Built for TiDB AgentX Hackathon 2025: Forge Agentic AI for Real-World Impact
DevAgent transforms any GitHub repository into an interactive AI assistant. Instead of navigating thousands of files, simply provide a repository URL and start asking questions about code architecture, security vulnerabilities, deployment strategies, and implementation details.
DevAgent combines TiDB's distributed vector database for lightning-fast semantic code search with Kimi LLM's web search capabilities to provide both deep code analysis and up-to-date deployment insights. Common questions are cached for sub-100ms responses, while complex queries leverage real-time web search.
- TiDB Account – free tier available at tidbcloud.com
- Kimi API Key – get yours at moonshot.cn
- Modal Account – serverless compute at modal.com
-
Clone the repository and install dependencies:
git clone https://github.com/yourusername/devagent.git cd devagent pip install -r requirements.txt -
Configure environment variables:
cp .env.example .env
Edit
.envwith your credentials:DATABASE_URL=your_tidb_connection_string KIMI_API_KEY=your_kimi_api_key
-
Deploy backend and run frontend:
- Setup Modal secrets:
modal secret create tidb-secret DATABASE_URL=your_tidb_url modal secret create kimi-secret KIMI_API_KEY=your_kimi_key
- Deploy serverless backend:
modal deploy agent_core.py
- Run Streamlit frontend:
streamlit run app.py
- Setup Modal secrets:
The system automatically handles repository ingestion, vector embedding, and intelligent caching. Visit the Streamlit URL to start analyzing repositories with AI.
- 🧠 Smart Code Analysis – Deep understanding of architecture and patterns
- 🌐 Real-Time Web Search – Current deployment guides and best practices
- ⚡ Ultra-Fast Responses – Sub-100ms cached answers for common questions
- 🛡️ Security Auditing – Automated vulnerability detection with suggested fixes
- 🔧 Function Explorer – Detailed explanations of specific code components
- Private repository support with GitHub authentication
- VS Code extension for seamless IDE workflow
- Team collaboration features with shared insights and annotations
- Multi-language expansion beyond current supported programming languages
├── .env.example # Environment variables template
├── agent_core.py # Modal serverless backend with AI agents
├── app.py # Streamlit frontend application
├── requirements.txt # Python dependencies
└── README.md # Project documentation
MIT License © 2025 Pratik Shah