Market Research AI

~95% of the research loop automated

The problem

Multi-agent LangChain + Gemini pipeline (BYOK) scanning 1000+ data points on the AI/LLM market into company reports and investment proposals; ~95% of the manual research loop automated. (Forked base, substantially extended.)

The constraint

First contact with multi-agent orchestration: where do agent boundaries belong? The answer that worked — follow output artifacts (report sections), not 'roles'.

What the git log says

• $ Automated market research via a web-browser tool; AI use-case generation; resource discovery.
• $ Final proposal generation summarizing research, use cases, and resources.
• $ BYOK model — no server-side key custody simplifies the whole architecture.

What I learnt

Agent boundaries should follow output artifacts, not roles. BYOK shapes architecture — no key custody simplifies everything.