DocQuery App

DocQuery is a document query application that uses Retrieval-Augmented Generation (RAG) to provide accurate answers to questions based on your documents. It supports PDF files, text files, and images with OCR capabilities.

Features

• Multiple Document Types: Support for PDFs, text files, and images (using OCR)
• Advanced Text Processing: Extracts, chunks, and indexes document content
• Vector Search: Fast similarity search using FAISS
• Open Source LLMs: Uses Hugging Face models for embeddings and text generation
• Optional Groq Integration: Can use Groq's API for faster inference
• REST API: Simple API for document upload and querying

System Architecture

docquery_app/
├── app.py                 # Main FastAPI application
├── extractors/            # Document extraction modules
│   ├── pdf_extractor.py   # PDF processing
│   ├── txt_extractor.py   # Plain text processing
│   └── ocr_extractor.py   # OCR for images and scanned PDFs
├── embeddings/            # Embedding generation and storage
│   └── embedder.py        # Creates and manages embeddings
├── rag/                   # RAG implementation
│   └── rag_chain.py       # Retrieval and answer generation
├── utils/                 # Utility functions
│   └── file_utils.py      # File handling utilities
├── requirements.txt       # Python dependencies
└── README.md              # This file

Setup

1. Clone the repository:

   git clone https://github.com/your-username/docquery-app.git
   cd docquery-app

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables: Create a .env file in the project root with the following:

   # API Keys
   HUGGINGFACE_TOKEN=your_huggingface_token_here
   # GROQ_API_KEY=your_groq_api_key_here  # Uncomment if using Groq
   
   # Model configuration
   MODEL_NAME=mistralai/Mistral-7B-Instruct-v0.2
   MODEL_PROVIDER=huggingface  # Options: huggingface, groq
   EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
   
   # Application settings
   DEBUG=True
   LOG_LEVEL=INFO
   MAX_DOCUMENT_SIZE_MB=10
   CHUNK_SIZE=512
   CHUNK_OVERLAP=50
   
   # OCR settings (if using)
   TESSERACT_PATH=/usr/bin/tesseract  # Update based on your system
   

5. Run the application:

   python app.py

OCR Setup (Optional)

For OCR functionality, you need to install Tesseract:

• Linux:

  sudo apt-get install tesseract-ocr

• macOS:

  brew install tesseract

• Windows: Download and install from Tesseract GitHub

API Usage

Upload a Document

curl -X POST "http://localhost:8000/upload" \
  -H "accept: application/json" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@/path/to/your/document.pdf"

Query Documents

curl -X POST "http://localhost:8000/query" \
  -H "Content-Type: application/json" \
  -d '{"question":"What is the main topic of the document?"}'

Clear Documents

curl -X DELETE "http://localhost:8000/documents"

Using Different Models

The application is configured to use Mistral 7B by default, but you can change to other models:

1. For local Hugging Face models:

   • Update the MODEL_NAME in your .env file
   • Set MODEL_PROVIDER=huggingface

2. For Groq API:

   • Get an API key from Groq
   • Add your GROQ_API_KEY to the .env file
   • Set