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

# search_graph

> Keyword search over the knowledge graph node labels

## Overview

Performs fast keyword search over graph node labels (function names, class names, file names). Returns structural metadata — kind, source file, line range, degree — without embedding or vector lookup.

Use this tool for structural and navigation queries: "where is class X defined", "what files import Y", "find all methods named Z".

## Parameters

| Parameter | Type      | Default  | Description                                  |
| --------- | --------- | -------- | -------------------------------------------- |
| `query`   | `string`  | required | Keyword or identifier to search for          |
| `path`    | `string`  | required | Absolute path to the codebase root directory |
| `limit`   | `integer` | `20`     | Maximum number of results to return          |

## Example

```
search_graph("authenticate_user", "/path/to/myproject")
```

**Returns:**

```
[1] function authenticate_user
    File: src/auth.py  Lines: 45–72
    Score: 1.0  Degree: 7

[2] method authenticate_user
    File: src/legacy/auth.py  Lines: 110–138
    Score: 0.91  Degree: 3
```

## Return fields

| Field         | Description                                                      |
| ------------- | ---------------------------------------------------------------- |
| `kind`        | Node type: `file`, `function`, `class`, `method`                 |
| `source_file` | Path to the source file containing the node                      |
| `start_line`  | First line of the node                                           |
| `end_line`    | Last line of the node                                            |
| `score`       | Keyword match score (0–1)                                        |
| `degree`      | Number of edges connected to this node (higher = more connected) |

## Notes

* Requires `index_graph` to have been run first
* Average \~47 tokens per result set — \~99.8% reduction vs raw file reads
* Average latency \~3ms — significantly faster than vector search
* Use `search_code` instead when you need semantic / behaviour queries