DocIntel

A powerful document intelligence platform that enables document uploading, processing, embedding, and semantic search with AI-powered chat capabilities.

Watch Demo Video

Features

• Document Processing: Upload and process PDF, DOCX, PPTX, Excel files
• Intelligent Chunking: Smart document chunking for optimal retrieval
• Vector Embeddings: Generate and store embeddings for semantic search
• AI-Powered Chat: Chat with your documents using AI
• Citation Support: Get citations with source document context
• Multi-User Support: Document collections separated by user
• Parallel Processing: Efficient document processing with parallel embedding generation
• OCR Capabilities: Extract text from scanned documents and images

Technology Stack

• Backend: FastAPI
• Vector Database: Qdrant
• Embeddings: Azure OpenAI
• Containerization: Docker
• Deployment: Fly.io

Installation

Prerequisites

• Python 3.8+
• Docker (optional)

Local Setup

1. Clone the repository:

   git clone https://github.com/yourusername/docintel.git
   cd docintel
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Run the application:

   uvicorn app.main:app --reload
   

Docker Setup

1. Build the Docker image:

   docker build -t docintel .
   

2. Run the container:

   docker run -p 8000:8000 docintel
   

Usage

Uploading Documents

Upload documents using the /upload endpoint. Supported file types include:

• PDF (.pdf)
• Word Documents (.docx)
• PowerPoint Presentations (.pptx)
• Excel Spreadsheets (.xlsx, .xls, .csv)

Querying Documents

Use the /query endpoint to search through processed documents with natural language queries.

Chat with Documents

Create a chat session with the /sessions endpoint and send messages to interact with your documents.

API Endpoints

Document Management

• POST /upload - Upload and process a document
• GET /list - List all documents
• GET /{document_id} - Get document details
• DELETE /{document_id} - Delete a document
• GET /{document_id}/file - Download original document
• GET /statistics - Get document statistics

Chat

• POST /sessions - Create a new chat session
• GET /sessions - List chat sessions
• GET /sessions/{session_id} - Get chat session details
• DELETE /sessions/{session_id} - Delete a chat session
• GET /sessions/{session_id}/messages - Get chat history
• POST /sessions/{session_id}/messages - Send a message
• GET /messages/{message_id} - Get a specific message
• GET /messages/{message_id}/citations - Get citations for a message
• GET /citations/{document_id}/{chunk_id} - Get citation source

Architecture

DocIntel uses a modular architecture:

1. Parser Module - Extracts text and metadata from different document types
2. Chunking Module - Divides documents into manageable chunks for processing
3. Embedding Module - Generates vector embeddings for document chunks
4. Storage Module - Manages vector storage using Qdrant
5. Chat Module - Handles chat sessions and message processing

Deployment

DocIntel can be deployed to Fly.io using the included fly.toml configuration:

fly launch

License

MIT License