Biotact Production - AI-Powered RAG System

Intelligent knowledge base assistant for Biotact pharmaceutical company

🚀 Live Demo: https://core.biotact.uz 📧 Support: ruslan.biotact@gmail.com, x.temrjan@gmail.com

🎯 Overview

Production-ready RAG (Retrieval-Augmented Generation) system that provides intelligent responses based on Biotact's knowledge base using OpenAI embeddings and GPT-4o.

Key Features

✅ Smart RAG Pipeline: Semantic search with GPT-4o responses
✅ Multi-Department Support: Marketing, Sales, Medical, Support
✅ File Upload: Web interface for uploading .md documents
✅ Incremental Indexing: Only indexes new/changed documents
✅ User Management: Admin approval workflow
✅ Modern UI: React + TypeScript + Tailwind CSS
✅ Production Ready: Deployed on Ubuntu with Nginx + SSL

🏗️ Current Architecture

Nginx (SSL/TLS)
  ├── Frontend (React SPA) → /frontend/dist
  └── API (FastAPI) → localhost:8000
        ├── Qdrant (Vector DB) → localhost:6333
        ├── OpenAI Embeddings (text-embedding-3-large)
        └── OpenAI GPT-4o (LLM)

Note: Simple single-server deployment. No Docker, no Kubernetes, no microservices overhead.

🔧 Technology Stack

Backend

Framework: FastAPI 0.115.0
Language: Python 3.11+
Vector DB: Qdrant 1.12
AI/ML:
- OpenAI text-embedding-3-large (3072 dimensions)
- OpenAI GPT-4o (LLM)
RAG Framework: LlamaIndex 0.11.20

Frontend

Framework: React 18 + TypeScript
Build Tool: Vite 6
UI Components: shadcn/ui + Tailwind CSS
State: React Context API
Icons: Lucide React

Infrastructure

Web Server: Nginx
SSL: Let's Encrypt
OS: Ubuntu 22.04 LTS
Deployment: systemd services

📦 Quick Start

Prerequisites

# Required
- Ubuntu 20.04+ / Debian 11+
- Python 3.11+
- Node.js 18+
- 8GB+ RAM (for Qdrant)
- OpenAI API key

# Optional
- Nginx (for production)
- SSL certificate

Installation (Local Development)

1. Clone repository

git clone https://github.com/temrjan/biotact_core.git biotact-production
cd biotact-production

2. Setup environment

# Copy example .env
cp .env.example .env

# Edit .env and add your API keys
nano .env

Required environment variables:

# OpenAI
OPENAI_API_KEY=sk-proj-...
OPENAI_EMBEDDING_MODEL=text-embedding-3-large
OPENAI_EMBEDDING_DIMENSIONS=3072

# Qdrant
QDRANT_HOST=localhost
QDRANT_PORT=6333
QDRANT_COLLECTION_NAME=biotact_knowledge_v3

# URLs (for production)
API_URL=https://core.biotact.uz

3. Run setup script

chmod +x setup_local_simple.sh
./setup_local_simple.sh

This will:

Create Python virtual environment
Install backend dependencies
Install Qdrant
Setup directories

4. Install frontend

cd frontend
npm install
npm run dev  # Development mode
# OR
npm run build  # Production build

5. Prepare documents

# Copy your .md files to:
cp your-docs/*.md data/documents/

6. Index documents

# Activate venv
source venv/bin/activate

# Run incremental indexing
python scripts/incremental_index.py

7. Start backend

# Development
venv/bin/uvicorn backend.main:app --reload --host 0.0.0.0 --port 8000

# Production (with logs)
venv/bin/uvicorn backend.main:app --host 0.0.0.0 --port 8000 >> logs/backend.log 2>&1 &

8. Access the application

# Frontend (dev): http://localhost:5173
# Backend API: http://localhost:8000
# API Docs: http://localhost:8000/docs

📚 Usage

Admin Login

Email: <admin-email>
Password: <your-secure-password>

Upload Documents

Login as admin
Go to right sidebar
Drag & drop .md files or click to select
Wait for upload to complete
Click "Запустить индексацию" button
Wait 30-60 seconds for indexing

Query the System

Select department (Marketing, Sales, Medical, Support)
Type your question
System will:
- Search relevant documents (vector similarity)
- Generate response using GPT-4o
- Show sources used

🔐 API Endpoints

Public

GET / - Root endpoint
GET /health - System health check
GET /departments - List available departments
POST /auth/login - User authentication
POST /auth/register - User registration

Authenticated

POST /chat - RAG chat with AI
POST /query - RAG query
POST /search - Vector search only
GET /stats - System statistics
POST /upload - Upload .md files (admin only)
POST /index - Run incremental indexing (admin only)

Admin Only

GET /auth/admin/users/pending - Pending users
POST /auth/admin/users/{id}/approve - Approve user
POST /auth/admin/users/{id}/reject - Reject user

API Documentation: http://localhost:8000/docs

🗂️ Project Structure

biotact-production/
├── backend/
│   ├── main.py              # FastAPI app + RAG system
│   ├── config/
│   │   └── departments.py   # Department configurations
│   └── prompts/
│       ├── marketing.txt    # Marketing prompt
│       ├── sales.txt        # Sales prompt
│       ├── medical.txt      # Medical prompt
│       └── support.txt      # Support prompt
├── frontend/
│   ├── src/
│   │   ├── components/      # React components
│   │   ├── contexts/        # React contexts
│   │   ├── api/            # API client
│   │   └── types/          # TypeScript types
│   ├── dist/               # Production build
│   └── package.json
├── data/
│   ├── documents/          # 📁 Upload .md files here
│   ├── .index_state.json   # Indexing state tracker
│   └── qdrant/            # Qdrant data (ignored by git)
├── scripts/
│   └── incremental_index.py # Incremental indexing script
├── docs/
│   └── INDEXING_GUIDE.md   # Indexing documentation
├── logs/                   # Application logs
├── venv/                   # Python virtual environment
├── .env                    # Environment variables (not in git)
├── .env.example           # Example environment variables
└── requirements.txt        # Python dependencies

📖 Documentation

Indexing Guide - Detailed indexing documentation
Indexing Quick Start - Quick reference
Deployment Status - Current deployment state
API Docs - Interactive API documentation

🔄 Incremental Indexing

The system uses hash-based incremental indexing to avoid reprocessing unchanged files.

How it works:

Calculates MD5 hash of each file
Compares with previous state (.index_state.json)
Only indexes new/modified files
Removes vectors for deleted files

Run indexing:

# Manual
venv/bin/python scripts/incremental_index.py

# Via API (admin only)
curl -X POST https://core.biotact.uz/index

See: INDEXING_GUIDE.md

🚀 Production Deployment

1. Install Nginx

sudo apt update
sudo apt install nginx

2. Configure Nginx

# Copy nginx config
sudo cp infrastructure/nginx/core.biotact.uz.conf /etc/nginx/sites-available/core.biotact.uz
sudo ln -s /etc/nginx/sites-available/core.biotact.uz /etc/nginx/sites-enabled/

# Test config
sudo nginx -t

# Reload
sudo systemctl reload nginx

3. Setup SSL (Let's Encrypt)

sudo apt install certbot python3-certbot-nginx
sudo certbot --nginx -d core.biotact.uz

4. Setup systemd service

# Create service file
sudo nano /etc/systemd/system/biotact-backend.service

[Unit]
Description=Biotact Backend API
After=network.target

[Service]
Type=simple
User=temrjan
WorkingDirectory=/home/temrjan/project/biotact-production
Environment="PATH=/home/temrjan/project/biotact-production/venv/bin"
ExecStart=/home/temrjan/project/biotact-production/venv/bin/uvicorn backend.main:app --host 0.0.0.0 --port 8000
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

# Enable and start
sudo systemctl daemon-reload
sudo systemctl enable biotact-backend
sudo systemctl start biotact-backend

🧪 Testing

Health Check

curl http://localhost:8000/health
# {"status":"healthy","postgres":true,"qdrant":true,"embeddings":true,"llm":true}

Test RAG Query

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Что такое биотакт?",
    "department": "marketing"
  }'

Check Qdrant

curl http://localhost:6333/collections/biotact_knowledge_v3

📊 System Requirements

Minimum (Development)

CPU: 2 cores
RAM: 4GB
Disk: 20GB
Bandwidth: 10 Mbps

Recommended (Production)

CPU: 4 cores
RAM: 8GB
Disk: 50GB SSD
Bandwidth: 100 Mbps

Current Deployment

Server: Ubuntu 22.04 LTS
CPU: 2 vCPU
RAM: 8GB
Disk: 80GB
Vectors in DB: 45
Documents: 5 .md files

🛠️ Troubleshooting

Backend won't start

# Check logs
tail -50 logs/backend.log

# Check port
lsof -i :8000

# Check Qdrant
curl http://localhost:6333/health

Frontend build fails

cd frontend
rm -rf node_modules package-lock.json
npm install
npm run build

Indexing errors

# Check documents directory
ls -la data/documents/

# Check Qdrant connection
curl http://localhost:6333/collections

# Re-run indexing with verbose output
venv/bin/python scripts/incremental_index.py

Qdrant connection errors

# Check Qdrant process
ps aux | grep qdrant

# Restart Qdrant
pkill qdrant
# Start Qdrant (see setup script)

🔒 Security Notes

⚠️ Current Implementation (Development):

User passwords in plain text (in-memory dictionary)
Simple token authentication (not JWT)
No rate limiting
No input validation

✅ TODO for Production:

📈 Performance

Current Metrics:

Response time: ~3-5 seconds (GPT-4o generation)
Concurrent users: ~10-20
Vector search: <100ms
Embedding generation: ~500ms
Total throughput: ~5-10 requests/minute

Optimization Tips:

Use GPT-4o-mini for faster/cheaper responses (17x cheaper)
Increase similarity_top_k for better context
Add Redis caching for frequent queries
Use streaming responses for better UX

🤝 Contributing

This is a private production system. For authorized team members:

Create feature branch: git checkout -b feature/your-feature
Make changes and test locally
Commit: git commit -m "feat: your feature"
Push: git push origin feature/your-feature
Create Pull Request

📄 License

Unauthorized copying, modification, or distribution is prohibited.

👥 Team

Project Lead: Temrjan (@temrjan)
Admin Contacts:
- ruslan.biotact@gmail.com
- x.temrjan@gmail.com

📞 Support

Issues: Create issue on GitHub Email: support@biotact.com Telegram: @biotact_support

🎯 Roadmap

✅ Completed (v1.0)

🚧 In Progress (v1.1)

PostgreSQL user storage
Better error handling
Response caching
Analytics dashboard

📋 Planned (v2.0)

Version: 1.0.1 Last Updated: November 20, 2025 Status: ✅ Production Ready

FilesExpand file tree

README.md

Latest commit

History