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

# GET /memories/search — semantic memory search

> Query parameters, response schema, and request examples for the OMP semantic and keyword search endpoint, including min_score threshold filtering.

`GET /memories/search` returns a ranked list of memories that match a query string. Depending on the provider, the ranking is driven by vector similarity, keyword matching, or a hybrid of both. Each result includes a `score` field so you can threshold results by confidence.

<Note>
  This endpoint requires the `vector_search` or `keyword_search` capability to be supported by your provider. Call `GET /capabilities` first to confirm. If your provider does not support either, `GET /memories` with a `scope` or `tag` filter is the appropriate fallback.
</Note>

## Query parameters

<ParamField query="q" type="string" required>
  The search query. The provider runs semantic and/or keyword matching against all memories for the given `user_id`.
</ParamField>

<ParamField query="user_id" type="string" required>
  Return only memories belonging to this user. Search is always scoped to a single user.
</ParamField>

<ParamField query="scope" type="string">
  Restrict results to memories whose scope starts with this prefix (e.g. `coding` matches `coding/preferences` and `coding/tools`).
</ParamField>

<ParamField query="limit" type="integer" default="10">
  Maximum number of results to return. Maximum is 100.
</ParamField>

<ParamField query="min_score" type="number">
  Minimum similarity score (0–1) for a result to be included. Results with a `score` below this threshold are filtered out before the response is returned. Omit to return all results up to `limit`.
</ParamField>

## Response — 200 OK

<ResponseField name="results" type="SearchResult[]" required>
  Ranked list of matching memories, highest score first.

  <Expandable title="SearchResult properties">
    <ResponseField name="memory" type="Memory" required>
      The full memory record. See [Memory CRUD](/api/memories) for the complete field reference.
    </ResponseField>

    <ResponseField name="score" type="number" required>
      Cosine similarity score between 0 and 1. Higher values indicate stronger relevance to the query.
    </ResponseField>
  </Expandable>
</ResponseField>

## Example

```bash theme={null}
curl "http://localhost:8080/memories/search?q=package+manager&user_id=u1&limit=5"
```

```json theme={null}
{
  "results": [
    {
      "memory": {
        "id": "mem_abc123",
        "content": "User prefers pnpm over npm",
        "user_id": "u1",
        "scope": "coding/preferences",
        "tags": ["tooling", "nodejs"],
        "confidence": 0.92,
        "created_at": "2026-04-27T10:00:00Z",
        "updated_at": "2026-04-27T10:00:00Z",
        "status": "done"
      },
      "score": 0.94
    },
    {
      "memory": {
        "id": "mem_def456",
        "content": "User has used yarn in legacy projects",
        "user_id": "u1",
        "scope": "coding/preferences",
        "tags": ["tooling"],
        "confidence": 0.75,
        "created_at": "2026-03-10T09:00:00Z",
        "updated_at": "2026-03-10T09:00:00Z",
        "status": "done"
      },
      "score": 0.71
    }
  ]
}
```

### Filtering by minimum score

```bash theme={null}
curl "http://localhost:8080/memories/search?q=package+manager&user_id=u1&min_score=0.8"
```

Only results with a similarity score of 0.8 or above are returned — useful when you want high-confidence matches for LLM prompt injection.