessage: userMessage,
+        responseMessage: response,
+      });
+      res.end();
+
+      if (parentMessageId === Constants.NO_PARENT && newConvo) {
+        addTitle(req, {
+          text,
+          response,
+          client,
+        });
+      }
+
+      response.plugins = plugins.map((p) => ({ ...p, loading: false }));
+      if (response.plugins?.length > 0) {
+        await updateMessage(
+          req,
+          { ...response, user },
+          { context: 'api/server/routes/ask/gptPlugins.js - save plugins used' },
+        );
+      }
+    } catch (error) {
+      const partialText = getPartialText();
+      handleAbortError(res, req, error, {
+        partialText,
+        conversationId,
+        sender,
+        messageId: responseMessageId,
+        parentMessageId: userMessageId ?? parentMessageId,
+      });
+    }
+  },
+);
+
+module.exports = router;

api/server/routes/ask/index.js

@@ -0,0 +1,47 @@
+const express = require('express');
+const { EModelEndpoint } = require('librechat-data-provider');
+const {
+  uaParser,
+  checkBan,
+  requireJwtAuth,
+  messageIpLimiter,
+  concurrentLimiter,
+  messageUserLimiter,
+  validateConvoAccess,
+} = require('~/server/middleware');
+const { isEnabled } = require('~/server/utils');
+const gptPlugins = require('./gptPlugins');
+const anthropic = require('./anthropic');
+const custom = require('./custom');
+const google = require('./google');
+const openAI = require('./openAI');
+
+const { LIMIT_CONCURRENT_MESSAGES, LIMIT_MESSAGE_IP, LIMIT_MESSAGE_USER } = process.env ?? {};
+
+const router = express.Router();
+
+router.use(requireJwtAuth);
+router.use(checkBan);
+router.use(uaParser);
+
+if (isEnabled(LIMIT_CONCURRENT_MESSAGES)) {
+  router.use(concurrentLimiter);
+}
+
+if (isEnabled(LIMIT_MESSAGE_IP)) {
+  router.use(messageIpLimiter);
+}
+
+if (isEnabled(LIMIT_MESSAGE_USER)) {
+  router.use(messageUserLimiter);
+}
+
+router.use(validateConvoAccess);
+
+router.use([`/${EModelEndpoint.azureOpenAI}`, `/${EModelEndpoint.openAI}`], openAI);
+router.use(`/${EModelEndpoint.gptPlugins}`, gptPlugins);
+router.use(`/${EModelEndpoint.anthropic}`, anthropic);
+router.use(`/${EModelEndpoint.google}`, google);
+router.use(`/${EModelEndpoint.custom}`, custom);
+
+module.exports = router;

api/server/routes/ask/openAI.js

@@ -0,0 +1,27 @@
+const express = require('express');
+const AskController = require('~/server/controllers/AskController');
+const { addTitle, initializeClient } = require('~/server/services/Endpoints/openAI');
+const {
+  handleAbort,
+  setHeaders,
+  validateModel,
+  validateEndpoint,
+  buildEndpointOption,
+  moderateText,
+} = require('~/server/middleware');
+
+const router = express.Router();
+router.use(moderateText);
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res, next) => {
+    await AskController(req, res, next, initializeClient, addTitle);
+  },
+);
+
+module.exports = router;

api/server/routes/bedrock/chat.js

@@ -0,0 +1,37 @@
+const express = require('express');
+
+const router = express.Router();
+const {
+  setHeaders,
+  handleAbort,
+  moderateText,
+  // validateModel,
+  // validateEndpoint,
+  buildEndpointOption,
+} = require('~/server/middleware');
+const { initializeClient } = require('~/server/services/Endpoints/bedrock');
+const AgentController = require('~/server/controllers/agents/request');
+const addTitle = require('~/server/services/Endpoints/agents/title');
+
+router.use(moderateText);
+
+/**
+ * @route POST /
+ * @desc Chat with an assistant
+ * @access Public
+ * @param {express.Request} req - The request object, containing the request data.
+ * @param {express.Response} res - The response object, used to send back a response.
+ * @returns {void}
+ */
+router.post(
+  '/',
+  // validateModel,
+  // validateEndpoint,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res, next) => {
+    await AgentController(req, res, next, initializeClient, addTitle);
+  },
+);
+
+module.exports = router;

api/server/routes/bedrock/index.js

@@ -0,0 +1,35 @@
+const express = require('express');
+const {
+  uaParser,
+  checkBan,
+  requireJwtAuth,
+  messageIpLimiter,
+  concurrentLimiter,
+  messageUserLimiter,
+} = require('~/server/middleware');
+const { isEnabled } = require('~/server/utils');
+const chat = require('./chat');
+
+const { LIMIT_CONCURRENT_MESSAGES, LIMIT_MESSAGE_IP, LIMIT_MESSAGE_USER } = process.env ?? {};
+
+const router = express.Router();
+
+router.use(requireJwtAuth);
+router.use(checkBan);
+router.use(uaParser);
+
+if (isEnabled(LIMIT_CONCURRENT_MESSAGES)) {
+  router.use(concurrentLimiter);
+}
+
+if (isEnabled(LIMIT_MESSAGE_IP)) {
+  router.use(messageIpLimiter);
+}
+
+if (isEnabled(LIMIT_MESSAGE_USER)) {
+  router.use(messageUserLimiter);
+}
+
+router.use('/chat', chat);
+
+module.exports = router;

api/server/routes/edit/gptPlugins.js

@@ -0,0 +1,207 @@
+const express = require('express');
+const { getResponseSender } = require('librechat-data-provider');
+const {
+  setHeaders,
+  moderateText,
+  validateModel,
+  handleAbortError,
+  validateEndpoint,
+  buildEndpointOption,
+  createAbortController,
+} = require('~/server/middleware');
+const { sendMessage, createOnProgress, formatSteps, formatAction } = require('~/server/utils');
+const { initializeClient } = require('~/server/services/Endpoints/gptPlugins');
+const { saveMessage, updateMessage } = require('~/models');
+const { validateTools } = require('~/app');
+const { logger } = require('~/config');
+
+const router = express.Router();
+
+router.use(moderateText);
+
+router.post(
+  '/',
+  validateEndpoint,
+  validateModel,
+  buildEndpointOption,
+  setHeaders,
+  async (req, res) => {
+    let {
+      text,
+      generation,
+      endpointOption,
+      conversationId,
+      responseMessageId,
+      isContinued = false,
+      parentMessageId = null,
+      overrideParentMessageId = null,
+    } = req.body;
+
+    logger.debug('[/edit/gptPlugins]', {
+      text,
+      generation,
+      isContinued,
+      conversationId,
+      ...endpointOption,
+    });
+
+    let userMessage;
+    let userMessagePromise;
+    let promptTokens;
+    const sender = getResponseSender({
+      ...endpointOption,
+      model: endpointOption.modelOptions.model,
+    });
+    const userMessageId = parentMessageId;
+    const user = req.user.id;
+
+    const plugin = {
+      loading: true,
+      inputs: [],
+      latest: null,
+      outputs: null,
+    };
+
+    const getReqData = (data = {}) => {
+      for (let key in data) {
+        if (key === 'userMessage') {
+          userMessage = data[key];
+        } else if (key === 'userMessagePromise') {
+          userMessagePromise = data[key];
+        } else if (key === 'responseMessageId') {
+          responseMessageId = data[key];
+        } else if (key === 'promptTokens') {
+          promptTokens = data[key];
+        }
+      }
+    };
+
+    const {
+      onProgress: progressCallback,
+      sendIntermediateMessage,
+      getPartialText,
+    } = createOnProgress({
+      generation,
+      onProgress: () => {
+        if (plugin.loading === true) {
+          plugin.loading = false;
+        }
+      },
+    });
+
+    const onChainEnd = (data) => {
+      let { intermediateSteps: steps } = data;
+      plugin.outputs = steps && steps[0].action ? formatSteps(steps) : 'An error occurred.';
+      plugin.loading = false;
+      saveMessage(
+        req,
+        { ...userMessage, user },
+        { context: 'api/server/routes/ask/gptPlugins.js - onChainEnd' },
+      );
+      sendIntermediateMessage(res, {
+        plugin,
+        parentMessageId: userMessage.messageId,
+        messageId: responseMessageId,
+      });
+      // logger.debug('CHAIN END', plugin.outputs);
+    };
+
+    const getAbortData = () => ({
+      sender,
+      conversationId,
+      userMessagePromise,
+      messageId: responseMessageId,
+      parentMessageId: overrideParentMessageId ?? userMessageId,
+      text: getPartialText(),
+      plugin: { ...plugin, loading: false },
+      userMessage,
+      promptTokens,
+    });
+    const { abortController, onStart } = createAbortController(req, res, getAbortData, getReqData);
+
+    try {
+      endpointOption.tools = await validateTools(user, endpointOption.tools);
+      const { client } = await initializeClient({ req, res, endpointOption });
+
+      const onAgentAction = (action, start = false) => {
+        const formattedAction = formatAction(action);
+        plugin.inputs.push(formattedAction);
+        plugin.latest = formattedAction.plugin;
+        if (!start && !client.skipSaveUserMessage) {
+          saveMessage(
+            req,
+            { ...userMessage, user },
+            { context: 'api/server/routes/ask/gptPlugins.js - onAgentAction' },
+          );
+        }
+        sendIntermediateMessage(res, {
+          plugin,
+          parentMessageId: userMessage.messageId,
+          messageId: responseMessageId,
+        });
+        // logger.debug('PLUGIN ACTION', formattedAction);
+      };
+
+      let response = await client.sendMessage(text, {
+        user,
+        generation,
+        isContinued,
+        isEdited: true,
+        conversationId,
+        parentMessageId,
+        responseMessageId,
+        overrideParentMessageId,
+        getReqData,
+        onAgentAction,
+        onChainEnd,
+        onStart,
+        ...endpointOption,
+        progressCallback,
+        progressOptions: {
+          res,
+          plugin,
+          // parentMessageId: overrideParentMessageId || userMessageId,
+        },
+        abortController,
+      });
+
+      if (overrideParentMessageId) {
+        response.parentMessageId = overrideParentMessageId;
+      }
+
+      logger.debug('[/edit/gptPlugins] CLIENT RESPONSE', response);
+
+      const { conversation = {} } = await response.databasePromise;
+      delete response.databasePromise;
+      conversation.title =
+        conversation && !conversation.title ? null : conversation?.title || 'New Chat';
+
+      sendMessage(res, {
+        title: conversation.title,
+        final: true,
+        conversation,
+        requestMessage: userMessage,
+        responseMessage: response,
+      });
+      res.end();
+
+      response.plugin = { ...plugin, loading: false };
+      await updateMessage(
+        req,
+        { ...response, user },
+        { context: 'api/server/routes/edit/gptPlugins.js' },
+      );
+    } catch (error) {
+      const partialText = getPartialText();
+      handleAbortError(res, req, error, {
+        partialText,
+        conversationId,
+        sender,
+        messageId: responseMessageId,
+        parentMessageId: userMessageId ?? parentMessageId,
+      });
+    }
+  },
+);
+
+module.exports = router;

api/server/routes/edit/index.js

@@ -3,6 +3,7 @@ const openAI = require('./openAI');
 const custom = require('./custom');
 const google = require('./google');
 const anthropic = require('./anthropic');
+const gptPlugins = require('./gptPlugins');
 const { isEnabled } = require('~/server/utils');
 const { EModelEndpoint } = require('librechat-data-provider');
 const {
@@ -38,6 +39,7 @@ if (isEnabled(LIMIT_MESSAGE_USER)) {
 router.use(validateConvoAccess);
 
 router.use([`/${EModelEndpoint.azureOpenAI}`, `/${EModelEndpoint.openAI}`], openAI);
+router.use(`/${EModelEndpoint.gptPlugins}`, gptPlugins);
 router.use(`/${EModelEndpoint.anthropic}`, anthropic);
 router.use(`/${EModelEndpoint.google}`, google);
 router.use(`/${EModelEndpoint.custom}`, custom);

api/server/routes/files/files.js

@@ -283,10 +283,7 @@ router.post('/', async (req, res) => {
       message += ': ' + error.message;
     }
 
-    if (
-      error.message?.includes('Invalid file format') ||
-      error.message?.includes('No OCR result')
-    ) {
+    if (error.message?.includes('Invalid file format')) {
       message = error.message;
     }
 

api/server/routes/index.js

@@ -9,6 +9,7 @@ const presets = require('./presets');
 const prompts = require('./prompts');
 const balance = require('./balance');
 const plugins = require('./plugins');
+const bedrock = require('./bedrock');
 const actions = require('./actions');
 const banner = require('./banner');
 const search = require('./search');
@@ -25,9 +26,13 @@ const auth = require('./auth');
 const edit = require('./edit');
 const keys = require('./keys');
 const user = require('./user');
+const ask = require('./ask');
+const accessPermissions = require('./accessPermissions');
 const mcp = require('./mcp');
 
 module.exports = {
+  ask,
+  mcp,
   edit,
   auth,
   keys,
@@ -43,6 +48,7 @@ module.exports = {
   search,
   config,
   models,
+  bedrock,
   prompts,
   plugins,
   actions,
@@ -55,5 +61,5 @@ module.exports = {
   assistants,
   categories,
   staticRoute,
-  mcp,
+  accessPermissions,
 };

api/server/routes/memories.js

@@ -1,43 +1,37 @@
 const express = require('express');
-const { Tokenizer, generateCheckAccess } = require('@librechat/api');
+const { Tokenizer } = require('@librechat/api');
 const { PermissionTypes, Permissions } = require('librechat-data-provider');
 const {
   getAllUserMemories,
   toggleUserMemories,
   createMemory,
-  deleteMemory,
   setMemory,
+  deleteMemory,
 } = require('~/models');
-const { requireJwtAuth } = require('~/server/middleware');
-const { getRoleByName } = require('~/models/Role');
+const { requireJwtAuth, generateCheckAccess } = require('~/server/middleware');
 
 const router = express.Router();
 
-const checkMemoryRead = generateCheckAccess({
-  permissionType: PermissionTypes.MEMORIES,
-  permissions: [Permissions.USE, Permissions.READ],
-  getRoleByName,
-});
-const checkMemoryCreate = generateCheckAccess({
-  permissionType: PermissionTypes.MEMORIES,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  getRoleByName,
-});
-const checkMemoryUpdate = generateCheckAccess({
-  permissionType: PermissionTypes.MEMORIES,
-  permissions: [Permissions.USE, Permissions.UPDATE],
-  getRoleByName,
-});
-const checkMemoryDelete = generateCheckAccess({
-  permissionType: PermissionTypes.MEMORIES,
-  permissions: [Permissions.USE, Permissions.UPDATE],
-  getRoleByName,
-});
-const checkMemoryOptOut = generateCheckAccess({
-  permissionType: PermissionTypes.MEMORIES,
-  permissions: [Permissions.USE, Permissions.OPT_OUT],
-  getRoleByName,
-});
+const checkMemoryRead = generateCheckAccess(PermissionTypes.MEMORIES, [
+  Permissions.USE,
+  Permissions.READ,
+]);
+const checkMemoryCreate = generateCheckAccess(PermissionTypes.MEMORIES, [
+  Permissions.USE,
+  Permissions.CREATE,
+]);
+const checkMemoryUpdate = generateCheckAccess(PermissionTypes.MEMORIES, [
+  Permissions.USE,
+  Permissions.UPDATE,
+]);
+const checkMemoryDelete = generateCheckAccess(PermissionTypes.MEMORIES, [
+  Permissions.USE,
+  Permissions.UPDATE,
+]);
+const checkMemoryOptOut = generateCheckAccess(PermissionTypes.MEMORIES, [
+  Permissions.USE,
+  Permissions.OPT_OUT,
+]);
 
 router.use(requireJwtAuth);
 

api/server/routes/oauth.js

@@ -9,6 +9,7 @@ const {
   setBalanceConfig,
   checkDomainAllowed,
 } = require('~/server/middleware');
+const { syncUserEntraGroupMemberships } = require('~/server/services/PermissionService');
 const { setAuthTokens, setOpenIDAuthTokens } = require('~/server/services/AuthService');
 const { isEnabled } = require('~/server/utils');
 const { logger } = require('~/config');
@@ -35,6 +36,7 @@ const oauthHandler = async (req, res) => {
       req.user.provider == 'openid' &&
       isEnabled(process.env.OPENID_REUSE_TOKENS) === true
     ) {
+      await syncUserEntraGroupMemberships(req.user, req.user.tokenset.access_token);
       setOpenIDAuthTokens(req.user.tokenset, res);
     } else {
       await setAuthTokens(req.user._id, res);

api/server/routes/prompts.js

@@ -1,7 +1,5 @@
 const express = require('express');
-const { logger } = require('@librechat/data-schemas');
-const { generateCheckAccess } = require('@librechat/api');
-const { Permissions, SystemRoles, PermissionTypes } = require('librechat-data-provider');
+const { PermissionTypes, Permissions, SystemRoles } = require('librechat-data-provider');
 const {
   getPrompt,
   getPrompts,
@@ -16,30 +14,24 @@ const {
   // updatePromptLabels,
   makePromptProduction,
 } = require('~/models/Prompt');
-const { requireJwtAuth } = require('~/server/middleware');
-const { getRoleByName } = require('~/models/Role');
+const { requireJwtAuth, generateCheckAccess } = require('~/server/middleware');
+const { logger } = require('~/config');
 
 const router = express.Router();
 
-const checkPromptAccess = generateCheckAccess({
-  permissionType: PermissionTypes.PROMPTS,
-  permissions: [Permissions.USE],
-  getRoleByName,
-});
-const checkPromptCreate = generateCheckAccess({
-  permissionType: PermissionTypes.PROMPTS,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  getRoleByName,
-});
+const checkPromptAccess = generateCheckAccess(PermissionTypes.PROMPTS, [Permissions.USE]);
+const checkPromptCreate = generateCheckAccess(PermissionTypes.PROMPTS, [
+  Permissions.USE,
+  Permissions.CREATE,
+]);
 
-const checkGlobalPromptShare = generateCheckAccess({
-  permissionType: PermissionTypes.PROMPTS,
-  permissions: [Permissions.USE, Permissions.CREATE],
-  bodyProps: {
+const checkGlobalPromptShare = generateCheckAccess(
+  PermissionTypes.PROMPTS,
+  [Permissions.USE, Permissions.CREATE],
+  {
     [Permissions.SHARED_GLOBAL]: ['projectIds', 'removeProjectIds'],
   },
-  getRoleByName,
-});
+);
 
 router.use(requireJwtAuth);
 router.use(checkPromptAccess);

api/server/routes/tags.js

@@ -1,24 +1,18 @@
 const express = require('express');
-const { logger } = require('@librechat/data-schemas');
-const { generateCheckAccess } = require('@librechat/api');
 const { PermissionTypes, Permissions } = require('librechat-data-provider');
 const {
-  updateTagsForConversation,
+  getConversationTags,
   updateConversationTag,
   createConversationTag,
   deleteConversationTag,
-  getConversationTags,
+  updateTagsForConversation,
 } = require('~/models/ConversationTag');
-const { requireJwtAuth } = require('~/server/middleware');
-const { getRoleByName } = require('~/models/Role');
+const { requireJwtAuth, generateCheckAccess } = require('~/server/middleware');
+const { logger } = require('~/config');
 
 const router = express.Router();
 
-const checkBookmarkAccess = generateCheckAccess({
-  permissionType: PermissionTypes.BOOKMARKS,
-  permissions: [Permissions.USE],
-  getRoleByName,
-});
+const checkBookmarkAccess = generateCheckAccess(PermissionTypes.BOOKMARKS, [Permissions.USE]);
 
 router.use(requireJwtAuth);
 router.use(checkBookmarkAccess);

api/server/services/AppService.interface.spec.js

@@ -1,5 +1,6 @@
 jest.mock('~/models', () => ({
   initializeRoles: jest.fn(),
+  seedDefaultRoles: jest.fn(),
 }));
 jest.mock('~/models/Role', () => ({
   updateAccessPermissions: jest.fn(),

api/server/services/AppService.js

@@ -17,6 +17,7 @@ const {
 const { azureAssistantsDefaults, assistantsConfigSetup } = require('./start/assistants');
 const { initializeAzureBlobService } = require('./Files/Azure/initialize');
 const { initializeFirebase } = require('./Files/Firebase/initialize');
+const { seedDefaultRoles, initializeRoles } = require('~/models');
 const loadCustomConfig = require('./Config/loadCustomConfig');
 const handleRateLimits = require('./Config/handleRateLimits');
 const { loadDefaultInterface } = require('./start/interface');
@@ -26,7 +27,6 @@ const { processModelSpecs } = require('./start/modelSpecs');
 const { initializeS3 } = require('./Files/S3/initialize');
 const { loadAndFormatTools } = require('./ToolService');
 const { isEnabled } = require('~/server/utils');
-const { initializeRoles } = require('~/models');
 const { setCachedTools } = require('./Config');
 const paths = require('~/config/paths');
 
@@ -37,6 +37,7 @@ const paths = require('~/config/paths');
  */
 const AppService = async (app) => {
   await initializeRoles();
+  await seedDefaultRoles();
   /** @type {TCustomConfig} */
   const config = (await loadCustomConfig()) ?? {};
   const configDefaults = getConfigDefaults();

api/server/services/AppService.spec.js

@@ -28,6 +28,7 @@ jest.mock('./Files/Firebase/initialize', () => ({
 }));
 jest.mock('~/models', () => ({
   initializeRoles: jest.fn(),
+  seedDefaultRoles: jest.fn(),
 }));
 jest.mock('~/models/Role', () => ({
   updateAccessPermissions: jest.fn(),

api/server/services/AssistantService.js

@@ -1,7 +1,4 @@
 const { klona } = require('klona');
-const { sleep } = require('@librechat/agents');
-const { sendEvent } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const {
   StepTypes,
   RunStatus,
@@ -14,10 +11,11 @@ const {
 } = require('librechat-data-provider');
 const { retrieveAndProcessFile } = require('~/server/services/Files/process');
 const { processRequiredActions } = require('~/server/services/ToolService');
+const { createOnProgress, sendMessage, sleep } = require('~/server/utils');
 const { RunManager, waitForRun } = require('~/server/services/Runs');
 const { processMessages } = require('~/server/services/Threads');
-const { createOnProgress } = require('~/server/utils');
 const { TextStream } = require('~/app/clients');
+const { logger } = require('~/config');
 
 /**
  * Sorts, processes, and flattens messages to a single string.
@@ -66,7 +64,7 @@ async function createOnTextProgress({
     };
 
     logger.debug('Content data:', contentData);
-    sendEvent(openai.res, contentData);
+    sendMessage(openai.res, contentData);
   };
 }
 

api/server/services/AuthService.js

@@ -1,5 +1,4 @@
 const bcrypt = require('bcryptjs');
-const jwt = require('jsonwebtoken');
 const { webcrypto } = require('node:crypto');
 const { isEnabled } = require('@librechat/api');
 const { logger } = require('@librechat/data-schemas');
@@ -500,18 +499,6 @@ const resendVerificationEmail = async (req) => {
     };
   }
 };
-/**
- * Generate a short-lived JWT token
- * @param {String} userId - The ID of the user
- * @param {String} [expireIn='5m'] - The expiration time for the token (default is 5 minutes)
- * @returns {String} - The generated JWT token
- */
-const generateShortLivedToken = (userId, expireIn = '5m') => {
-  return jwt.sign({ id: userId }, process.env.JWT_SECRET, {
-    expiresIn: expireIn,
-    algorithm: 'HS256',
-  });
-};
 
 module.exports = {
   logoutUser,
@@ -519,8 +506,7 @@ module.exports = {
   registerUser,
   setAuthTokens,
   resetPassword,
-  setOpenIDAuthTokens,
   requestPasswordReset,
   resendVerificationEmail,
-  generateShortLivedToken,
+  setOpenIDAuthTokens,
 };

api/server/services/Config/EndpointService.js

@@ -1,6 +1,5 @@
-const { isUserProvided } = require('@librechat/api');
 const { EModelEndpoint } = require('librechat-data-provider');
-const { generateConfig } = require('~/server/utils/handleText');
+const { isUserProvided, generateConfig } = require('~/server/utils');
 
 const {
   OPENAI_API_KEY: openAIApiKey,

api/server/services/Config/getCustomConfig.js

@@ -40,7 +40,6 @@ async function getBalanceConfig() {
 /**
  *
  * @param {string | EModelEndpoint} endpoint
- * @returns {Promise<TEndpoint | undefined>}
  */
 const getCustomEndpointConfig = async (endpoint) => {
   const customConfig = await getCustomConfig();

api/server/services/Config/loadCustomConfig.js

@@ -1,18 +1,18 @@
 const path = require('path');
-const axios = require('axios');
-const yaml = require('js-yaml');
-const keyBy = require('lodash/keyBy');
-const { loadYaml } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const {
   CacheKeys,
   configSchema,
-  paramSettings,
   EImageOutputType,
-  agentParamSettings,
   validateSettingDefinitions,
+  agentParamSettings,
+  paramSettings,
 } = require('librechat-data-provider');
 const getLogStores = require('~/cache/getLogStores');
+const loadYaml = require('~/utils/loadYaml');
+const { logger } = require('~/config');
+const axios = require('axios');
+const yaml = require('js-yaml');
+const keyBy = require('lodash/keyBy');
 
 const projectRoot = path.resolve(__dirname, '..', '..', '..', '..');
 const defaultConfigPath = path.resolve(projectRoot, 'librechat.yaml');

api/server/services/Config/loadCustomConfig.spec.js

@@ -1,9 +1,6 @@
 jest.mock('axios');
 jest.mock('~/cache/getLogStores');
-jest.mock('@librechat/api', () => ({
-  ...jest.requireActual('@librechat/api'),
-  loadYaml: jest.fn(),
-}));
+jest.mock('~/utils/loadYaml');
 jest.mock('librechat-data-provider', () => {
   const actual = jest.requireActual('librechat-data-provider');
   return {
@@ -33,22 +30,11 @@ jest.mock('librechat-data-provider', () => {
   };
 });
 
-jest.mock('@librechat/data-schemas', () => {
-  return {
-    logger: {
-      info: jest.fn(),
-      warn: jest.fn(),
-      debug: jest.fn(),
-      error: jest.fn(),
-    },
-  };
-});
-
 const axios = require('axios');
-const { loadYaml } = require('@librechat/api');
-const { logger } = require('@librechat/data-schemas');
 const loadCustomConfig = require('./loadCustomConfig');
 const getLogStores = require('~/cache/getLogStores');
+const loadYaml = require('~/utils/loadYaml');
+const { logger } = require('~/config');
 
 describe('loadCustomConfig', () => {
   const mockSet = jest.fn();

api/server/services/Endpoints/agents/agent.js

@@ -1,9 +1,5 @@
 const { Providers } = require('@librechat/agents');
-const {
-  primeResources,
-  extractLibreChatParams,
-  optionalChainWithEmptyCheck,
-} = require('@librechat/api');
+const { primeResources, optionalChainWithEmptyCheck } = require('@librechat/api');
 const {
   ErrorTypes,
   EModelEndpoint,
@@ -11,12 +7,30 @@ const {
   replaceSpecialVars,
   providerEndpointMap,
 } = require('librechat-data-provider');
-const { getProviderConfig } = require('~/server/services/Endpoints');
+const initAnthropic = require('~/server/services/Endpoints/anthropic/initialize');
+const getBedrockOptions = require('~/server/services/Endpoints/bedrock/options');
+const initOpenAI = require('~/server/services/Endpoints/openAI/initialize');
+const initCustom = require('~/server/services/Endpoints/custom/initialize');
+const initGoogle = require('~/server/services/Endpoints/google/initialize');
 const generateArtifactsPrompt = require('~/app/clients/prompts/artifacts');
+const { getCustomEndpointConfig } = require('~/server/services/Config');
 const { processFiles } = require('~/server/services/Files/process');
-const { getFiles, getToolFilesByIds } = require('~/models/File');
 const { getConvoFiles } = require('~/models/Conversation');
+const { getToolFilesByIds } = require('~/models/File');
 const { getModelMaxTokens } = require('~/utils');
+const { getFiles } = require('~/models/File');
+
+const providerConfigMap = {
+  [Providers.XAI]: initCustom,
+  [Providers.OLLAMA]: initCustom,
+  [Providers.DEEPSEEK]: initCustom,
+  [Providers.OPENROUTER]: initCustom,
+  [EModelEndpoint.openAI]: initOpenAI,
+  [EModelEndpoint.google]: initGoogle,
+  [EModelEndpoint.azureOpenAI]: initOpenAI,
+  [EModelEndpoint.anthropic]: initAnthropic,
+  [EModelEndpoint.bedrock]: getBedrockOptions,
+};
 
 /**
  * @param {object} params
@@ -57,7 +71,7 @@ const initializeAgent = async ({
     ),
   );
 
-  const { resendFiles, maxContextTokens, modelOptions } = extractLibreChatParams(_modelOptions);
+  const { resendFiles = true, ...modelOptions } = _modelOptions;
 
   if (isInitialAgent && conversationId != null && resendFiles) {
     const fileIds = (await getConvoFiles(conversationId)) ?? [];
@@ -97,9 +111,17 @@ const initializeAgent = async ({
     })) ?? {};
 
   agent.endpoint = provider;
-  const { getOptions, overrideProvider } = await getProviderConfig(provider);
-  if (overrideProvider) {
-    agent.provider = overrideProvider;
+  let getOptions = providerConfigMap[provider];
+  if (!getOptions && providerConfigMap[provider.toLowerCase()] != null) {
+    agent.provider = provider.toLowerCase();
+    getOptions = providerConfigMap[agent.provider];
+  } else if (!getOptions) {
+    const customEndpointConfig = await getCustomEndpointConfig(provider);
+    if (!customEndpointConfig) {
+      throw new Error(`Provider ${provider} not supported`);
+    }
+    getOptions = initCustom;
+    agent.provider = Providers.OPENAI;
   }
 
   const _endpointOption =
@@ -123,8 +145,9 @@ const initializeAgent = async ({
     modelOptions.maxTokens,
     0,
   );
-  const agentMaxContextTokens = optionalChainWithEmptyCheck(
-    maxContextTokens,
+  const maxContextTokens = optionalChainWithEmptyCheck(
+    modelOptions.maxContextTokens,
+    modelOptions.max_context_tokens,
     getModelMaxTokens(tokensModel, providerEndpointMap[provider]),
     4096,
   );
@@ -166,7 +189,7 @@ const initializeAgent = async ({
     attachments,
     resendFiles,
     toolContextMap,
-    maxContextTokens: (agentMaxContextTokens - maxTokens) * 0.9,
+    maxContextTokens: (maxContextTokens - maxTokens) * 0.9,
   };
 };
 

api/server/services/Endpoints/agents/build.js

@@ -1,9 +1,10 @@
-const { isAgentsEndpoint, removeNullishValues, Constants } = require('librechat-data-provider');
+const { isAgentsEndpoint, Constants } = require('librechat-data-provider');
 const { loadAgent } = require('~/models/Agent');
 const { logger } = require('~/config');
 
 const buildOptions = (req, endpoint, parsedBody, endpointType) => {
-  const { spec, iconURL, agent_id, instructions, ...model_parameters } = parsedBody;
+  const { spec, iconURL, agent_id, instructions, maxContextTokens, ...model_parameters } =
+    parsedBody;
   const agentPromise = loadAgent({
     req,
     agent_id: isAgentsEndpoint(endpoint) ? agent_id : Constants.EPHEMERAL_AGENT_ID,
@@ -14,16 +15,19 @@ const buildOptions = (req, endpoint, parsedBody, endpointType) => {
     return undefined;
   });
 
-  return removeNullishValues({
+  const endpointOption = {
     spec,
     iconURL,
     endpoint,
     agent_id,
     endpointType,
     instructions,
+    maxContextTokens,
     model_parameters,
     agent: agentPromise,
-  });
+  };
+
+  return endpointOption;
 };
 
 module.exports = { buildOptions };

api/server/services/Endpoints/agents/initialize.js

@@ -1,17 +1,11 @@
 const { logger } = require('@librechat/data-schemas');
 const { createContentAggregator } = require('@librechat/agents');
+const { Constants, EModelEndpoint, getResponseSender } = require('librechat-data-provider');
 const {
-  Constants,
-  EModelEndpoint,
-  isAgentsEndpoint,
-  getResponseSender,
-} = require('librechat-data-provider');
-const {
-  createToolEndCallback,
   getDefaultHandlers,
+  createToolEndCallback,
 } = require('~/server/controllers/agents/callbacks');
 const { initializeAgent } = require('~/server/services/Endpoints/agents/agent');
-const { getCustomEndpointConfig } = require('~/server/services/Config');
 const { loadAgentTools } = require('~/server/services/ToolService');
 const AgentClient = require('~/server/controllers/agents/client');
 const { getAgent } = require('~/models/Agent');
@@ -67,7 +61,6 @@ const initializeClient = async ({ req, res, endpointOption }) => {
   }
 
   const primaryAgent = await endpointOption.agent;
-  delete endpointOption.agent;
   if (!primaryAgent) {
     throw new Error('Agent not found');
   }
@@ -115,25 +108,11 @@ const initializeClient = async ({ req, res, endpointOption }) => {
     }
   }
 
-  let endpointConfig = req.app.locals[primaryConfig.endpoint];
-  if (!isAgentsEndpoint(primaryConfig.endpoint) && !endpointConfig) {
-    try {
-      endpointConfig = await getCustomEndpointConfig(primaryConfig.endpoint);
-    } catch (err) {
-      logger.error(
-        '[api/server/controllers/agents/client.js #titleConvo] Error getting custom endpoint config',
-        err,
-      );
-    }
-  }
-
   const sender =
     primaryAgent.name ??
     getResponseSender({
       ...endpointOption,
       model: endpointOption.model_parameters.model,
-      modelDisplayLabel: endpointConfig?.modelDisplayLabel,
-      modelLabel: endpointOption.model_parameters.modelLabel,
     });
 
   const client = new AgentClient({

api/server/services/Endpoints/agents/title.js

@@ -23,7 +23,7 @@ const addTitle = async (req, { text, response, client }) => {
   let timeoutId;
   try {
     const timeoutPromise = new Promise((_, reject) => {
-      timeoutId = setTimeout(() => reject(new Error('Title generation timeout')), 45000);
+      timeoutId = setTimeout(() => reject(new Error('Title generation timeout')), 25000);
     }).catch((error) => {
       logger.error('Title error:', error);
     });

api/server/services/Endpoints/anthropic/initialize.js

@@ -41,7 +41,7 @@ const initializeClient = async ({ req, res, endpointOption, overrideModel, optio
       {
         reverseProxyUrl: ANTHROPIC_REVERSE_PROXY ?? null,
         proxy: PROXY ?? null,
-        modelOptions: endpointOption?.model_parameters ?? {},
+        modelOptions: endpointOption.model_parameters,
       },
       clientOptions,
     );

api/server/services/Endpoints/anthropic/llm.js

@@ -75,7 +75,6 @@ function getLLMConfig(apiKey, options = {}) {
 
   if (options.reverseProxyUrl) {
     requestOptions.clientOptions.baseURL = options.reverseProxyUrl;
-    requestOptions.anthropicApiUrl = options.reverseProxyUrl;
   }
 
   return {

api/server/services/Endpoints/anthropic/llm.spec.js

@@ -1,45 +1,11 @@
-const { anthropicSettings, removeNullishValues } = require('librechat-data-provider');
+const { anthropicSettings } = require('librechat-data-provider');
 const { getLLMConfig } = require('~/server/services/Endpoints/anthropic/llm');
-const { checkPromptCacheSupport, getClaudeHeaders, configureReasoning } = require('./helpers');
 
 jest.mock('https-proxy-agent', () => ({
   HttpsProxyAgent: jest.fn().mockImplementation((proxy) => ({ proxy })),
 }));
 
-jest.mock('./helpers', () => ({
-  checkPromptCacheSupport: jest.fn(),
-  getClaudeHeaders: jest.fn(),
-  configureReasoning: jest.fn((requestOptions) => requestOptions),
-}));
-
-jest.mock('librechat-data-provider', () => ({
-  anthropicSettings: {
-    model: { default: 'claude-3-opus-20240229' },
-    maxOutputTokens: { default: 4096, reset: jest.fn(() => 4096) },
-    thinking: { default: false },
-    promptCache: { default: false },
-    thinkingBudget: { default: null },
-  },
-  removeNullishValues: jest.fn((obj) => {
-    const result = {};
-    for (const key in obj) {
-      if (obj[key] !== null && obj[key] !== undefined) {
-        result[key] = obj[key];
-      }
-    }
-    return result;
-  }),
-}));
-
 describe('getLLMConfig', () => {
-  beforeEach(() => {
-    jest.clearAllMocks();
-    checkPromptCacheSupport.mockReturnValue(false);
-    getClaudeHeaders.mockReturnValue(undefined);
-    configureReasoning.mockImplementation((requestOptions) => requestOptions);
-    anthropicSettings.maxOutputTokens.reset.mockReturnValue(4096);
-  });
-
   it('should create a basic configuration with default values', () => {
     const result = getLLMConfig('test-api-key', { modelOptions: {} });
 
@@ -70,7 +36,6 @@ describe('getLLMConfig', () => {
     });
 
     expect(result.llmConfig.clientOptions).toHaveProperty('baseURL', 'http://reverse-proxy');
-    expect(result.llmConfig).toHaveProperty('anthropicApiUrl', 'http://reverse-proxy');
   });
 
   it('should include topK and topP for non-Claude-3.7 models', () => {
@@ -100,11 +65,6 @@ describe('getLLMConfig', () => {
   });
 
   it('should NOT include topK and topP for Claude-3-7 models (hyphen notation)', () => {
-    configureReasoning.mockImplementation((requestOptions) => {
-      requestOptions.thinking = { type: 'enabled' };
-      return requestOptions;
-    });
-
     const result = getLLMConfig('test-api-key', {
       modelOptions: {
         model: 'claude-3-7-sonnet',
@@ -118,11 +78,6 @@ describe('getLLMConfig', () => {
   });
 
   it('should NOT include topK and topP for Claude-3.7 models (decimal notation)', () => {
-    configureReasoning.mockImplementation((requestOptions) => {
-      requestOptions.thinking = { type: 'enabled' };
-      return requestOptions;
-    });
-
     const result = getLLMConfig('test-api-key', {
       modelOptions: {
         model: 'claude-3.7-sonnet',
@@ -199,160 +154,4 @@ describe('getLLMConfig', () => {
     expect(result3.llmConfig).toHaveProperty('topK', 10);
     expect(result3.llmConfig).toHaveProperty('topP', 0.9);
   });
-
-  describe('Edge cases', () => {
-    it('should handle missing apiKey', () => {
-      const result = getLLMConfig(undefined, { modelOptions: {} });
-      expect(result.llmConfig).not.toHaveProperty('apiKey');
-    });
-
-    it('should handle empty modelOptions', () => {
-      expect(() => {
-        getLLMConfig('test-api-key', {});
-      }).toThrow("Cannot read properties of undefined (reading 'thinking')");
-    });
-
-    it('should handle no options parameter', () => {
-      expect(() => {
-        getLLMConfig('test-api-key');
-      }).toThrow("Cannot read properties of undefined (reading 'thinking')");
-    });
-
-    it('should handle temperature, stop sequences, and stream settings', () => {
-      const result = getLLMConfig('test-api-key', {
-        modelOptions: {
-          temperature: 0.7,
-          stop: ['\n\n', 'END'],
-          stream: false,
-        },
-      });
-
-      expect(result.llmConfig).toHaveProperty('temperature', 0.7);
-      expect(result.llmConfig).toHaveProperty('stopSequences', ['\n\n', 'END']);
-      expect(result.llmConfig).toHaveProperty('stream', false);
-    });
-
-    it('should handle maxOutputTokens when explicitly set to falsy value', () => {
-      anthropicSettings.maxOutputTokens.reset.mockReturnValue(8192);
-      const result = getLLMConfig('test-api-key', {
-        modelOptions: {
-          model: 'claude-3-opus',
-          maxOutputTokens: null,
-        },
-      });
-
-      expect(anthropicSettings.maxOutputTokens.reset).toHaveBeenCalledWith('claude-3-opus');
-      expect(result.llmConfig).toHaveProperty('maxTokens', 8192);
-    });
-
-    it('should handle both proxy and reverseProxyUrl', () => {
-      const result = getLLMConfig('test-api-key', {
-        modelOptions: {},
-        proxy: 'http://proxy:8080',
-        reverseProxyUrl: 'https://reverse-proxy.com',
-      });
-
-      expect(result.llmConfig.clientOptions).toHaveProperty('fetchOptions');
-      expect(result.llmConfig.clientOptions.fetchOptions).toHaveProperty('dispatcher');
-      expect(result.llmConfig.clientOptions.fetchOptions.dispatcher).toBeDefined();
-      expect(result.llmConfig.clientOptions.fetchOptions.dispatcher.constructor.name).toBe(
-        'ProxyAgent',
-      );
-      expect(result.llmConfig.clientOptions).toHaveProperty('baseURL', 'https://reverse-proxy.com');
-      expect(result.llmConfig).toHaveProperty('anthropicApiUrl', 'https://reverse-proxy.com');
-    });
-
-    it('should handle prompt cache with supported model', () => {
-      checkPromptCacheSupport.mockReturnValue(true);
-      getClaudeHeaders.mockReturnValue({ 'anthropic-beta': 'prompt-caching-2024-07-31' });
-
-      const result = getLLMConfig('test-api-key', {
-        modelOptions: {
-          model: 'claude-3-5-sonnet',
-          promptCache: true,
-        },
-      });
-
-      expect(checkPromptCacheSupport).toHaveBeenCalledWith('claude-3-5-sonnet');
-      expect(getClaudeHeaders).toHaveBeenCalledWith('claude-3-5-sonnet', true);
-      expect(result.llmConfig.clientOptions.defaultHeaders).toEqual({
-        'anthropic-beta': 'prompt-caching-2024-07-31',
-      });
-    });
-
-    it('should handle thinking and thinkingBudget options', () => {
-      configureReasoning.mockImplementation((requestOptions, systemOptions) => {
-        if (systemOptions.thinking) {
-          requestOptions.thinking = { type: 'enabled' };
-        }
-        if (systemOptions.thinkingBudget) {
-          requestOptions.thinking = {
-            ...requestOptions.thinking,
-            budget_tokens: systemOptions.thinkingBudget,
-          };
-        }
-        return requestOptions;
-      });
-
-      getLLMConfig('test-api-key', {
-        modelOptions: {
-          model: 'claude-3-7-sonnet',
-          thinking: true,
-          thinkingBudget: 5000,
-        },
-      });
-
-      expect(configureReasoning).toHaveBeenCalledWith(
-        expect.any(Object),
-        expect.objectContaining({
-          thinking: true,
-          promptCache: false,
-          thinkingBudget: 5000,
-        }),
-      );
-    });
-
-    it('should remove system options from modelOptions', () => {
-      const modelOptions = {
-        model: 'claude-3-opus',
-        thinking: true,
-        promptCache: true,
-        thinkingBudget: 1000,
-        temperature: 0.5,
-      };
-
-      getLLMConfig('test-api-key', { modelOptions });
-
-      expect(modelOptions).not.toHaveProperty('thinking');
-      expect(modelOptions).not.toHaveProperty('promptCache');
-      expect(modelOptions).not.toHaveProperty('thinkingBudget');
-      expect(modelOptions).toHaveProperty('temperature', 0.5);
-    });
-
-    it('should handle all nullish values removal', () => {
-      removeNullishValues.mockImplementation((obj) => {
-        const cleaned = {};
-        Object.entries(obj).forEach(([key, value]) => {
-          if (value !== null && value !== undefined) {
-            cleaned[key] = value;
-          }
-        });
-        return cleaned;
-      });
-
-      const result = getLLMConfig('test-api-key', {
-        modelOptions: {
-          temperature: null,
-          topP: undefined,
-          topK: 0,
-          stop: [],
-        },
-      });
-
-      expect(result.llmConfig).not.toHaveProperty('temperature');
-      expect(result.llmConfig).not.toHaveProperty('topP');
-      expect(result.llmConfig).toHaveProperty('topK', 0);
-      expect(result.llmConfig).toHaveProperty('stopSequences', []);
-    });
-  });
 });

api/server/services/Endpoints/azureAssistants/initialize.js

@@ -1,7 +1,12 @@
 const OpenAI = require('openai');
 const { HttpsProxyAgent } = require('https-proxy-agent');
-const { constructAzureURL, isUserProvided, resolveHeaders } = require('@librechat/api');
-const { ErrorTypes, EModelEndpoint, mapModelToAzureConfig } = require('librechat-data-provider');
+const { constructAzureURL, isUserProvided } = require('@librechat/api');
+const {
+  ErrorTypes,
+  EModelEndpoint,
+  resolveHeaders,
+  mapModelToAzureConfig,
+} = require('librechat-data-provider');
 const {
   getUserKeyValues,
   getUserKeyExpiry,
@@ -109,14 +114,11 @@ const initializeClient = async ({ req, res, version, endpointOption, initAppClie
 
     apiKey = azureOptions.azureOpenAIApiKey;
     opts.defaultQuery = { 'api-version': azureOptions.azureOpenAIApiVersion };
-    opts.defaultHeaders = resolveHeaders(
-      {
-        ...headers,
-        'api-key': apiKey,
-        'OpenAI-Beta': `assistants=${version}`,
-      },
-      req.user,
-    );
+    opts.defaultHeaders = resolveHeaders({
+      ...headers,
+      'api-key': apiKey,
+      'OpenAI-Beta': `assistants=${version}`,
+    });
     opts.model = azureOptions.azureOpenAIApiDeploymentName;
 
     if (initAppClient) {

api/server/services/Endpoints/bedrock/options.js

@@ -64,7 +64,7 @@ const getOptions = async ({ req, overrideModel, endpointOption }) => {
 
   /** @type {BedrockClientOptions} */
   const requestOptions = {
-    model: overrideModel ?? endpointOption?.model,
+    model: overrideModel ?? endpointOption.model,
     region: BEDROCK_AWS_DEFAULT_REGION,
   };
 
@@ -76,7 +76,7 @@ const getOptions = async ({ req, overrideModel, endpointOption }) => {
 
   const llmConfig = bedrockOutputParser(
     bedrockInputParser.parse(
-      removeNullishValues(Object.assign(requestOptions, endpointOption?.model_parameters ?? {})),
+      removeNullishValues(Object.assign(requestOptions, endpointOption.model_parameters)),
     ),
   );
 

api/server/services/Endpoints/custom/initialize.js

@@ -6,7 +6,7 @@ const {
   extractEnvVariable,
 } = require('librechat-data-provider');
 const { Providers } = require('@librechat/agents');
-const { getOpenAIConfig, createHandleLLMNewToken, resolveHeaders } = require('@librechat/api');
+const { getOpenAIConfig, createHandleLLMNewToken } = require('@librechat/api');
 const { getUserKeyValues, checkUserKeyExpiry } = require('~/server/services/UserService');
 const { getCustomEndpointConfig } = require('~/server/services/Config');
 const { fetchModels } = require('~/server/services/ModelService');
@@ -28,7 +28,12 @@ const initializeClient = async ({ req, res, endpointOption, optionsOnly, overrid
   const CUSTOM_API_KEY = extractEnvVariable(endpointConfig.apiKey);
   const CUSTOM_BASE_URL = extractEnvVariable(endpointConfig.baseURL);
 
-  let resolvedHeaders = resolveHeaders(endpointConfig.headers, req.user);
+  let resolvedHeaders = {};
+  if (endpointConfig.headers && typeof endpointConfig.headers === 'object') {
+    Object.keys(endpointConfig.headers).forEach((key) => {
+      resolvedHeaders[key] = extractEnvVariable(endpointConfig.headers[key]);
+    });
+  }
 
   if (CUSTOM_API_KEY.match(envVarRegex)) {
     throw new Error(`Missing API Key for ${endpoint}.`);
@@ -129,7 +134,7 @@ const initializeClient = async ({ req, res, endpointOption, optionsOnly, overrid
   };
 
   if (optionsOnly) {
-    const modelOptions = endpointOption?.model_parameters ?? {};
+    const modelOptions = endpointOption.model_parameters;
     if (endpoint !== Providers.OLLAMA) {
       clientOptions = Object.assign(
         {

api/server/services/Endpoints/custom/initialize.spec.js

@@ -1,93 +0,0 @@
-const initializeClient = require('./initialize');
-
-jest.mock('@librechat/api', () => ({
-  resolveHeaders: jest.fn(),
-  getOpenAIConfig: jest.fn(),
-  createHandleLLMNewToken: jest.fn(),
-}));
-
-jest.mock('librechat-data-provider', () => ({
-  CacheKeys: { TOKEN_CONFIG: 'token_config' },
-  ErrorTypes: { NO_USER_KEY: 'NO_USER_KEY', NO_BASE_URL: 'NO_BASE_URL' },
-  envVarRegex: /\$\{([^}]+)\}/,
-  FetchTokenConfig: {},
-  extractEnvVariable: jest.fn((value) => value),
-}));
-
-jest.mock('@librechat/agents', () => ({
-  Providers: { OLLAMA: 'ollama' },
-}));
-
-jest.mock('~/server/services/UserService', () => ({
-  getUserKeyValues: jest.fn(),
-  checkUserKeyExpiry: jest.fn(),
-}));
-
-jest.mock('~/server/services/Config', () => ({
-  getCustomEndpointConfig: jest.fn().mockResolvedValue({
-    apiKey: 'test-key',
-    baseURL: 'https://test.com',
-    headers: { 'x-user': '{{LIBRECHAT_USER_ID}}', 'x-email': '{{LIBRECHAT_USER_EMAIL}}' },
-    models: { default: ['test-model'] },
-  }),
-}));
-
-jest.mock('~/server/services/ModelService', () => ({
-  fetchModels: jest.fn(),
-}));
-
-jest.mock('~/app/clients/OpenAIClient', () => {
-  return jest.fn().mockImplementation(() => ({
-    options: {},
-  }));
-});
-
-jest.mock('~/server/utils', () => ({
-  isUserProvided: jest.fn().mockReturnValue(false),
-}));
-
-jest.mock('~/cache/getLogStores', () =>
-  jest.fn().mockReturnValue({
-    get: jest.fn(),
-  }),
-);
-
-describe('custom/initializeClient', () => {
-  const mockRequest = {
-    body: { endpoint: 'test-endpoint' },
-    user: { id: 'user-123', email: 'test@example.com' },
-    app: { locals: {} },
-  };
-  const mockResponse = {};
-
-  beforeEach(() => {
-    jest.clearAllMocks();
-  });
-
-  it('calls resolveHeaders with headers and user', async () => {
-    const { resolveHeaders } = require('@librechat/api');
-    await initializeClient({ req: mockRequest, res: mockResponse, optionsOnly: true });
-    expect(resolveHeaders).toHaveBeenCalledWith(
-      { 'x-user': '{{LIBRECHAT_USER_ID}}', 'x-email': '{{LIBRECHAT_USER_EMAIL}}' },
-      { id: 'user-123', email: 'test@example.com' },
-    );
-  });
-
-  it('throws if endpoint config is missing', async () => {
-    const { getCustomEndpointConfig } = require('~/server/services/Config');
-    getCustomEndpointConfig.mockResolvedValueOnce(null);
-    await expect(
-      initializeClient({ req: mockRequest, res: mockResponse, optionsOnly: true }),
-    ).rejects.toThrow('Config not found for the test-endpoint custom endpoint.');
-  });
-
-  it('throws if user is missing', async () => {
-    await expect(
-      initializeClient({
-        req: { ...mockRequest, user: undefined },
-        res: mockResponse,
-        optionsOnly: true,
-      }),
-    ).rejects.toThrow("Cannot read properties of undefined (reading 'id')");
-  });
-});

api/server/services/Endpoints/google/initialize.js

@@ -1,6 +1,7 @@
-const { getGoogleConfig, isEnabled } = require('@librechat/api');
 const { EModelEndpoint, AuthKeys } = require('librechat-data-provider');
 const { getUserKey, checkUserKeyExpiry } = require('~/server/services/UserService');
+const { getLLMConfig } = require('~/server/services/Endpoints/google/llm');
+const { isEnabled } = require('~/server/utils');
 const { GoogleClient } = require('~/app');
 
 const initializeClient = async ({ req, res, endpointOption, overrideModel, optionsOnly }) => {
@@ -17,7 +18,7 @@ const initializeClient = async ({ req, res, endpointOption, overrideModel, optio
   let serviceKey = {};
   try {
     serviceKey = require('~/data/auth.json');
-  } catch (_e) {
+  } catch (e) {
     // Do nothing
   }
 
@@ -57,14 +58,14 @@ const initializeClient = async ({ req, res, endpointOption, overrideModel, optio
   if (optionsOnly) {
     clientOptions = Object.assign(
       {
-        modelOptions: endpointOption?.model_parameters ?? {},
+        modelOptions: endpointOption.model_parameters,
       },
       clientOptions,
     );
     if (overrideModel) {
       clientOptions.modelOptions.model = overrideModel;
     }
-    return getGoogleConfig(credentials, clientOptions);
+    return getLLMConfig(credentials, clientOptions);
   }
 
   const client = new GoogleClient(credentials, clientOptions);

api/server/services/Endpoints/google/llm.js

@@ -1,15 +1,13 @@
-import { Providers } from '@librechat/agents';
-import { googleSettings, AuthKeys } from 'librechat-data-provider';
-import type { GoogleClientOptions, VertexAIClientOptions } from '@librechat/agents';
-import type * as t from '~/types';
-import { isEnabled } from '~/utils';
+const { Providers } = require('@librechat/agents');
+const { AuthKeys } = require('librechat-data-provider');
+const { isEnabled } = require('~/server/utils');
 
-function getThresholdMapping(model: string) {
+function getThresholdMapping(model) {
   const gemini1Pattern = /gemini-(1\.0|1\.5|pro$|1\.0-pro|1\.5-pro|1\.5-flash-001)/;
   const restrictedPattern = /(gemini-(1\.5-flash-8b|2\.0|exp)|learnlm)/;
 
   if (gemini1Pattern.test(model)) {
-    return (value: string) => {
+    return (value) => {
       if (value === 'OFF') {
         return 'BLOCK_NONE';
       }
@@ -18,7 +16,7 @@ function getThresholdMapping(model: string) {
   }
 
   if (restrictedPattern.test(model)) {
-    return (value: string) => {
+    return (value) => {
       if (value === 'OFF' || value === 'HARM_BLOCK_THRESHOLD_UNSPECIFIED') {
         return 'BLOCK_NONE';
       }
@@ -26,16 +24,19 @@ function getThresholdMapping(model: string) {
     };
   }
 
-  return (value: string) => value;
+  return (value) => value;
 }
 
-export function getSafetySettings(
-  model?: string,
-): Array<{ category: string; threshold: string }> | undefined {
+/**
+ *
+ * @param {string} model
+ * @returns {Array<{category: string, threshold: string}> | undefined}
+ */
+function getSafetySettings(model) {
   if (isEnabled(process.env.GOOGLE_EXCLUDE_SAFETY_SETTINGS)) {
     return undefined;
   }
-  const mapThreshold = getThresholdMapping(model ?? '');
+  const mapThreshold = getThresholdMapping(model);
 
   return [
     {
@@ -73,27 +74,24 @@ export function getSafetySettings(
  * Replicates core logic from GoogleClient's constructor and setOptions, plus client determination.
  * Returns an object with the provider label and the final options that would be passed to createLLM.
  *
- * @param credentials - Either a JSON string or an object containing Google keys
- * @param options - The same shape as the "GoogleClient" constructor options
+ * @param {string | object} credentials - Either a JSON string or an object containing Google keys
+ * @param {object} [options={}]         - The same shape as the "GoogleClient" constructor options
  */
 
-export function getGoogleConfig(
-  credentials: string | t.GoogleCredentials | undefined,
-  options: t.GoogleConfigOptions = {},
-) {
-  let creds: t.GoogleCredentials = {};
+function getLLMConfig(credentials, options = {}) {
+  // 1. Parse credentials
+  let creds = {};
   if (typeof credentials === 'string') {
     try {
       creds = JSON.parse(credentials);
-    } catch (err: unknown) {
-      throw new Error(
-        `Error parsing string credentials: ${err instanceof Error ? err.message : 'Unknown error'}`,
-      );
+    } catch (err) {
+      throw new Error(`Error parsing string credentials: ${err.message}`);
     }
   } else if (credentials && typeof credentials === 'object') {
     creds = credentials;
   }
 
+  // Extract from credentials
   const serviceKeyRaw = creds[AuthKeys.GOOGLE_SERVICE_KEY] ?? {};
   const serviceKey =
     typeof serviceKeyRaw === 'string' ? JSON.parse(serviceKeyRaw) : (serviceKeyRaw ?? {});
@@ -104,15 +102,9 @@ export function getGoogleConfig(
   const reverseProxyUrl = options.reverseProxyUrl;
   const authHeader = options.authHeader;
 
-  const {
-    thinking = googleSettings.thinking.default,
-    thinkingBudget = googleSettings.thinkingBudget.default,
-    ...modelOptions
-  } = options.modelOptions || {};
-
-  const llmConfig: GoogleClientOptions | VertexAIClientOptions = {
-    ...(modelOptions || {}),
-    model: modelOptions?.model ?? '',
+  /** @type {GoogleClientOptions | VertexAIClientOptions} */
+  let llmConfig = {
+    ...(options.modelOptions || {}),
     maxRetries: 2,
   };
 
@@ -129,30 +121,16 @@ export function getGoogleConfig(
 
   // If we have a GCP project => Vertex AI
   if (project_id && provider === Providers.VERTEXAI) {
-    (llmConfig as VertexAIClientOptions).authOptions = {
+    /** @type {VertexAIClientOptions['authOptions']} */
+    llmConfig.authOptions = {
       credentials: { ...serviceKey },
       projectId: project_id,
     };
-    (llmConfig as VertexAIClientOptions).location = process.env.GOOGLE_LOC || 'us-central1';
+    llmConfig.location = process.env.GOOGLE_LOC || 'us-central1';
   } else if (apiKey && provider === Providers.GOOGLE) {
     llmConfig.apiKey = apiKey;
   }
 
-  const shouldEnableThinking =
-    thinking && thinkingBudget != null && (thinkingBudget > 0 || thinkingBudget === -1);
-
-  if (shouldEnableThinking && provider === Providers.GOOGLE) {
-    (llmConfig as GoogleClientOptions).thinkingConfig = {
-      thinkingBudget: thinking ? thinkingBudget : googleSettings.thinkingBudget.default,
-      includeThoughts: Boolean(thinking),
-    };
-  } else if (shouldEnableThinking && provider === Providers.VERTEXAI) {
-    (llmConfig as VertexAIClientOptions).thinkingBudget = thinking
-      ? thinkingBudget
-      : googleSettings.thinkingBudget.default;
-    (llmConfig as VertexAIClientOptions).includeThoughts = Boolean(thinking);
-  }
-
   /*
   let legacyOptions = {};
   // Filter out any "examples" that are empty
@@ -174,11 +152,11 @@ export function getGoogleConfig(
   */
 
   if (reverseProxyUrl) {
-    (llmConfig as GoogleClientOptions).baseUrl = reverseProxyUrl;
+    llmConfig.baseUrl = reverseProxyUrl;
   }
 
   if (authHeader) {
-    (llmConfig as GoogleClientOptions).customHeaders = {
+    llmConfig.customHeaders = {
       Authorization: `Bearer ${apiKey}`,
     };
   }
@@ -191,3 +169,8 @@ export function getGoogleConfig(
     llmConfig,
   };
 }
+
+module.exports = {
+  getLLMConfig,
+  getSafetySettings,
+};

api/server/services/Endpoints/gptPlugins/build.js

@@ -0,0 +1,41 @@
+const { removeNullishValues } = require('librechat-data-provider');
+const generateArtifactsPrompt = require('~/app/clients/prompts/artifacts');
+
+const buildOptions = (endpoint, parsedBody) => {
+  const {
+    modelLabel,
+    chatGptLabel,
+    promptPrefix,
+    agentOptions,
+    tools = [],
+    iconURL,
+    greeting,
+    spec,
+    maxContextTokens,
+    artifacts,
+    ...modelOptions
+  } = parsedBody;
+  const endpointOption = removeNullishValues({
+    endpoint,
+    tools: tools
+      .map((tool) => tool?.pluginKey ?? tool)
+      .filter((toolName) => typeof toolName === 'string'),
+    modelLabel,
+    chatGptLabel,
+    promptPrefix,
+    agentOptions,
+    iconURL,
+    greeting,
+    spec,
+    maxContextTokens,
+    modelOptions,
+  });
+
+  if (typeof artifacts === 'string') {
+    endpointOption.artifactsPrompt = generateArtifactsPrompt({ endpoint, artifacts });
+  }
+
+  return endpointOption;
+};
+
+module.exports = buildOptions;

api/server/services/Endpoints/gptPlugins/index.js

@@ -0,0 +1,7 @@
+const buildOptions = require('./build');
+const initializeClient = require('./initialize');
+
+module.exports = {
+  buildOptions,
+  initializeClient,
+};

api/server/services/Endpoints/gptPlugins/initialize.js

@@ -0,0 +1,134 @@
+const {
+  EModelEndpoint,
+  resolveHeaders,
+  mapModelToAzureConfig,
+} = require('librechat-data-provider');
+const { isEnabled, isUserProvided, getAzureCredentials } = require('@librechat/api');
+const { getUserKeyValues, checkUserKeyExpiry } = require('~/server/services/UserService');
+const { PluginsClient } = require('~/app');
+
+const initializeClient = async ({ req, res, endpointOption }) => {
+  const {
+    PROXY,
+    OPENAI_API_KEY,
+    AZURE_API_KEY,
+    PLUGINS_USE_AZURE,
+    OPENAI_REVERSE_PROXY,
+    AZURE_OPENAI_BASEURL,
+    OPENAI_SUMMARIZE,
+    DEBUG_PLUGINS,
+  } = process.env;
+
+  const { key: expiresAt, model: modelName } = req.body;
+  const contextStrategy = isEnabled(OPENAI_SUMMARIZE) ? 'summarize' : null;
+
+  let useAzure = isEnabled(PLUGINS_USE_AZURE);
+  let endpoint = useAzure ? EModelEndpoint.azureOpenAI : EModelEndpoint.openAI;
+
+  /** @type {false | TAzureConfig} */
+  const azureConfig = req.app.locals[EModelEndpoint.azureOpenAI];
+  useAzure = useAzure || azureConfig?.plugins;
+
+  if (useAzure && endpoint !== EModelEndpoint.azureOpenAI) {
+    endpoint = EModelEndpoint.azureOpenAI;
+  }
+
+  const credentials = {
+    [EModelEndpoint.openAI]: OPENAI_API_KEY,
+    [EModelEndpoint.azureOpenAI]: AZURE_API_KEY,
+  };
+
+  const baseURLOptions = {
+    [EModelEndpoint.openAI]: OPENAI_REVERSE_PROXY,
+    [EModelEndpoint.azureOpenAI]: AZURE_OPENAI_BASEURL,
+  };
+
+  const userProvidesKey = isUserProvided(credentials[endpoint]);
+  const userProvidesURL = isUserProvided(baseURLOptions[endpoint]);
+
+  let userValues = null;
+  if (expiresAt && (userProvidesKey || userProvidesURL)) {
+    checkUserKeyExpiry(expiresAt, endpoint);
+    userValues = await getUserKeyValues({ userId: req.user.id, name: endpoint });
+  }
+
+  let apiKey = userProvidesKey ? userValues?.apiKey : credentials[endpoint];
+  let baseURL = userProvidesURL ? userValues?.baseURL : baseURLOptions[endpoint];
+
+  const clientOptions = {
+    contextStrategy,
+    debug: isEnabled(DEBUG_PLUGINS),
+    reverseProxyUrl: baseURL ? baseURL : null,
+    proxy: PROXY ?? null,
+    req,
+    res,
+    ...endpointOption,
+  };
+
+  if (useAzure && azureConfig) {
+    const { modelGroupMap, groupMap } = azureConfig;
+    const {
+      azureOptions,
+      baseURL,
+      headers = {},
+      serverless,
+    } = mapModelToAzureConfig({
+      modelName,
+      modelGroupMap,
+      groupMap,
+    });
+
+    clientOptions.reverseProxyUrl = baseURL ?? clientOptions.reverseProxyUrl;
+    clientOptions.headers = resolveHeaders({ ...headers, ...(clientOptions.headers ?? {}) });
+
+    clientOptions.titleConvo = azureConfig.titleConvo;
+    clientOptions.titleModel = azureConfig.titleModel;
+    clientOptions.titleMethod = azureConfig.titleMethod ?? 'completion';
+
+    const azureRate = modelName.includes('gpt-4') ? 30 : 17;
+    clientOptions.streamRate = azureConfig.streamRate ?? azureRate;
+
+    const groupName = modelGroupMap[modelName].group;
+    clientOptions.addParams = azureConfig.groupMap[groupName].addParams;
+    clientOptions.dropParams = azureConfig.groupMap[groupName].dropParams;
+    clientOptions.forcePrompt = azureConfig.groupMap[groupName].forcePrompt;
+
+    apiKey = azureOptions.azureOpenAIApiKey;
+    clientOptions.azure = !serverless && azureOptions;
+    if (serverless === true) {
+      clientOptions.defaultQuery = azureOptions.azureOpenAIApiVersion
+        ? { 'api-version': azureOptions.azureOpenAIApiVersion }
+        : undefined;
+      clientOptions.headers['api-key'] = apiKey;
+    }
+  } else if (useAzure || (apiKey && apiKey.includes('{"azure') && !clientOptions.azure)) {
+    clientOptions.azure = userProvidesKey ? JSON.parse(userValues.apiKey) : getAzureCredentials();
+    apiKey = clientOptions.azure.azureOpenAIApiKey;
+  }
+
+  /** @type {undefined | TBaseEndpoint} */
+  const pluginsConfig = req.app.locals[EModelEndpoint.gptPlugins];
+
+  if (!useAzure && pluginsConfig) {
+    clientOptions.streamRate = pluginsConfig.streamRate;
+  }
+
+  /** @type {undefined | TBaseEndpoint} */
+  const allConfig = req.app.locals.all;
+  if (allConfig) {
+    clientOptions.streamRate = allConfig.streamRate;
+  }
+
+  if (!apiKey) {
+    throw new Error(`${endpoint} API key not provided. Please provide it again.`);
+  }
+
+  const client = new PluginsClient(apiKey, clientOptions);
+  return {
+    client,
+    azure: clientOptions.azure,
+    openAIApiKey: apiKey,
+  };
+};
+
+module.exports = initializeClient;

api/server/services/Endpoints/gptPlugins/initialize.spec.js

@@ -0,0 +1,410 @@
+// gptPlugins/initializeClient.spec.js
+jest.mock('~/cache/getLogStores');
+const { EModelEndpoint, ErrorTypes, validateAzureGroups } = require('librechat-data-provider');
+const { getUserKey, getUserKeyValues } = require('~/server/services/UserService');
+const initializeClient = require('./initialize');
+const { PluginsClient } = require('~/app');
+
+// Mock getUserKey since it's the only function we want to mock
+jest.mock('~/server/services/UserService', () => ({
+  getUserKey: jest.fn(),
+  getUserKeyValues: jest.fn(),
+  checkUserKeyExpiry: jest.requireActual('~/server/services/UserService').checkUserKeyExpiry,
+}));
+
+describe('gptPlugins/initializeClient', () => {
+  // Set up environment variables
+  const originalEnvironment = process.env;
+  const app = {
+    locals: {},
+  };
+
+  const validAzureConfigs = [
+    {
+      group: 'librechat-westus',
+      apiKey: 'WESTUS_API_KEY',
+      instanceName: 'librechat-westus',
+      version: '2023-12-01-preview',
+      models: {
+        'gpt-4-vision-preview': {
+          deploymentName: 'gpt-4-vision-preview',
+          version: '2024-02-15-preview',
+        },
+        'gpt-3.5-turbo': {
+          deploymentName: 'gpt-35-turbo',
+        },
+        'gpt-3.5-turbo-1106': {
+          deploymentName: 'gpt-35-turbo-1106',
+        },
+        'gpt-4': {
+          deploymentName: 'gpt-4',
+        },
+        'gpt-4-1106-preview': {
+          deploymentName: 'gpt-4-1106-preview',
+        },
+      },
+    },
+    {
+      group: 'librechat-eastus',
+      apiKey: 'EASTUS_API_KEY',
+      instanceName: 'librechat-eastus',
+      deploymentName: 'gpt-4-turbo',
+      version: '2024-02-15-preview',
+      models: {
+        'gpt-4-turbo': true,
+      },
+      baseURL: 'https://eastus.example.com',
+      additionalHeaders: {
+        'x-api-key': 'x-api-key-value',
+      },
+    },
+    {
+      group: 'mistral-inference',
+      apiKey: 'AZURE_MISTRAL_API_KEY',
+      baseURL:
+        'https://Mistral-large-vnpet-serverless.region.inference.ai.azure.com/v1/chat/completions',
+      serverless: true,
+      models: {
+        'mistral-large': true,
+      },
+    },
+    {
+      group: 'llama-70b-chat',
+      apiKey: 'AZURE_LLAMA2_70B_API_KEY',
+      baseURL:
+        'https://Llama-2-70b-chat-qmvyb-serverless.region.inference.ai.azure.com/v1/chat/completions',
+      serverless: true,
+      models: {
+        'llama-70b-chat': true,
+      },
+    },
+  ];
+
+  const { modelNames, modelGroupMap, groupMap } = validateAzureGroups(validAzureConfigs);
+
+  beforeEach(() => {
+    jest.resetModules(); // Clears the cache
+    process.env = { ...originalEnvironment }; // Make a copy
+  });
+
+  afterAll(() => {
+    process.env = originalEnvironment; // Restore original env vars
+  });
+
+  test('should initialize PluginsClient with OpenAI API key and default options', async () => {
+    process.env.OPENAI_API_KEY = 'test-openai-api-key';
+    process.env.PLUGINS_USE_AZURE = 'false';
+    process.env.DEBUG_PLUGINS = 'false';
+    process.env.OPENAI_SUMMARIZE = 'false';
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    const { client, openAIApiKey } = await initializeClient({ req, res, endpointOption });
+
+    expect(openAIApiKey).toBe('test-openai-api-key');
+    expect(client).toBeInstanceOf(PluginsClient);
+  });
+
+  test('should initialize PluginsClient with Azure credentials when PLUGINS_USE_AZURE is true', async () => {
+    process.env.AZURE_API_KEY = 'test-azure-api-key';
+    (process.env.AZURE_OPENAI_API_INSTANCE_NAME = 'some-value'),
+      (process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME = 'some-value'),
+      (process.env.AZURE_OPENAI_API_VERSION = 'some-value'),
+      (process.env.AZURE_OPENAI_API_COMPLETIONS_DEPLOYMENT_NAME = 'some-value'),
+      (process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME = 'some-value'),
+      (process.env.PLUGINS_USE_AZURE = 'true');
+    process.env.DEBUG_PLUGINS = 'false';
+    process.env.OPENAI_SUMMARIZE = 'false';
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'test-model' } };
+
+    const { client, azure } = await initializeClient({ req, res, endpointOption });
+
+    expect(azure.azureOpenAIApiKey).toBe('test-azure-api-key');
+    expect(client).toBeInstanceOf(PluginsClient);
+  });
+
+  test('should use the debug option when DEBUG_PLUGINS is enabled', async () => {
+    process.env.OPENAI_API_KEY = 'test-openai-api-key';
+    process.env.DEBUG_PLUGINS = 'true';
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    const { client } = await initializeClient({ req, res, endpointOption });
+
+    expect(client.options.debug).toBe(true);
+  });
+
+  test('should set contextStrategy to summarize when OPENAI_SUMMARIZE is enabled', async () => {
+    process.env.OPENAI_API_KEY = 'test-openai-api-key';
+    process.env.OPENAI_SUMMARIZE = 'true';
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    const { client } = await initializeClient({ req, res, endpointOption });
+
+    expect(client.options.contextStrategy).toBe('summarize');
+  });
+
+  // ... additional tests for reverseProxyUrl, proxy, user-provided keys, etc.
+
+  test('should throw an error if no API keys are provided in the environment', async () => {
+    // Clear the environment variables for API keys
+    delete process.env.OPENAI_API_KEY;
+    delete process.env.AZURE_API_KEY;
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    await expect(initializeClient({ req, res, endpointOption })).rejects.toThrow(
+      `${EModelEndpoint.openAI} API key not provided.`,
+    );
+  });
+
+  // Additional tests for gptPlugins/initializeClient.spec.js
+
+  // ... (previous test setup code)
+
+  test('should handle user-provided OpenAI keys and check expiry', async () => {
+    process.env.OPENAI_API_KEY = 'user_provided';
+    process.env.PLUGINS_USE_AZURE = 'false';
+
+    const futureDate = new Date(Date.now() + 10000).toISOString();
+    const req = {
+      body: { key: futureDate },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    getUserKeyValues.mockResolvedValue({ apiKey: 'test-user-provided-openai-api-key' });
+
+    const { openAIApiKey } = await initializeClient({ req, res, endpointOption });
+
+    expect(openAIApiKey).toBe('test-user-provided-openai-api-key');
+  });
+
+  test('should handle user-provided Azure keys and check expiry', async () => {
+    process.env.AZURE_API_KEY = 'user_provided';
+    process.env.PLUGINS_USE_AZURE = 'true';
+
+    const futureDate = new Date(Date.now() + 10000).toISOString();
+    const req = {
+      body: { key: futureDate },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'test-model' } };
+
+    getUserKeyValues.mockResolvedValue({
+      apiKey: JSON.stringify({
+        azureOpenAIApiKey: 'test-user-provided-azure-api-key',
+        azureOpenAIApiDeploymentName: 'test-deployment',
+      }),
+    });
+
+    const { azure } = await initializeClient({ req, res, endpointOption });
+
+    expect(azure.azureOpenAIApiKey).toBe('test-user-provided-azure-api-key');
+  });
+
+  test('should throw an error if the user-provided key has expired', async () => {
+    process.env.OPENAI_API_KEY = 'user_provided';
+    process.env.PLUGINS_USE_AZURE = 'FALSE';
+    const expiresAt = new Date(Date.now() - 10000).toISOString(); // Expired
+    const req = {
+      body: { key: expiresAt },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    await expect(initializeClient({ req, res, endpointOption })).rejects.toThrow(
+      /expired_user_key/,
+    );
+  });
+
+  test('should throw an error if the user-provided Azure key is invalid JSON', async () => {
+    process.env.AZURE_API_KEY = 'user_provided';
+    process.env.PLUGINS_USE_AZURE = 'true';
+
+    const req = {
+      body: { key: new Date(Date.now() + 10000).toISOString() },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    // Simulate an invalid JSON string returned from getUserKey
+    getUserKey.mockResolvedValue('invalid-json');
+    getUserKeyValues.mockImplementation(() => {
+      let userValues = getUserKey();
+      try {
+        userValues = JSON.parse(userValues);
+      } catch (e) {
+        throw new Error(
+          JSON.stringify({
+            type: ErrorTypes.INVALID_USER_KEY,
+          }),
+        );
+      }
+      return userValues;
+    });
+
+    await expect(initializeClient({ req, res, endpointOption })).rejects.toThrow(
+      /invalid_user_key/,
+    );
+  });
+
+  test('should correctly handle the presence of a reverse proxy', async () => {
+    process.env.OPENAI_REVERSE_PROXY = 'http://reverse.proxy';
+    process.env.PROXY = 'http://proxy';
+    process.env.OPENAI_API_KEY = 'test-openai-api-key';
+
+    const req = {
+      body: { key: null },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = { modelOptions: { model: 'default-model' } };
+
+    const { client } = await initializeClient({ req, res, endpointOption });
+
+    expect(client.options.reverseProxyUrl).toBe('http://reverse.proxy');
+    expect(client.options.proxy).toBe('http://proxy');
+  });
+
+  test('should throw an error when user-provided values are not valid JSON', async () => {
+    process.env.OPENAI_API_KEY = 'user_provided';
+    const req = {
+      body: { key: new Date(Date.now() + 10000).toISOString(), endpoint: 'openAI' },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = {};
+
+    // Mock getUserKey to return a non-JSON string
+    getUserKey.mockResolvedValue('not-a-json');
+    getUserKeyValues.mockImplementation(() => {
+      let userValues = getUserKey();
+      try {
+        userValues = JSON.parse(userValues);
+      } catch (e) {
+        throw new Error(
+          JSON.stringify({
+            type: ErrorTypes.INVALID_USER_KEY,
+          }),
+        );
+      }
+      return userValues;
+    });
+
+    await expect(initializeClient({ req, res, endpointOption })).rejects.toThrow(
+      /invalid_user_key/,
+    );
+  });
+
+  test('should initialize client correctly for Azure OpenAI with valid configuration', async () => {
+    const req = {
+      body: {
+        key: null,
+        endpoint: EModelEndpoint.gptPlugins,
+        model: modelNames[0],
+      },
+      user: { id: '123' },
+      app: {
+        locals: {
+          [EModelEndpoint.azureOpenAI]: {
+            plugins: true,
+            modelNames,
+            modelGroupMap,
+            groupMap,
+          },
+        },
+      },
+    };
+    const res = {};
+    const endpointOption = {};
+
+    const client = await initializeClient({ req, res, endpointOption });
+    expect(client.client.options.azure).toBeDefined();
+  });
+
+  test('should initialize client with default options when certain env vars are not set', async () => {
+    delete process.env.OPENAI_SUMMARIZE;
+    process.env.OPENAI_API_KEY = 'some-api-key';
+
+    const req = {
+      body: { key: null, endpoint: EModelEndpoint.gptPlugins },
+      user: { id: '123' },
+      app,
+    };
+    const res = {};
+    const endpointOption = {};
+
+    const client = await initializeClient({ req, res, endpointOption });
+    expect(client.client.options.contextStrategy).toBe(null);
+  });
+
+  test('should correctly use user-provided apiKey and baseURL when provided', async () => {
+    process.env.OPENAI_API_KEY = 'user_provided';
+    process.env.OPENAI_REVERSE_PROXY = 'user_provided';
+    const req = {
+      body: {
+        key: new Date(Date.now() + 10000).toISOString(),
+        endpoint: 'openAI',
+      },
+      user: {
+        id: '123',
+      },
+      app,
+    };
+    const res = {};
+    const endpointOption = {};
+
+    getUserKeyValues.mockResolvedValue({
+      apiKey: 'test',
+      baseURL: 'https://user-provided-url.com',
+    });
+
+    const result = await initializeClient({ req, res, endpointOption });
+
+    expect(result.openAIApiKey).toBe('test');
+    expect(result.client.options.reverseProxyUrl).toBe('https://user-provided-url.com');
+  });
+});

api/server/services/Endpoints/index.js

@@ -1,58 +0,0 @@
-const { Providers } = require('@librechat/agents');
-const { EModelEndpoint } = require('librechat-data-provider');
-const initAnthropic = require('~/server/services/Endpoints/anthropic/initialize');
-const getBedrockOptions = require('~/server/services/Endpoints/bedrock/options');
-const initOpenAI = require('~/server/services/Endpoints/openAI/initialize');
-const initCustom = require('~/server/services/Endpoints/custom/initialize');
-const initGoogle = require('~/server/services/Endpoints/google/initialize');
-const { getCustomEndpointConfig } = require('~/server/services/Config');
-
-const providerConfigMap = {
-  [Providers.XAI]: initCustom,
-  [Providers.OLLAMA]: initCustom,
-  [Providers.DEEPSEEK]: initCustom,
-  [Providers.OPENROUTER]: initCustom,
-  [EModelEndpoint.openAI]: initOpenAI,
-  [EModelEndpoint.google]: initGoogle,
-  [EModelEndpoint.azureOpenAI]: initOpenAI,
-  [EModelEndpoint.anthropic]: initAnthropic,
-  [EModelEndpoint.bedrock]: getBedrockOptions,
-};
-
-/**
- * Get the provider configuration and override endpoint based on the provider string
- * @param {string} provider - The provider string
- * @returns {Promise<{
- * getOptions: Function,
- * overrideProvider?: string,
- * customEndpointConfig?: TEndpoint
- * }>}
- */
-async function getProviderConfig(provider) {
-  let getOptions = providerConfigMap[provider];
-  let overrideProvider;
-  /** @type {TEndpoint | undefined} */
-  let customEndpointConfig;
-
-  if (!getOptions && providerConfigMap[provider.toLowerCase()] != null) {
-    overrideProvider = provider.toLowerCase();
-    getOptions = providerConfigMap[overrideProvider];
-  } else if (!getOptions) {
-    customEndpointConfig = await getCustomEndpointConfig(provider);
-    if (!customEndpointConfig) {
-      throw new Error(`Provider ${provider} not supported`);
-    }
-    getOptions = initCustom;
-    overrideProvider = Providers.OPENAI;
-  }
-
-  return {
-    getOptions,
-    overrideProvider,
-    customEndpointConfig,
-  };
-}
-
-module.exports = {
-  getProviderConfig,
-};

api/server/services/Endpoints/openAI/initialize.js

@@ -1,7 +1,11 @@
-const { ErrorTypes, EModelEndpoint, mapModelToAzureConfig } = require('librechat-data-provider');
+const {
+  ErrorTypes,
+  EModelEndpoint,
+  resolveHeaders,
+  mapModelToAzureConfig,
+} = require('librechat-data-provider');
 const {
   isEnabled,
-  resolveHeaders,
   isUserProvided,
   getOpenAIConfig,
   getAzureCredentials,
@@ -80,10 +84,7 @@ const initializeClient = async ({
     });
 
     clientOptions.reverseProxyUrl = baseURL ?? clientOptions.reverseProxyUrl;
-    clientOptions.headers = resolveHeaders(
-      { ...headers, ...(clientOptions.headers ?? {}) },
-      req.user,
-    );
+    clientOptions.headers = resolveHeaders({ ...headers, ...(clientOptions.headers ?? {}) });
 
     clientOptions.titleConvo = azureConfig.titleConvo;
     clientOptions.titleModel = azureConfig.titleModel;
@@ -138,7 +139,7 @@ const initializeClient = async ({
   }
 
   if (optionsOnly) {
-    const modelOptions = endpointOption?.model_parameters ?? {};
+    const modelOptions = endpointOption.model_parameters;
     modelOptions.model = modelName;
     clientOptions = Object.assign({ modelOptions }, clientOptions);
     clientOptions.modelOptions.user = req.user.id;