• 🧠 Smart Workflows: Multi-step reasoning and adaptive plan execution with LangGraph
• 🌊 Real-time Streaming: Live updates with Server-Sent Events (SSE)
• 🌐 Modern Web UI: Responsive frontend built with Next.js 15
• 🔄 Human-in-the-Loop: Interactive plan confirmation and feedback
• 🌐 Multi-language: Support for both English and Chinese
• 🔌 MCP Integration: Model Context Protocol for extended capabilities
• 📊 Multiple Formats: Text, charts, podcasts, and more
• 🐳 Docker Ready: Full containerization support

• Core: LangGraph, LangChain, FastAPI
• Package Manager: uv (recommended)
• Database: PostgreSQL (Supabase)
• Search: Tavily, DuckDuckGo, Brave Search, arXiv

• Framework: Next.js 15 (App Router)
• Language: TypeScript 5.0+
• Package Manager: pnpm
• UI: Tailwind CSS, Radix UI
• State: Zustand

• Python 3.12+
• Node.js 18+ and pnpm
• uv package manager (recommended)
• PostgreSQL database (Supabase recommended)

git clone https://github.com/lc708/YADRA-YetAnotherDeepResearchAgent.git
cd YADRA-YetAnotherDeepResearchAgent

Using bootstrap script (Recommended)

# Development mode with hot reload
./bootstrap.sh --dev

# Windows users
bootstrap.bat --dev

This will start:

• 🌐 Frontend: http://localhost:3000
• 🔌 Backend API: http://localhost:8000
• 📚 API Docs: http://localhost:8000/docs

Create .env file in project root:

# Database
DATABASE_URL=your_postgresql_url
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_key

# API
NEXT_PUBLIC_API_URL="http://localhost:8000/api"
DEBUG=True
APP_ENV=development

# Search
SEARCH_API=tavily
TAVILY_API_KEY=your_tavily_key

# Copy config template
cp conf.yaml.example conf.yaml

# Edit with your API keys
nano conf.yaml

Example conf.yaml:

BASIC_MODEL:
  base_url: https://api.openai.com/v1
  model: "gpt-4o"
  api_key: "your-openai-api-key"

REASONING_MODEL:
  base_url: https://api.anthropic.com
  model: "claude-3-5-sonnet-20241022"
  api_key: "your-anthropic-api-key"

1. Visit http://localhost:3000
2. Enter your research question
3. Configure research parameters (optional)
4. Send and watch real-time streaming responses
5. View generated artifacts, podcasts, and results

# Create research session
POST /research/create
{
  "query": "Your research question",
  "research_config": {
    "max_plan_iterations": 2,
    "max_step_num": 5,
    "auto_accepted_plan": false
  }
}

# Get session status
GET /research/status/{session_id}

# SSE streaming
GET /research/stream/{session_id}

python main.py [options] [query]

Options:
  --interactive              Interactive mode
  --debug                    Enable debug logging
  --max_plan_iterations N    Max plan iterations (default: 1)
  --max_step_num N          Max steps (default: 3)

# Local development (recommended)
./bootstrap.sh --dev

# Docker development
docker-compose up -d

Frontend: Deploy to Vercel
Backend: Use ./deploy.sh script for cloud deployment

For detailed deployment instructions, please refer to the configuration examples above and the deployment scripts in the repository.

YADRA/
├── src/              # Backend source
│   ├── server/       # FastAPI server
│   ├── graph/        # LangGraph workflows
│   ├── agents/       # AI agents
│   └── tools/        # Tools and utilities
├── web/              # Frontend source
│   ├── src/app/      # Next.js pages
│   ├── components/   # React components
│   └── core/         # Core functionality
└── scripts/          # Database and utility scripts

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Make your changes
4. Run tests (pytest tests/)
5. Commit your changes (git commit -m 'Add amazing feature')
6. Push to the branch (git push origin feature/amazing-feature)
7. Open a Pull Request

# Backend
ruff format src/
ruff check src/ --fix
pytest tests/

# Frontend
cd web
pnpm lint
pnpm format
pnpm test

This project is licensed under the MIT License.

Please see our Security Policy for reporting vulnerabilities.

• 🐛 Bug Reports: GitHub Issues
• 💡 Feature Requests: GitHub Issues
• 💬 Discussions: GitHub Discussions

YADRA stands on the shoulders of giants. We're grateful to these amazing open-source projects:

• LangGraph - Multi-agent workflow orchestration
• LangChain - LLM application framework
• LiteLLM - Unified LLM API interface

• Next.js 15 - React framework with App Router
• TipTap - Headless rich text editor
• Novel - Notion-style WYSIWYG editor
• Radix UI - Unstyled, accessible UI primitives
• Tailwind CSS - Utility-first CSS framework
• Framer Motion - Animation library
• Zustand - State management

• FastAPI - Modern Python web framework
• Supabase - Open-source Firebase alternative
• PostgreSQL - Advanced relational database
• uv - Fast Python package manager

• Tavily - AI-powered web search API
• DuckDuckGo Search - Privacy-focused search
• arXiv - Academic paper search
• Jina AI - Web content extraction

YADRA draws inspiration from these pioneering research projects:

• DeerFlow - Prompt templates and workflow patterns
• local-deep-researcher - Local AI research methodologies
• OpenDeepResearch - Open-source research frameworks

Special thanks to these projects for their prompt templates and research methodologies that helped shape YADRA's approach.

• Thanks to all contributors who make YADRA better
• Inspired by the open-source AI research community
• Built with love for researchers, by researchers