DocuMind

Multi-modal AI agent that extracts information from PDFs, images, and documents to answer questions. Combines vision models with RAG architecture for intelligent document understanding.

Features

• PDF document processing and text extraction
• Image analysis using vision models
• Multi-modal document understanding
• RAG (Retrieval-Augmented Generation) architecture
• Question answering over uploaded documents

Getting Started

Prerequisites

• Python 3.9+
• OpenAI API key

Installation

pip install -r requirements.txt

Configuration

Copy .env.example to .env and add your API keys:

cp .env.example .env

Usage

Python API

from src.main import DocuMind

# Initialize DocuMind
app = DocuMind(use_cache=True)

# Process a PDF document
result = app.process_pdf("path/to/document.pdf")
print(f"Processed {result['chunks_processed']} chunks")

# Query the knowledge base
answer = app.query("What is this document about?")
print(f"Answer: {answer['answer']}")
print(f"Sources: {answer['num_sources']}")

# Get application statistics
stats = app.get_stats()
print(f"Documents in store: {stats['documents_in_store']}")

REST API

Start the API server:

python -m uvicorn src.api:app --reload

Upload a document:

curl -X POST "http://localhost:8000/upload" \
  -F "file=@document.pdf"

Query documents:

curl -X POST "http://localhost:8000/query" \
  -H "Content-Type: application/json" \
  -d '{"query": "What is the main topic?", "top_k": 5}'

Running Tests

pytest tests/

Architecture

DocuMind uses a multi-modal RAG architecture to process various document types:

1. Document Processing: Extract text and images from PDFs
2. Vision Analysis: Analyze images and diagrams using vision models
3. Embedding Generation: Create vector embeddings for text and visual content
4. Retrieval: Find relevant content based on user queries
5. Generation: Generate accurate answers using retrieved context

License

MIT