FinSight
by RUC-NLPIR
Notebook-driven multi-agent pipeline for publication-ready financial reports
How It Works
Generates publication-ready financial research reports from a single ticker using multi-agent workflows and RAG pipelines. Coordinates specialized agents for data retrieval, analysis, and narrative drafting, then composes a polished report in a notebook-first workflow. Notable for its end-to-end finance focus and reproducible Jupyter demo that stitches retrieval, analysis, and summarization together. RAG pipelines and its architectural approach align with the Hierarchical multi-agent pattern.
Why It Matters
When to Use
Researchers and engineers prototyping finance-focused multi-agent RAG pipelines who want a reproducible demo for data-to-report workflows. This aligns well with the Blackboard Pattern approach to shared understanding and integration of results across agents.
How It's Used
- Prototyping end-to-end finance research: fetch market data, run analyses, and produce a report
- Evaluating different LLMs and retrieval stacks on the same financial task for comparison
- Teaching teams how to split finance workflows into specialized agents and add evaluation checkpoints