# QuantTrade

Local-first quantitative backtesting on Coolify: Streamlit UI, VectorBT engine, Parquet market data, nightly Yahoo Finance sync, in-app Authentik OIDC, and SQLite strategy persistence.

## Architecture

| Layer | Technology |
|-------|------------|
| Auth | Authentik OIDC inside Streamlit (`OIDC_CLIENT_SECRET` in app env) |
| UI | Streamlit (`streamlit` service, port 8501) |
| Engine | VectorBT + NumPy |
| Market data | Parquet volume (`parquet-data`) |
| Ingestion | `harvester` cron @ 17:00 America/New_York (weekdays) |
| Strategies | SQLite on `strategy-data` volume, keyed by OIDC username |
| Telemetry | Bugsink via `sentry-sdk` (`bugsink.aexoradao.com`) |

## Coolify deployment

1. Create a **Docker Compose** resource pointing at this repo.
2. Assign your public domain to the **`streamlit`** service on port **8501**.
3. **Do not** enable Authentik forward auth on the Coolify proxy for this app — login happens inside Streamlit.
4. Set environment variables in Coolify:
   - `OIDC_CLIENT_SECRET` — **required**; copy from your Authentik provider
   - `OIDC_CLIENT_ID` — Authentik provider slug (default `quant-web`)
   - `OIDC_ISSUER` — Authentik issuer URL for the provider
   - `OIDC_REDIRECT_URI` — optional; defaults to `SERVICE_URL_STREAMLIT_8501`
   - `BUGSINK_DSN` — DSN from your Bugsink project
   - `AUTH_REQUIRED` — keep `true` in production
   - `DEV_USER` — only when `AUTH_REQUIRED=false` for local testing

**Coolify note:** if you change any default after the first deploy, update the value manually in Coolify UI > Environment Variables.

### Authentik provider setup

In Authentik, create an **OAuth2/OIDC provider** for this app:

1. **Client type:** Confidential
2. **Client ID:** `quant-web` (or match `OIDC_CLIENT_ID`)
3. **Client secret:** paste into Coolify as `OIDC_CLIENT_SECRET`
4. **Redirect URIs:** your public app URL, e.g. `https://quant.example.com`
   - Must match `OIDC_REDIRECT_URI` or the Coolify-generated `SERVICE_URL_STREAMLIT_8501`
5. **Signing key:** RS256 (Authentik default)

Users hit the app → Streamlit redirects to Authentik → callback with auth code → app exchanges code using `OIDC_CLIENT_SECRET` → userinfo drives strategy ownership.

## Services

- **`data-seed`** — one-shot 5-year historical download into Parquet (idempotent).
- **`harvester`** — cron container; appends daily bars after US cash close.
- **`streamlit`** — dashboard, OIDC login, backtests, save/load strategies.

## Local development

```bash
cp .env.example .env
# Set AUTH_REQUIRED=false and DEV_USER for local testing without Authentik
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python sync.py --seed
streamlit run app.py
```

For full OIDC locally, set `OIDC_CLIENT_SECRET` and register `http://localhost:8501` as a redirect URI in Authentik.

## Research workflow

1. **Backtest** — run a single parameter set and inspect equity curve, drawdown, trades.
2. **Optimize** — exhaustive parameter scan (grid search) ranked by Sharpe, Sortino, return, or drawdown.
3. **Python** — view builtin source or author custom strategies with `generate_signals()`.
4. **Library** — save/load strategies per user (SQLite), including custom Python source code.

### Custom Python strategy contract

```python
PARAM_GRID = {"fast_window": list(range(10, 41, 5)), "slow_window": list(range(50, 151, 10))}
DEFAULT_PARAMS = {"fast_window": 20, "slow_window": 50}

def generate_signals(close, high, low, volume, **params):
    # return boolean entry/exit Series aligned to close
    return entries, exits
```

Builtins: `ma_crossover` (vectorized VectorBT comb scan), `rsi_reversion` (grid scan).

## Docker build speed

- Harvester image installs only `requirements-harvester.txt` (no VectorBT/Streamlit).
- App image uses BuildKit pip cache (`RUN --mount=type=cache`).
- `.dockerignore` keeps git/cache out of build context.

Enable BuildKit on Coolify/build host for cache mounts.


```bash
python sync.py --seed    # full history
python sync.py --daily   # append latest bars
```

## Strategy storage

SQLite path: `/data/strategies/strategies.db` (Docker volume `strategy-data`). Each row stores username, name, ticker, and JSON parameters (MA windows, cash, fees).

## Bugsink

Set `BUGSINK_DSN` to your project DSN, e.g.:

```text
http://<public-key>@bugsink.aexoradao.com/<project-id>
```

Both `app.py` and `sync.py` initialize the Sentry-compatible SDK with tracing disabled for Bugsink compatibility.