Django RAG Docs

A Retrieval-Augmented Generation (RAG) system built with Django, PostgreSQL (pgvector), and LangChain. This system allows users to upload documents, search them using vector embeddings, and ask questions to receive AI-generated answers based on the document context.

Features

• Document Management: Upload and manage text documents via Django Admin.
• Vector Search: Automatically generates embeddings (using all-MiniLM-L6-v2) and performs semantic search using PostgreSQL pgvector.
• RAG QA System: Uses LangChain and zephyr-7b-beta (via Hugging Face API) to answer user questions based on stored documents.
• Dockerized: Fully containerized setup for easy deployment.

Prerequisites

• Docker and Docker Compose installed on your machine.
• A Hugging Face Access Token (for the LLM feature).

Installation & Setup

1. Clone the project: Extract the project files to your local machine.

2. Configure Environment: Rename .env.sample file to .env and find the variables and change its values to the desire value.

3. Build and Run:

docker compose up -d --build

4. Apply Migrations:

docker compose run web python manage.py migrate

5. Create Superuser (Admin):

docker compose run web python manage.py createsuperuser

Importing Sample Data

You can import documents from a CSV file (e.g., data.csv):

docker compose run web python manage.py import_documents data.csv

Usage

1. Admin Panel

Access the admin panel at http://localhost:8000/admin to manage documents and view query logs.

2. API Endpoints

Search Documents (Vector Search):

• URL: POST /api/search/
• Body:

{
    "query": "What is reinforcement learning?",
    "top_k": 3
}

Ask a Question (RAG):

• URL: POST /api/ask/
• Body:

{
    "question": "How does Git help developers?"
}

• Response: Returns the AI-generated answer and the source documents used.