Document Query Assistant

Overview

This project is a document query assistant that combines a FastAPI backend with a Gradio frontend, featuring OCR capabilities. It uses LangChain and Ollama for document processing and conversational AI, with a focus on secure inter-service communication.

Features

• Document Processing: Support for multiple file formats including PDF, DOCX, TXT, HTML with OCR capabilities
• Interactive Chat: Context-aware chatbot interface for querying document content
• Secure Architecture: Isolated service communication with Docker networks
• OCR Integration: Automated text extraction from images and scanned PDFs
• LLM Integration: Powered by Ollama with optimized RAG configuration

Project Structure

.
├── front/
│   ├── interface.py          # Main Gradio application
│   ├── chat.py         # Chat functionality
│   ├── ingest.py       # Document ingestion
│   ├── Dockerfile
│   ├── requirements.txt
├── back/
│   ├── routes/
│   │   ├── chat.py     # Chat endpoints
│   │   ├── ingest.py   # Ingestion endpoints
│   ├── utils/
│   │   ├── app_langchain/
│   │   │   ├── data_parser.py
│   │   │   ├── process_vector.py
│   ├── config.py       # Application configuration
│   ├── Dockerfile
│   ├── requirements.txt
├── docker-compose.yml
├── .env.example

Network Architecture

The application uses isolated Docker networks for security:

• front-back: Communication between frontend and backend
• back-ollama: Communication between backend and Ollama
• External access only to frontend (7860) and backend API (8000)

Setup Instructions

Prerequisites

• Docker and Docker Compose
• Git

Installation

1. Clone the Repository

git clone https://github.com/sybtra/document-query-assistant.git
cd document-query-assistant

2. Configure Environment Create a .env file:

API_URL=http://backend:8000
DB_NAME=/app/data/chroma
APP_MODEL=llama2
MODEL_BASE_URL=http://ollama:11434

3. Build and Run

docker compose up --build

Accessing the Application

• Frontend Interface: http://localhost:7860
• API Documentation: http://localhost:8000/docs

API Endpoints

Document Ingestion

• Endpoint: POST /ingest/{collection_name}
• Purpose: Upload and process documents
• Supported Formats:
  • Text: .txt, .json
  • Documents: .pdf, .docx, .doc
  • Web: .html, .htm
  • Images (via OCR): .png, .jpg, .jpeg, .tiff, .bmp

Chat Interface

• Endpoint: POST /chat/{collection_name}
• Purpose: Process queries against ingested documents
• Features:
  • Context-aware responses
  • RAG-optimized configuration
  • Conversation memory

Development

Backend Requirements

fastapi
uvicorn
python-multipart
langchain
langchain_community
langchain_core
langchain_chroma
langchain_ollama
pytesseract
pdf2image
python-magic
python-dotenv
loguru

Frontend Requirements

gradio
python-dotenv
requests

Docker Configuration

Services

1. Frontend

   • Port: 7860
   • Network: front-back
   • Dependencies: backend

2. Backend

   • Port: 8000
   • Networks: front-back, back-ollama
   • Volumes: chroma_data

3. Ollama

   • Internal Port: 11434
   • Network: back-ollama
   • Volumes: ollama_data

Troubleshooting

Common Issues

1. Connection Refused

   • Verify network configurations in docker-compose.yml
   • Check service health status
   • Ensure correct environment variables

2. File Processing Errors

   • Verify file format support
   • Check OCR configuration
   • Ensure sufficient permissions

Maintenance

• Regular model updates via Ollama
• Vector store maintenance
• Log monitoring

License

MIT License

Repository

https://github.com/sybtra/document-query-assistant.git