An autonomous AI agent system that integrates Jira and GitHub to automate software engineering tasks. This system listens for Jira assignments, generates code changes using DeepSeek, submits Pull Requests, and handles feedback loops automatically.
- Autonomous Coding: Uses DeepSeek and LangGraph to analyze codebases and implement features/fixes.
- Jira Integration: Automatically picks up assigned issues and transitions them to "In Progress" or "Done".
- GitHub Automation: Creates branches, commits code, and opens Pull Requests.
- Real-time Dashboard: A React-based UI to monitor task status (Pending, In Progress, Completed, Failed).
- Feedback Loop: Automatically handles PR review comments or manual triggers via
[agent-fix]comments. - Reliable Queueing: PostgreSQL-backed task queue with
LISTEN/NOTIFYand polling fallbacks.
The system follows a modular MVC-like pattern:
- Backend: Node.js/Express with modular controllers and middlewares.
- Agent: LangGraph-orchestrated worker using DeepSeek LLM.
- UI: Vite-powered React dashboard for observability.
- Database: PostgreSQL for persistent task management.
For a deep dive into the technical workflow, see Design/system_workflow.md.
- Node.js (v18+) & Docker
- DeepSeek API Key
- GitHub Personal Access Token
- Jira API Token
Start the database using Docker Compose:
docker-compose up -dcd Backend
npm install
npm run devcd UI
npm install
npm run devUse ngrok to expose your local server (port 3000) to receive webhooks:
ngrok http 3000Register the ngrok URL in your Jira Webhooks and GitHub repository settings.
- GEMINI.md - Detailed setup, environment variables, and development guidelines.
- Design/system_workflow.md - Technical walkthrough of the agent's lifecycle.
- Check DB Status:
docker exec -i coding_automation_db psql -U agentuser -d agentdb -c "SELECT * FROM agent_instructions ORDER BY id;"
- Deploy to Vercel:
vercel(Requires environment variable configuration).