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

# Search Depth

> Choose between precise, balanced, and exhaustive extraction modes

The `search_depth` parameter controls how Snippex's extraction engine approaches each page. It's a trade-off between **speed**, **coverage**, and **confidence**.

## Modes

<CardGroup cols={3}>
  <Card title="precise" icon="crosshairs">
    Focuses on the highest-confidence matches. Returns fewer snippets but with tighter relevance. Fastest mode.
  </Card>

  <Card title="balanced" icon="scale-balanced">
    The default. A good mix of speed, coverage, and confidence — suitable for most use cases.
  </Card>

  <Card title="exhaustive" icon="magnifying-glass">
    Maximises the number of snippets extracted per page. Best when you need comprehensive coverage at the cost of slightly more latency.
  </Card>
</CardGroup>

## When to use each

| Use case                          | Recommended depth |
| --------------------------------- | ----------------- |
| Quick answer / fact lookup        | `precise`         |
| General-purpose RAG pipeline      | `balanced`        |
| Research / comprehensive coverage | `exhaustive`      |
| Real-time UX (latency-sensitive)  | `precise`         |
| Offline batch processing          | `exhaustive`      |

## Example

```json theme={null}
{
  "query": "how does CRISPR gene editing work",
  "search_depth": "exhaustive",
  "num_results": 5
}
```

## Effect on `max_snippets_per_result`

`search_depth` works in conjunction with `max_snippets_per_result` (default: 3, max: 5). In `exhaustive` mode, the engine will try harder to find up to the maximum number of snippets per page. In `precise` mode, it may return fewer snippets than the maximum if confidence thresholds aren't met.