BacktestEngine Dashboard

A dark "quant terminal" for backtesting intraday NIFTY/SENSEX options strategies. This folder is self-contained — the backtesting engine, strategies, FastAPI backend, and React frontend all live here, so you can commit/clone just dashboard/.

dashboard/
  backtest/     the engine (replays a strategy day-by-day off cached candles)
  strategies/   strategy modules (example_strategy.py is the committed template)
  backend/      FastAPI — strategy discovery, background jobs, analytics, persistence
  frontend/     React + Vite + Tailwind v4 (Config → Running → Results → History → Compare)

Run it

# from this folder
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
cp .env.example .env          # add your Groww keys (only needed to FETCH uncached ranges)
cd frontend && npm install && cd ..
./run.sh                      # backend :8000 + frontend :5173 → open http://localhost:5173

Cached date ranges run fully offline (no creds). Uncached ranges are fetched from Groww on demand and cached to backtest/cache/ for next time.

The backend has no auto-reload — restart it after editing backend/*.py or backtest/*.py. The frontend hot-reloads.

What it does

• New backtest — pick a strategy, underlying, date range, capital, interval, compound; optionally override parameters. A badge shows whether the range is cached (offline) or needs Groww keys.
• Running — the run is a background job; the UI streams progress (per-day) and the engine log, with a Stop button.
• Results — summary cards, equity curve + underwater drawdown, a full risk panel, Yearly/Monthly-heatmap/Daily breakdown, time-of-day edge, and the order tape. A Gross / Net of STT toggle shows P&L before/after the 0.15% options STT.
• History — sortable leaderboard of past runs (running jobs show live; failed runs show their logs + a Retry; Clear-history button). Same Gross/Net-of-STT toggle.
• Compare — overlay two runs' equity curves with a stat-by-stat comparison.
• Pages are URL-addressable (#/history, #/compare, #/run/<id>) so refresh/back work.

Strategies

Strategy modules live in strategies/ and must follow the engine's convention: a module-level UNDERLYING = "NIFTY"|"SENSEX" and a main() the engine calls once per trading day, talking to the broker only through the openalgo client. UPPERCASE module-level literals become overridable parameters in the UI.

• strategies/example_strategy.py is a minimal, documented template — copy it.
• Add one in-app ("Add a strategy" on the New-backtest screen → paste/upload .py; validated by POST /api/strategies and written to strategies/), or drop a file in strategies/ manually.

Only example_strategy.py is committed — real strategies, the candle cache, results, tokens, and runtime persistence are gitignored (see .gitignore). Keep your alphas private.

API

Method	Path	Purpose
GET	/api/strategies	discovered strategies + cache coverage
POST	/api/strategies	validate + save a pasted/uploaded strategy
GET	/api/cache-status	is a range cached (offline) or needs keys
POST	/api/backtest	dispatch a background run → {job_id}
GET	/api/jobs/{job_id}	poll status / log / result
POST	/api/jobs/{job_id}/cancel	request cancellation
GET	/api/history	running jobs + persisted past runs
DELETE	/api/history	clear all history
GET	/api/runs/{run_id}	re-view a persisted run

Engineering notes

See CLAUDE.md for architecture details and the non-obvious gotchas (offline cache reuse, serialized jobs, lightweight-charts v5, STT model, hash routing).