Indian IPO Analyzer
A multi-agent pipeline that reads live Indian IPO listings and writes a report where every number can be traced back to its source.
Context
IPO write-ups are either marketing or a wall of ratios with no explanation. A language model can explain ratios well, but left alone it will also invent them, and a wrong P/E in a buy note is worse than no note.
The idea
Split the work so the model never produces a figure. Five specialist agents (fundamentals, valuation, sentiment, risk profiling, synthesis) run as a LangGraph graph over scraped data. Ratios are computed in plain Python first; the agents only get to interpret them.
Details
- P/E, P/B, ROE, ROCE, debt to equity and CAGR are deterministic functions with unit tests, not prompts.
- Each agent writes into one shared state, so the synthesis step can only cite what an earlier node produced.
- Listings are scraped live and filterable by board (mainboard or SME) and status (open or upcoming).
- The report keeps the source of every number next to it, so a reader can check it without trusting the model.
What shipped
Shipped as a public Streamlit app. Pick any open or upcoming IPO and it returns a full report in one run. Try it.