NSE Trading Engine.

the problem

Algo-Bot kept missing SEBI circuit-breaker timing and retail APIs throttled mid-session. A production rewrite was the only way to trade inside compliance windows.

overview

Retail algorithmic trading in Indian equity markets lives inside a hard compliance envelope: SEBI circuit-breaker timings, broker-level throttles, per-symbol position limits. Miss one and the trade doesn't just lose money — it gets rejected, or worse, fills outside the window you meant. Algo-Bot taught me what retail APIs can't do. This is what it looked like to rewrite around those constraints.

how it works

• 01Electron desktop frontend — a trader needs a kill switch within reach, not behind a browser tab.
• 02FastAPI backend with an event-driven order loop — the loop itself enforces circuit-breaker windows; routes can't bypass them.
• 03Angel One SmartAPI integration with exponential backoff on throttle and idempotent order IDs on retry.
• 04Local SQLite position ledger — source of truth for P&L even when the broker API is lagging.

measured impact

what I'd change

Electron was the right call for the kill switch, wrong call for the resource footprint on older Windows machines. Next pass is probably Tauri — same UX, fraction of the memory, and the Rust backend would unify with the FastAPI rewrite I've been planning for the order loop.