This is a multi-agent AI system built for the Oosc3.0 hackathon. It provides real-time talking points, action items, and summaries to a sales representative during a live meeting, with a focus on transparency and provenance.
Click Here to Watch the Full Demo Video on Google Drive
- Multi-Agent Architecture: A system of 17 independent microservices collaborating via a Redis event bus.
- Real-Time Intelligence: Generates talking points, action items, and a final summary using a Large Language Model (LLM).
- Transparent UI: A professional, dark-themed dashboard that visualizes the agent workflow in real-time with a live event log and progress bar.
- Provenance & Auditability: A dedicated Logger agent captures every event for a complete audit trail.
- Built for Resilience: Gracefully handles API failures with a fallback system.
- Backend: Python, FastAPI, Redis
- Frontend: React.js
- AI: OpenRouter API (Nous Hermes 2 Mixtral & Google Gemini Pro)
- Communication: Server-Sent Events (SSE)
- Ensure you have Python, Node.js, and Redis installed and running.
- Clone the repository.
- For each agent in the
backend/directory, runpip install -r requirements.txt. - Create a
.envfile in each agent directory that requires it with yourOPENROUTER_API_KEY. - Start all agents using the provided
start_agents.pyscript, or start each of the 17 backend agents manually usinguvicorn main:app --reload. - In a separate terminal, navigate to
frontend/and runnpm start.
- Member-1 Name: Ayush Singh (Backend Developer, Domain Expertise)
- Member-1 Email-ID: [email protected]
- Member-2 Name: Lucky Tiwari (Front-End Developer)
- Member-2 Email-ID: [email protected]
- Member-3 Name: Saksham Tiwari (Documentation)
- Member-3 Email-ID: [email protected]
- Member-4 Name: Satyam Shukla (Testing)
- Member-4 Email-ID: [email protected]
- Member-5 Name: Aman Mishra (Project Manager)
- Member-5 Email-ID: [email protected]