DocClaw - Document Intelligence Platform

A web-based document Q&A platform where companies upload documents and employees chat with AI that only answers from uploaded content.

Features

• Multi-tenant: Companies with multiple employees
• Document Upload: PDF, DOCX, PPTX, XLSX support
• AI-Powered Q&A: Chat with your documents using MiniMax-M2.7
• Vector Search: Semantic search with Qwen3 embeddings + Qdrant
• Chat History: Persistent conversation storage
• Auto Re-indexing: Documents persist and re-index on startup

Tech Stack

Frontend: Next.js 14, TypeScript, Tailwind CSS
Backend: FastAPI, Python
Database: PostgreSQL (Prisma ORM)
Vector Store: Qdrant
AI Models: MiniMax-M2.7-highspeed (chat), Qwen3-embedding-8b (embeddings)

Quick Start

Prerequisites

• Node.js 18+
• Python 3.10+
• PostgreSQL running on port 5432
• Qdrant running on port 6333

Setup

1. Clone and install dependencies

# Backend
cd backend
pip install -r requirements.txt

# Frontend
cd frontend
npm install

2. Configure environment

# Backend - copy and edit
cp backend/.env.example backend/.env
# Edit backend/.env with your API keys

# Frontend
cp frontend/.env.example frontend/.env

3. Set up database

cd backend
export DATABASE_URL="postgresql://user:pass@localhost:5432/doclaw"
npx prisma db push

4. Start services

# Terminal 1 - Backend
cd backend
python3 -m uvicorn app.main:app --host 0.0.0.0 --port 8000

# Terminal 2 - Frontend
cd frontend
npm run dev

5. Open browser

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000/docs

Environment Variables

Backend (backend/.env)

Variable	Description	Required
DATABASE_URL	PostgreSQL connection string	Yes
MINIMAX_API_KEY	MiniMax Token Plan API key	Yes
MINIMAX_BASE_URL	MiniMax API base URL	Yes
OPENROUTER_API_KEY	OpenRouter API key	Yes
OPENROUTER_BASE_URL	OpenRouter API base URL	Yes
QDRANT_HOST	Qdrant host	Yes
QDRANT_PORT	Qdrant port	Yes
SECRET_KEY	JWT secret key	Yes

Frontend (frontend/.env)

Variable	Description
NEXT_PUBLIC_API_URL	Backend API URL (default: http://localhost:8000)

Project Structure

doc-ai/
├── backend/
│   ├── app/
│   │   ├── routers/          # API endpoints
│   │   │   ├── auth.py       # Authentication
│   │   │   ├── chat.py       # Chat & AI
│   │   │   └── documents.py  # Document management
│   │   ├── services/         # Business logic
│   │   │   ├── auth_service.py
│   │   │   ├── document_processor.py
│   │   │   ├── embeddings.py
│   │   │   ├── indexer.py    # Auto re-indexing
│   │   │   ├── rag_engine.py
│   │   │   └── vector_store.py
│   │   ├── config.py
│   │   └── main.py
│   ├── prisma/
│   │   └── schema.prisma
│   ├── .env.example
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── app/              # Next.js pages
│   │   │   ├── chat/
│   │   │   ├── documents/
│   │   │   ├── login/
│   │   │   └── signup/
│   │   ├── components/
│   │   └── lib/
│   ├── .env.example
│   └── package.json
├── .gitignore
└── README.md

API Endpoints

Authentication

Method	Endpoint	Description
POST	/api/auth/signup	Register company & admin user
POST	/api/auth/signin	Login, returns JWT token
GET	/api/auth/me	Get current user info

Documents

Method	Endpoint	Description
POST	/api/documents/upload	Upload document (admin only)
GET	/api/documents/	List company documents
DELETE	/api/documents/{id}	Delete document (admin only)

Chat

Method	Endpoint	Description
POST	/api/chat/	Send message, returns AI response
GET	/api/chat/history/{id}	Get conversation messages
GET	/api/chat/conversations	List user conversations

Database Schema

• Company: Multi-tenant organizations
• Employee: Users with roles (ADMIN/EMPLOYEE)
• Document: Uploaded files with metadata & text content
• Conversation: Chat sessions
• Message: Individual messages with sources

How It Works

1. Upload: Admin uploads documents (PDF, DOCX, PPTX, XLSX)
2. Extract: Text is extracted and chunked
3. Embed: Chunks are embedded using Qwen3 via OpenRouter
4. Store: Embeddings stored in Qdrant, text in PostgreSQL
5. Chat: User asks question → search Qdrant → generate answer with MiniMax
6. Persist: On restart, documents auto re-index from PostgreSQL

License

MIT