| name | ngat-brokers |
| description | Implement and maintain broker connectivity for Zerodha in brokers/ including auth, session retry behavior, order/position APIs, and historical data sync helpers. Use when touching login flow, order execution, broker resilience, or market data backfill behavior. |
Ngat Brokers
Overview
Own the boundary between internal engine components and Zerodha endpoints. Use this skill for reliability and security improvements in broker-facing logic.
Submodules
brokers/utils.py: TOTP generation, login, cache read/write (.keystore.json), token validation.brokers/kite_trade.py: ZerodhaBroker REST wrapper and retry behavior.brokers/zerodha.py: historical backfill and realtime sync into RedisTimeSeries.
Pros
- Keep provider-specific behavior isolated from core strategy code.
- Reuse cached
enctoken and refresh on retriable auth failures.
- Backfill and realtime sync share one candle ingestion path.
Cons
- Keep direct
os.system installs in import path (kite_trade.py) which is unsafe in production.
- Keep secret token cache in project root without encryption.
- Keep historical loops broad and potentially expensive when partitions are unset.
Caveats
- Never log secrets;
print statements currently risk leaking operational internals.
- Treat retry login and request replay as non-idempotent for order endpoints.
- Treat timeframe mapping carefully (
day vs intraday intervals) to avoid API errors.
- Handle Zerodha API limits explicitly; current code relies on chunking but not adaptive throttling.
Forward Features
- Add secure credential management (OS keychain or vault) for token cache.
- Add idempotency keys and order state reconciliation layer.
- Add broker abstraction coverage for alternate providers beyond Zerodha.
