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

# Search

> Content search with lexical scoring and embedding integration

## SearchCells

Convenience wrapper over `QueryCells` for backward compatibility. Returns `[]CellSearchResult` sorted by score; each result includes a `Coord` for use as a context-pack seed.

```go theme={null}
results, err := tx.SearchCells(ctx, hexxladb.CellSearchConfig{
    Query:         "integration test",
    RequireTags:   []string{"fact"},
    MinConfidence: 0.5,
    SourceID:      "session-2",
    Center:        &centerCoord,
    MaxScanRadius: 5,
    Embedding:     queryVector,
    MaxResults:    10,
})
```

For `ExcludeTags`, `SortBy`, `Explain`, or temporal filters, use `QueryCells` directly.

## CellSearchConfig fields

| Field           | Description                                  |
| --------------- | -------------------------------------------- |
| `Query`         | Full-text search query                       |
| `RequireTags`   | Tags that must all be present (AND)          |
| `AnyTags`       | Tags where at least one must be present (OR) |
| `MinConfidence` | Minimum confidence threshold                 |
| `MaxConfidence` | Maximum confidence threshold                 |
| `SourceID`      | Filter by source ID                          |
| `Center`        | Spatial query center coordinate              |
| `MaxScanRadius` | Maximum ring radius to scan                  |
| `Embedding`     | Vector for ANN-accelerated seed selection    |
| `MaxResults`    | Maximum results to return                    |

## Content search scoring

| Condition                                      | Score contribution |
| ---------------------------------------------- | ------------------ |
| Query matches a tag exactly (case-insensitive) | +1.0               |
| Query is a prefix of a tag                     | +0.8               |
| Query found verbatim in `RawContent`           | +0.6               |
| Query found case-insensitively in `RawContent` | +0.5               |
| Query matches `SourceID` exactly               | +0.3               |
| Confidence bonus                               | +0.1 × Confidence  |

## CellSearchResult

```go theme={null}
type CellSearchResult struct {
    Cell  CellView
    Score float64
}
```

Each result's `Cell.Coord` can be used as a seed for context assembly.

## Multi-seed context assembly

A **seed** is a `Coord` — the centre point of a ring-walk expansion. `SearchCells` returns `CellSearchResult` values each carrying a `Coord`; those coords are the seeds passed to the assembly APIs, which expand each matched location's neighbourhood into context. One query → N matched coords → N seeds → one merged `ContextPack`.

```go theme={null}
// Search for relevant cells
results, err := tx.SearchCells(ctx, config)

// Extract seeds
seeds := make([]hexxladb.Coord, len(results))
for i, result := range results {
    seeds[i] = result.Cell.Coord
}

// Assemble context from multiple seeds
pack, err := tx.LoadContextPackFrom(ctx, 2, 4096, hexxladb.ByteLenBudgeter{}, hexxladb.LoadContextBudgetConfig{}, seeds...)
```

## See also

* [Query](/api/query) — Full query engine with more options
* [Context](/api/context) — Context assembly from seeds