unstract/backend

Unstract Backend

Contains the backend services for Unstract written with Django and DRF.

Dependencies

1. Postgres
2. Redis
3. RabbitMQ

Getting started

Install and run locally

Create your virtual env

All commands assumes that you have activated your venv.

Install UV: https://docs.astral.sh/uv/getting-started/installation/

# Create venv and install dependencies
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv sync

Installing dependencies

Go to service dir and install dependencies listed in corresponding pyproject.toml.

# Install dependencies
uv sync

# Install specific dev dependency group
uv sync --group dev
# Install production dependencies only
uv sync --group deploy

Running scripts

UV allows you to run python scripts applicable within the service dir.

uv run sample_script.py

Running commands

• If you plan to run the django server locally, make sure the dependent services are up (either locally or through docker compose)
• Copy sample.env into .env and update the necessary variables. For eg:

DJANGO_SETTINGS_MODULE='backend.settings.dev'
DB_HOST='localhost'
DB_USER='unstract_dev'
DB_PASSWORD='unstract_pass'
DB_NAME='unstract_db'
DB_PORT=5432

• If you've made changes to the model, run uv run manage.py makemigrations, else ignore this step
• Run the following to apply any migrations to the DB and start the server

uv run manage.py migrate
uv run manage.py runserver localhost:8000

• Server will start and run at port 8000. (http://localhost:8000)

Authentication

The default username is unstract and the default password is unstract.

Initial Setup

To customize the username or password:

1. Navigate to /backend/.env created from /backend/sample.env
2. Update the values for DEFAULT_AUTH_USERNAME and DEFAULT_AUTH_PASSWORD with strong, unique credentials of your choosing
3. Save the /backend/.env file and restart the server to apply changes

Updating Credentials

To update the username or password after initial setup:

1. Modify the credentials in /backend/.env
   • DEFAULT_AUTH_USERNAME=your_new_username
   • DEFAULT_AUTH_PASSWORD=your_new_password
2. Save changes and restart backend service

Now you can login with the new credentials.

Important Notes

• DEFAULT_AUTH_USERNAME must not match the username of any Django superuser or admin account. Keeping them distinct ensures security and avoids potential conflicts.
• Use strong and unique credentials to protect your system.
• The authentication system validates credentials against the values specified in the /backend/.env file.

Asynchronous Execution

This project uses Celery for handling asynchronous execution. Celery tasks are managed through various queues and consumed by workers.

ETL, TASK, and API Deployment tasks are handled by these asynchronous workers. Log management also utilizes Celery.

Queues

Queue Name	Description	Tasks
celery	Default queue for general Celery tasks, including those without a specific queue.	Webhook notifications, Pipeline (ETL, Tasks) Executions.
celery_periodic_logs	Queue for persisting logs into the database.
celery_log_task_queue	Queue for publishing logs to WebSocket clients.
celery_api_deployments	Queue for managing API deployment tasks.

Run Execution Worker

To start a Celery worker, use the following command:

celery -A backend worker --loglevel=info -Q <queue_name>

Autoscaling Workers

  celery -A backend worker --loglevel=info -Q <queue_name> --autoscale=<max_workers>,<min_workers>

Celery supports autoscaling of worker processes, allowing you to dynamically adjust the number of workers based on workload.

• Max Workers (max_workers): This value is related to your CPU resources and the level of concurrency you need.

  • For CPU-bound tasks: Consider setting max_workers close to or slightly above the number of CPU cores.
  • For I/O-bound tasks: You can set a higher max_workers value, typically 2-3 times the number of CPU cores.

• Min Workers (min_workers): This is the minimum number of worker processes that will always be running.

Worker Dashboard

• We have to ensure the package flower is installed in the current environment
• Run command

celery -A backend flower

This command will start Flower on the default port (5555) and can be accessed via a web browser. Flower provides a user-friendly interface for monitoring and managing Celery tasks

Broker Dashboard

• RabbitMQ provides a web interface for monitoring and managing the broker
• Access the dashboard at http://localhost:15672
• Default credentials: admin / password
• These can be configured with RABBITMQ_DEFAULT_USER and RABBITMQ_DEFAULT_PASS envs in essentials.env

Connecting to Postgres

Follow the below steps to connect to the postgres DB running with docker compose.

1. Exec into a shell within the postgres container

docker compose exec -it db bash

1. Connect to the db as the specified user

psql -d unstract_db -U unstract_dev

1. Execute PSQL commands within the shell.

API Docs

While running the backend server locally, access the API documentation that's auto generated at the backend endpoint /api/v1/doc/.

NOTE: There exists issues accessing this when the django server is run with gunicorn (in case of running with a container)

• Account
• FileManagement

Connectors

Google Drive

The Google Drive connector makes use of PyDrive2 library and supports only OAuth 2.0 authentication. To set it up, follow the first step higlighted in Google's docs and set the client ID and client secret as envs in backend/.env

GOOGLE_OAUTH2_KEY="<client-id>"
GOOGLE_OAUTH2_SECRET="<client-secret>"

Tool Registry

Information regarding how tools are added and maintained can be found here.

Archived - (EXPERIMENTAL)

Accessing the admin site

• If its the first time, create a super user and follow the on-screen instructions

python manage.py createsuperuser

• Register your models in <app>/admin.py, for example

from django.contrib import admin
from .models import Prompt

admin.site.register(Prompt)

• Make sure the server is running and hit the /admin endpoint

Running unit tests

Units tests are run with pytest and pytest-django

pytest
pytest prompt # To run for an app named prompt

All tests are organized within an app, for eg: prompt/tests/test_urls.py

NOTE: The django server need not be up to run the tests, however the DB needs to be running.