KJStockScreener — AI-native NSE stock screener with a conversational trading agent, live option chain & ML gap prediction

Show the Community!
, , , , , , , , ,
1

Hey everyone :waving_hand:

I’ve been building this for a while and finally got it to a place I’m happy sharing — KJStockScreener, a Streamlit app for screening Indian (NSE) stocks that goes a bit further than a typical breakout scanner.

:rocket: Live demo: https://kjstockscreener.streamlit.app/
:laptop: Source (MIT licensed): GitHub - 85599/KJStockScreener: AI-native NSE stock screener — breakouts, live option chain & AI trading agent, built with Streamlit · GitHub

classic-screening
classic-screening1360×768 146 KB

What it does

  • Classic technical screening — multi-threaded scans across NSE indices for breakouts, consolidation, RSI reversals, MA/EMA crossovers, and candlestick patterns
  • An actual AI trading agent tab, built on openai-agents, where you chat in plain English instead of clicking through filters — e.g. “find me swing setups in Nifty 500 with RSI between 50 and 65.” Works with OpenAI, Anthropic, Groq, or any OpenAI-compatible endpoint (bring your own key)
  • 5 built-in trading personas (SwingTrader, MomentumAnalyst, OptionBuyer, ValueScreener, and a Nifty 50 technical-analysis persona), each with its own strategy — plus scheduled/cron-style agent runs
  • Live NSE option chain for NIFTY, BANKNIFTY and F&O stocks, with one-click TradingView links per strike
  • An ML model that predicts next-day Nifty gap-up/gap-down, shipped as a pure NumPy .npz file — no TensorFlow dependency, so it stays lightweight and Python 3.13-compatible

Why I’m posting here specifically

This community has genuinely shaped how I think about building with Streamlit — the multi-tab structure, session-state handling for the chat agent, and the way I stream the AI responses all came out of patterns I picked up reading other people’s posts here. Figured it was time to give something back.

Try it

Hosted demo needs no install — just open it. If you’d rather self-host:

docker pull callmejainsahab/kjstockscreener:latest
docker run -p 8501:8501 -p 8000:8000 callmejainsahab/kjstockscreener:latest

Or run from source — instructions are in the README.

I’d love feedback on

  • The AI Native chat tab’s UX — is the persona-switching intuitive, or confusing?
  • Performance on the option chain tab — I’m polling fairly aggressively for “live” data and would love thoughts on a better pattern
  • Anything that breaks for you on first run — please open an issue, I read all of them

If you find it useful, a :star: on the repo is hugely appreciated — and if you build something similar or have ideas for the roadmap, I’d love to hear about it in this thread.

Disclaimer: this is a research/educational tool, not investment advice — always verify signals yourself before trading.

Thanks for reading :folded_hands: