vtempest/ai-research-agent

Combines web search, multi-format content extraction (HTML, PDF, YouTube transcripts), and semantic topic modeling to build structured research datasets with LLM-generated answers and follow-up suggestions. Uses embedding-based reranking to identify and highlight the most topically relevant sentences within extracted content, mapping keyphrases to document structure via noun n-grams and vector similarity. Targets Node.js and serverless environments (Cloudflare Workers, Bun) with integrations for Groq, OpenAI, and other LLM providers to power crowdsourced knowledge graphs and AI-assisted research workflows.