ETHGlobalHackerAgent
Find sponsor implementation answers in seconds with multi-agent AI.
Project Description
This project provides an AI-powered documentation assistant for ETH Global hackathons. ETH Global Admins create hackathons and add sponsors, who then upload their documentation. The platform instantly indexes it and makes it available to all participants through a unified search interface. Hackers can ask implementation questions in natural language, and the AI responds based on the active hackathon’s sponsors, combining context from multiple docs using multi-agent reasoning (ASI-1 + MeTTa). By removing the friction of reading dozens of docs, it helps hackers build faster, integrate more sponsor tools, and ultimately increase high-quality project submissions for hackathon organizers and sponsors.
How it's Made
The frontend is built with Next.js 15, React 19, and Tailwind CSS, providing a live admin dashboard for managing hackathons, sponsors, and documents. Uploaded documentation is chunked, embedded using OpenAI’s text-embedding-3-small, and stored in Qdrant Cloud for vector search and semantic retrieval.
The backend uses a multi-agent architecture of four specialized agents:
🧩 Metadata Extractor Agent – parses and extracts keywords, stack, and domain from sponsor docs.
🧠 Query Understanding Agent – interprets user intent (e.g., wants code, framework, or explanation).
💬 Main Agent – orchestrates the pipeline, retrieves results from Qdrant, and generates answers using ASI-1.
⚙️ MeTTa Agent – applies symbolic reasoning to detect dependencies and enhance response accuracy.
Main agent is deployed on Fetch.ai Agentverse, communicating via the mailbox protocol, while a Python service on Render exposes the query-understanding API (agent).
This setup forms a production-ready Retrieval-Augmented Generation (RAG) pipeline with symbolic reasoning, combining neural and symbolic AI to make hackathon knowledge instantly accessible.
The result: hackers spend less time searching, more time building, and organizers see higher engagement and more sponsor-integrated project submissions.
Refer to the repo for more details and diagrams:
https://github.com/gilbertsahumada/agent-eth-global
