Steward includes a built-in BM25-based search engine with relevant scoring and typo tolerance that is significantly faster than the native Obsidian search.

## Before you start

You need to build the search index first. Open the chat and run:

```
/ Build search index
/ Build search index from the Projects folder
```

The index is built and stored locally in your vault. You only need to do this once; the index updates automatically as you modify your notes.

## Two search modes

### 1. Search without AI

Wrap your query in quotation marks or search by tags only for instant, direct results:

```
/search "quarterly report"
/search #project #urgent
```

This searches for either exact matches or relevant results based on what you type — no AI processing needed.

### 2. Search with AI

Type `/search` followed by your query in natural language:

```
/search Note names start with "2026"
/search All PDF files in the Media folder
```

Steward uses the AI to break your query into keywords, file names, folders, tags, and frontmatter properties to find the most relevant notes.

#### Property operators

You can search frontmatter properties with comparison operators:

| Operator | Description           |
| -------- | --------------------- |
| `==`     | equals (default)      |
| `!=`     | not equal             |
| `>`      | greater than          |
| `<`      | less than             |
| `>=`     | greater than or equal |
| `<=`     | less than or equal    |

```
/search Notes with priority greater than 3
/search Notes where rating is at least 4
/search Notes are created last week
/search Tasks with status not equal to "done"
/search Tasks are still within due date
```

## Tips

- Use **tags** or wrap your queries in **quotation marks** for instant results (no AI round-trip).
- Full results are stored in an **artifact** that you or Steward can reference later — for example, to [edit](Edit.md) or perform bulk operations on the matched notes.