HackathAttest Agent

Problem Statement

Problem StatementHackathons are a great way for devs to train their skills and build a portfolio. Unfortunately, the data around that is scattered in multiple organization, which prevents discovery around past projects and portfolio building.SolutionBlockchain acts as the universal API (and EAS in this specific case) and is therefore the perfect solution. Sprinkle some AI on top of it to digest the enormous amount of data and you get the Hackathon Attestations AI Agent. The Hackathon Attestations AI Agent is an Agent that uses RAG on past Hackathon attestations to help participants navigate past projects. It also uses votes and assessments from past hackathons to make their own judgement.

Solution

What we built🚀 Complete RAG Pipeline with On-Chain DataBuilt a comprehensive RAG system that mines, processes, and serves hackathon attestation data from the blockchain:Data Mining & Processing:Envio Integration: Used Envio's HyperSync to fetch massive amounts of EAS attestation events from ArbitrumData Enrichment: Scraped and processed 1,556 Devfolio attestations and 1,735 quadratic voting attestationsSmart Data Processing: Built custom pipelines to extract GitHub links, project descriptions, and hackathon metadataVector Database: Uploaded processed data into Qdrant for semantic search capabilitiesBackend Architecture:FastAPI Backend: Built robust REST API with/ragendpoint for query processingHybrid Search: Implemented both semantic and keyword search for optimal retrievalContext Management: Smart context processing and ranking for relevant attestation dataCaching Layer: Intelligent caching system for improved performanceFrontend Interface:Streamlit Dashboard: Created interactive web interface at https://hackathon-assistant.theguild.dev/Real-time Queries: Live RAG system for exploring hackathon projectsAnalytics Dashboard: Visual representation of attestation data and patterns🤖 Multi-Agent Architecture with ASI IntegrationAgentverse Deployment:RAG Client Agent: Built with uAgents framework for chat interface and query processingRetrieval Agent: Specialized agent for context retrieval with intelligent cachingAgent Communication: Implemented custom protocols for inter-agent communicationASI One Integration: Deployed agents on ASI's decentralized AI platformTechnical Implementation:uAgents Framework: Used for building decentralized agents with chat protocolsCustom Message Models: BuiltRetrieveAttestationRequestandRetrieveAttestationResponsemodelsAgent Discovery: Integrated with ASI One for automatic agent discovery and routingCaching Strategy: Implemented agent-level caching for improved performance🔧 Technology Stack & IntegrationCore Technologies:Envio HyperSync: For efficient blockchain data extractionQdrant Vector DB: For semantic search and similarity matchingFastAPI: High-performance backend with automatic API documentationStreamlit: Rapid frontend development for data visualizationuAgents: Decentralized agent framework for ASI integrationPartner Technology Benefits:Envio: Enabled us to fetch massive amounts of attestation data efficiently, processing thousands of events in minutesASI: Provided decentralized AI infrastructure, allowing us to deploy agents that can be discovered and used by anyone🎯 Notable Technical AchievementsHacky but Effective Solutions:Smart Data Extraction: Built custom regex patterns to extract GitHub URLs and project metadata from unstructured attestation dataHybrid Search Implementation: Combined semantic search with keyword matching for optimal retrieval accuracyAgent Caching: Implemented intelligent caching at the agent level to reduce API calls and improve response timesMulti-Agent Coordination: Created seamless communication between specialized agents for different tasksData Processing Pipeline:Processed 3,291 total attestations from ArbitrumExtracted and vectorized project descriptions, GitHub links, and hackathon metadataBuilt comprehensive search index for semantic similarity matchingImplemented real-time data updates and synchronizationDeployment Architecture:Live RAG System: https://hackathon-assistant.theguild.dev/Agentverse Agents: Deployed and discoverable through ASI OneMulti-Agent Communication: Seamless interaction between RAG client and retrieval agentsScalable Infrastructure: Built for handling multiple concurrent users and queries