Agentify.

Agentify is a production-grade discovery platform for developers seeking the best AI agents, Claude skills, MCP tools, and agentic development resources. A 5-agent LangGraph pipeline continuously scrapes GitHub repositories, Reddit threads, HuggingFace model pages, and developer blogs, then parses, embeds, evaluates, and surfaces the most relevant tools — personalized per developer based on their stack, goals, and history. The retrieval layer combines BM25 keyword search with pgvector semantic search, re-ranked by a HuggingFace Text Ranking model for p95 latency under 200ms. An automated eval harness measures Precision@5, NDCG, and MRR against a golden test set. LLMOps observability runs through LangSmith; scheduled ingestion pipelines run on Prefect — all on free-tier infrastructure. The system demonstrates every pattern expected of a mid-level AI engineer in 2026: RAG, multi-agent orchestration, hybrid retrieval, LLM evaluation, and LLMOps observability.

• 015-agent LangGraph pipeline: scrape → parse → embed → evaluate → surface
• 02Hybrid BM25 + pgvector retrieval re-ranked with HuggingFace Text Ranking model; p95 latency < 200ms
• 038 HuggingFace task types integrated across the production pipeline
• 04LLM eval harness with Precision@5, NDCG, and MRR on a 50+ query golden test set
• 05LLMOps observability via LangSmith; scheduled ingestion via Prefect — entirely free-tier infrastructure

• 500+Agents Indexed
• 5LangGraph Agents
• 8HF Task Types
• > 0.80Target P@5

• Live demo