> ## Documentation Index
> Fetch the complete documentation index at: https://mathematicalcompany.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Sentinel

> Always-on portfolio risk monitoring: correlation alerts, graduated drawdown response, tail risk hedging, regime-aware risk budgets.

<Warning>
  **Ultra Feature.** Requires an Ultra subscription. [Get started at api.mathematicalcompany.com](https://api.mathematicalcompany.com)
</Warning>

# Sentinel

Continuous portfolio risk surveillance that runs inside `hz.run()` or standalone. Monitors correlations, manages drawdowns with graduated responses, detects regime changes, and suggests hedges.

<CardGroup cols={2}>
  <Card title="Drawdown Management" icon="arrow-trend-down">
    4-level graduated response: alert → reduce → pause → exit.
  </Card>

  <Card title="Regime Detection" icon="chart-line">
    Classifies volatility regime and adjusts risk budget multiplier.
  </Card>

  <Card title="Correlation Alerts" icon="diagram-project">
    Ledoit-Wolf shrinkage detects correlation spikes across positions.
  </Card>

  <Card title="Hedge Suggestions" icon="umbrella">
    Delta-based hedge recommendations from prediction Greeks.
  </Card>
</CardGroup>

***

## Standalone Report

```python theme={null}
import horizon as hz

engine = hz.Engine()
report = hz.sentinel_report(engine)
print(report.regime)          # "normal", "high_vol", "crisis", etc.
print(report.var_95)          # Cornish-Fisher VaR at 95%
print(report.risk_budget)     # Budget utilisation with regime multiplier
```

***

## Hedge Suggestions

```python theme={null}
hedges = hz.suggest_hedges(engine, max_hedges=3, min_reduction_pct=5.0)
for h in hedges:
    print(f"{h.market_id}: {h.side} {h.size} @ {h.price} → {h.risk_reduction_pct}% reduction")
```

***

## Pipeline Mode

```python theme={null}
hz.run(
    name="guarded-strategy",
    markets=["mkt-a", "mkt-b"],
    pipeline=[
        hz.sentinel(config=hz.SentinelConfig(
            auto_deleverage=True,
            drawdown_thresholds=[
                (-0.05, "alert"),
                (-0.10, "reduce"),
                (-0.20, "pause"),
                (-0.30, "exit"),
            ],
        )),
        model,
        quoter,
    ],
    interval=1.0,
)
```

Injects `ctx.params["sentinel_report"]` every N cycles.

***

## Graduated Drawdown Response

| Threshold | Action   | Behaviour                                         |
| --------- | -------- | ------------------------------------------------- |
| `-5%`     | `alert`  | Log warning only                                  |
| `-10%`    | `reduce` | Cancel 50% of open orders (largest first)         |
| `-20%`    | `pause`  | Cancel all open orders                            |
| `-30%`    | `exit`   | Cancel all + submit exit orders for all positions |

***

## Regime Detection

Portfolio volatility determines the regime and risk budget multiplier:

| Regime     | Volatility | Multiplier            |
| ---------- | ---------- | --------------------- |
| `low_vol`  | \< 5%      | 1.5× (take more risk) |
| `normal`   | 5-15%      | 1.0×                  |
| `high_vol` | 15-30%     | 0.5× (reduce risk)    |
| `crisis`   | > 30%      | 0.25× (minimal risk)  |

***

## SentinelConfig

| Parameter                     | Default                  | Description                                |
| ----------------------------- | ------------------------ | ------------------------------------------ |
| `drawdown_thresholds`         | `[(-0.05,"alert"), ...]` | Graduated response levels                  |
| `correlation_spike_threshold` | `0.3`                    | Flag when correlation jumps by this        |
| `auto_hedge`                  | `False`                  | Automatically execute hedge suggestions    |
| `auto_deleverage`             | `True`                   | Automatically reduce positions on drawdown |
| `regime_sensitivity`          | `1.0`                    | Scaling factor for regime thresholds       |
| `var_confidence`              | `0.95`                   | VaR confidence level                       |
| `lookback_returns`            | `100`                    | Rolling window for return calculations     |
| `risk_budget_total`           | `10000.0`                | Total risk budget in USDC                  |