Document Chat Assistant

This project is a RESTful API built with FastAPI that allows users to upload PDF documents, extract their text content, and chat with the document's content using a Gemini-based LLM integration.

Features

• User Authentication: JWT-based authentication for secure access.
• PDF Management: Upload, list, parse, and select PDF documents.
• LLM Chat: Engage in conversations about selected PDF documents using the Gemini API.
• Data Storage: PostgreSQL for user and chat history, MongoDB (GridFS) for PDF files and metadata.
• Containerization: Docker support for easy setup and deployment.

Architecture

For a detailed overview of the system's architecture, see the ARCHITECTURE.md file.

For information on API errors and domain exceptions, see the ERRORS.md file.

Tech Stack

• Backend Framework: FastAPI
• Databases: PostgreSQL, MongoDB
• PDF Processing: PyPDF2
• LLM Integration: Gemini API (via curl)
• Authentication: JWT
• Containerization: Docker

Setup and Installation

• Clone the repository:

  git clone https://github.com/jryusuf/pdf-llm-chat
  cd pdf-llm-chat

Running the Project using Docker

1. Build the Docker image:

   docker build -t document-chat-assistant .

2. Run the Docker container: If using docker-compose.yml, you can use:

   docker-compose up --build

   Otherwise, you can run the container directly, ensuring you map ports and provide environment variables:

   docker run -p 8000:8000 --env-file .env document-chat-assistant

   (Adjust port mapping if necessary)

The API should be accessible at http://localhost:8000.

Development and Testing with Makefile

The project includes a Makefile to streamline common development and testing tasks. Ensure you have make installed on your system.

Here are the available targets:

• make all: Runs setup-dev and install-dev. This is the default target.
• make setup-dev: Creates a Python virtual environment in the ./venv directory.
• make install-dev: Installs the development dependencies listed in requirements-dev.txt into the virtual environment.
• make run-server: Starts the FastAPI development server using Uvicorn with auto-reloading enabled. The server runs on http://0.0.0.0:8000.
• make test: Runs all tests, including pytest unit/integration tests and behave feature tests for all modules (account, chat, pdf). Includes coverage reporting.
• make clean: Removes the virtual environment directory (./venv), __pycache__ directories, .pyc, .pyo, .pytest_cache, and .coverage files.

To use the Makefile targets, first ensure you have run make all or make setup-dev followed by make install-dev. Then, you can run commands like:

make run-server
make test

Environment Variables

Create a .env file in the project root with the following variables, based on the .env.example file:

GEMINI_API_KEY=your_gemini_api_key
GEMINI_SYSTEM_PROMPT='Your desired system prompt for the AI assistant.'

• GEMINI_API_KEY: Your API key for the Google Gemini API.
• GEMINI_SYSTEM_PROMPT: The initial prompt used to configure the AI assistant's behavior.

Note: Additional environment variables for database connections (DATABASE_URL, MONGO_URL) are configured in app/core/config.py and may need to be set depending on your deployment method (e.g., if not using the provided docker-compose.yml).

Known Issues and Limitations

• LLM requests and parsing currently implemented as async methods, this can move to background tasks or using celery/procastinator workers

Improvement Ideas (Optional)

• Fix feature test authentication errors
• Add support for other document types (e.g., DOCX).
• Improve error handling and logging details.
• Implement rate limiting for API endpoints.
• Add more advanced chat features (e.g., streaming responses).
• Seperate dependency between chat and pdf services to reduce coupling
• Deploy each service as seperate microservices
• Add API gateway useing Ngnix or other reverse proxies

API Usage Examples

(Note: Replace your_token with the actual JWT token obtained from the login endpoint)

1. Register a new user:

curl -X POST \
  http://localhost:8000/register \
  -H 'Content-Type: application/json' \
  -d '{
    "email": "[email protected]",
    "password": "securepassword"
  }'

2. Login a user:

curl -X POST \
  http://localhost:8000/login \
  -H 'Content-Type: application/json' \
  -d '{
    "email": "[email protected]",
    "password": "securepassword"
  }'

3. Upload a PDF:

curl -X POST \
  http://localhost:8000/pdf-upload \
  -H 'Authorization: Bearer your_token' \
  -F 'file=@/path/to/your/document.pdf'

4. List uploaded PDFs:

curl -X GET \
  http://localhost:8000/pdf-list \
  -H 'Authorization: Bearer your_token'

5. Parse a PDF (replace pdf_id with the actual ID):

curl -X POST \
  http://localhost:8000/pdf-parse \
  -H 'Authorization: Bearer your_token' \
  -H 'Content-Type: application/json' \
  -d '{
    "pdf_id": "your_pdf_id"
  }'

6. Select a PDF for chat (replace pdf_id with the actual ID):

curl -X POST \
  http://localhost:8000/pdf-select \
  -H 'Authorization: Bearer your_token' \
  -H 'Content-Type: application/json' \
  -d '{
    "pdf_id": "your_pdf_id"
  }'

7. Chat with the selected PDF:

curl -X POST \
  http://localhost:8000/pdf-chat \
  -H 'Authorization: Bearer your_token' \
  -H 'Content-Type: application/json' \
  -d '{
    "message": "What is this document about?"
  }'

8. Get chat history:

curl -X GET \
  http://localhost:8000/chat-history \
  -H 'Authorization: Bearer your_token'