> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nemoclaw/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemoclaw/_mcp/server.

# Configure Memory Search

> Configure OpenClaw memory search to use an embedding model from a host Ollama container.

Configure OpenClaw memory search to use an embedding model from a host Ollama container.
This setup lets another local server, such as managed vLLM, continue to serve the chat model.

## Prepare the Embedding Server

Start the host Ollama container with the embedding model.
This example uses `qwen3-embedding:4b` on port `11434`.

The sandbox reaches the container through `http://host.openshell.internal:11434`.
Keep the container bound to the local host or a reviewed host interface.
Port `11434` is commonly unauthenticated.
Allow it only from the OpenShell Docker bridge or an equivalent local container network.

## Permit the Bridge Route

The `local-inference` policy preset permits the sandbox to reach the host bridge.
Apply the preset, then verify the Ollama API from the sandbox.

```bash
nemoclaw my-assistant policy add local-inference --yes
nemoclaw my-assistant exec -- curl -fsS http://host.openshell.internal:11434/api/tags
```

## Configure the Embedding Provider

Open a bounded shields-down window before you change the OpenClaw configuration.
Define an Ollama provider, select it for memory search, restart OpenClaw, and restore lockdown.

Shields down temporarily relaxes selected lockdown controls for five minutes.
If a configuration command or restart fails, run `nemoclaw my-assistant shields up` before you retry.

```bash
nemoclaw my-assistant shields down \
  --timeout 5m \
  --reason "configure memory search"

nemoclaw my-assistant config set \
  --key models.providers.ollama-mem \
  --value '{"api":"ollama","baseUrl":"http://host.openshell.internal:11434","apiKey":"x","models":[{"id":"qwen3-embedding:4b","name":"Qwen3 embedding 4B"}]}' \
  --config-accept-new-path

nemoclaw my-assistant config set \
  --key agents.defaults.memorySearch.provider \
  --value ollama-mem \
  --config-accept-new-path \
  --restart

nemoclaw my-assistant shields up
```

Verify that Shields up restored lockdown:

```bash
nemoclaw my-assistant shields status
```

Continue only when the command succeeds and reports `Shields: UP (lockdown active)`.

The host-side `config set` command accepts the bridge URL only for supported provider `baseUrl` fields.
Generic configuration keys and other private URL shapes remain rejected.

## Rebuild and Verify the Index

An existing index can report an identity mismatch after its embedding provider changes.
Inspect the index, rebuild it, and verify the configured memory provider.

```bash
nemoclaw my-assistant exec -- openclaw memory status --index
nemoclaw my-assistant exec -- openclaw memory index --force
nemoclaw my-assistant exec -- openclaw memory status --deep
```

Add `--agent <id>` to each `openclaw memory` command for a non-default OpenClaw agent.

## Related Topics

* [Set Up Ollama](../inference/local-inference/set-up-ollama) for the host Ollama lifecycle and authenticated chat proxy.
* [Apply Policy Presets](../network-policy/configure-policies/apply-policy-presets) for managed network-policy access.
* [Understand Runtime Changes](../manage-sandboxes/configure-sandboxes/understand-runtime-changes) for shields and configuration mutations.