🧠 DocuMind AI

Enterprise-grade RAG (Retrieval-Augmented Generation) platform — Upload documents, ask questions, get cited answers powered by OpenAI, Anthropic Claude, or Ollama.

---

✨ Features

Feature	Description
Multi-format Ingestion	PDF (with OCR), DOCX, TXT, Markdown, Web URLs
Hybrid RAG Search	Semantic (ChromaDB) + BM25 keyword search
Cross-encoder Re-ranking	MS-MARCO re-ranking for top results
Multi-LLM Support	OpenAI GPT-4, Anthropic Claude, Ollama (local)
LLM Fallback Chain	Auto-retry with next provider on failure
SSE Streaming	Real-time token-by-token answer streaming
Source Citations	Every answer references exact document chunks
Async Processing	Celery + Redis for non-blocking document ingestion
Conversation Memory	Per-session chat history with question condensation
Collections	Organize docs into named knowledge bases
REST API	Full OpenAPI/Swagger documented FastAPI backend
Web UI	Flask + Tailwind chat, upload, and collections UI
Docker	One-command docker-compose up deployment

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│          Flask Web UI  (chat, upload, collections)      │
└──────────────────────┬──────────────────────────────────┘
                       │ HTTP (via Nginx)
┌──────────────────────▼──────────────────────────────────┐
│              FastAPI Backend (REST API)                 │
│    /upload  /query  /collections  /health  /stream     │
└────┬──────────────┬──────────────────┬──────────────────┘
     │              │                  │
 Ingestion      RAG Query          LLM Router
 Service        Service            (LangChain)
     │              │                  │
     ▼              ▼                  ▼
 Celery +      ChromaDB          OpenAI / Claude
 Redis         VectorDB          / Ollama
     │
 PostgreSQL
 (metadata)

---

🚀 Quick Start

Prerequisites

• Docker & Docker Compose
• At least one LLM API key (OpenAI or Anthropic) or Ollama running locally

1. Clone & Configure

git clone https://github.com/yourname/documind-ai.git
cd documind-ai

# Copy and edit env file
cp .env.example .env

Edit .env and set your API keys:

DEFAULT_LLM_PROVIDER=openai
OPENAI_API_KEY=sk-your-key-here

# OR for Claude:
# DEFAULT_LLM_PROVIDER=anthropic
# ANTHROPIC_API_KEY=sk-ant-your-key-here

# OR for local Ollama:
# DEFAULT_LLM_PROVIDER=ollama
# OLLAMA_BASE_URL=http://host.docker.internal:11434

2. Launch

docker-compose up --build

3. Access

Service	URL
Web UI	http://localhost
API Docs (Swagger)	http://localhost/api/docs
ChromaDB	http://localhost:8001
Flower (Celery UI)	http://localhost:5555

---

📡 API Reference

Collections

GET    /api/v1/collections              # List all
POST   /api/v1/collections              # Create
GET    /api/v1/collections/{id}         # Get one
DELETE /api/v1/collections/{id}         # Delete (with all docs)

Documents

POST   /api/v1/documents/upload         # Upload files (multipart)
POST   /api/v1/documents/url            # Ingest from URL
GET    /api/v1/documents/{id}/status    # Check ingestion status
DELETE /api/v1/documents/{id}           # Delete document

Query

POST   /api/v1/query                    # Blocking RAG query
GET    /api/v1/query/stream             # SSE streaming query
POST   /api/v1/sessions                 # Create chat session
GET    /api/v1/sessions/{id}/messages   # Get session history

System

GET    /api/v1/health                   # Health of all services
GET    /api/v1/metrics                  # Usage statistics

Example Query

curl -X POST http://localhost/api/v1/query \
  -H "Content-Type: application/json" \
  -d '{
    "question": "What are the key findings?",
    "collection_id": "your-collection-uuid",
    "llm_provider": "openai",
    "top_k": 5,
    "include_sources": true
  }'

---

🐳 Docker Services

Service	Port	Description
nginx	80	Reverse proxy & load balancer
api	8000	FastAPI backend
web	5000	Flask frontend
chromadb	8001	Vector store
postgres	5432	Metadata & chat history
redis	6379	Celery broker & cache
celery_worker	—	Async ingestion worker

---

📁 Project Structure

documind-ai/
├── docker-compose.yml          # All services
├── .env.example                # Environment template
│
├── api/                        # FastAPI backend
│   ├── main.py                 # App entry point
│   ├── core/
│   │   ├── config.py           # Pydantic settings
│   │   └── dependencies.py     # DB, ChromaDB deps
│   ├── routers/
│   │   ├── collections.py      # Collection CRUD
│   │   ├── documents.py        # File upload / URL ingest
│   │   ├── query.py            # RAG query + SSE stream
│   │   └── health.py           # Health + metrics
│   ├── services/
│   │   ├── rag_service.py      # Core RAG pipeline
│   │   ├── ingestion_service.py # Document loading + chunking
│   │   ├── vector_store.py     # ChromaDB abstraction
│   │   └── llm_router.py       # Multi-LLM provider router
│   ├── models/
│   │   ├── database.py         # SQLAlchemy ORM
│   │   └── schemas.py          # Pydantic v2 schemas
│   └── workers/
│       └── celery_worker.py    # Async Celery tasks
│
├── web/                        # Flask frontend
│   ├── app.py                  # Flask app
│   └── templates/
│       ├── base.html           # Nav + layout
│       ├── index.html          # Dashboard
│       ├── chat.html           # Chat UI
│       ├── upload.html         # Upload UI
│       └── collections.html    # Collections UI
│
└── infra/
    ├── nginx/nginx.conf        # Reverse proxy config
    └── postgres/init.sql       # DB schema

---

🔧 Configuration

All settings are in .env. Key options:

Variable	Default	Description
DEFAULT_LLM_PROVIDER	openai	openai, anthropic, or ollama
CHUNK_SIZE	1000	Characters per document chunk
CHUNK_OVERLAP	200	Overlap between chunks
TOP_K_RESULTS	5	Number of chunks to retrieve
SIMILARITY_THRESHOLD	0.3	Min similarity score for retrieval
MAX_UPLOAD_SIZE_MB	50	Max file size in MB

---

🧪 Development

# Run without Docker (development)
cd api && pip install -r requirements.txt
uvicorn main:app --reload

# Run Celery worker
celery -A workers.celery_worker worker --loglevel=info

# Run Flask UI
cd web && pip install -r requirements.txt
python app.py

---

📄 License

MIT License — see LICENSE file for details.