An agentic equity research copilot for Taiwan-listed companies
Data → Analyst → Risk → Reporter — every conclusion ships with traceable citations.
Multi-agent collaboration
A Supervisor orchestrates Data / Analyst / Risk / Reporter with a Risk self-critique loop.
Hybrid RAG
BM25 + dense embedding + cross-encoder rerank over MOPS filings, transcripts, and notes.
Fully auditable
Every tool call and every revision is logged to Langfuse and streamed to the trace UI.