FinSight-AI

Why we included this project

FinSight AI is worth a look if you want an AI research agent whose output you can actually re-run and verify. It pairs a Spring Boot backend with pgvector-based hybrid retrieval to power a real A-share research workspace, and it treats long-running analysis jobs as recoverable: a task that dies halfway can be resumed and still end with a report tied to the exact data snapshot it started from. Redis single-flight leases stop duplicate requests from multiplying expensive model calls, and every conclusion keeps an inspectable evidence trail back to the filings and metrics that shaped it. Teams building their own analyst-style agents will find a solid engineering reference here, with recoverable workflows, snapshot-consistent reports, and RAG evaluation, rather than a thin chat wrapper.