openclaw - 💡(How to fix) Fix [Bug]: If two types of api methods are used (API_KEY + localhost LLM) UI does not allow switching over. [1 participants]

Error: Failed to set model: GatewayRequestError: model not allowed: openai/deepseek-v3.2-lite-q6_k

Fix Plan

To resolve the issue of the UI not allowing switching between API methods, we need to update the openclaw.json configuration file to properly handle multiple LM and API endpoints.

Step-by-Step Solution

1. Edit openclaw.json: Open the openclaw.json file located at ~/.openclaw/ in a text editor.
2. Add Endpoint Configuration: Ensure that each endpoint is configured with a unique id and type. For example:

{
  "endpoints": [
    {
      "id": "openai-api",
      "type": "api",
      "apiKey": "YOUR_OPENAI_API_KEY",
      "models": ["openai/gpt-5.1-codex"]
    },
    {
      "id": "lm-studio",
      "type": "llm",
      "url": "http://localhost:8080",
      "models": ["openai/deepseek-v3.2-lite-q6_k"]
    }
  ]
}

1. Update Gateway Configuration: Verify that the gateway is configured to use the openclaw.json file. You can do this by checking the openclaw daemon configuration.
2. Restart OpenClaw Daemon: Restart the openclaw daemon to apply the changes.
3. Test Endpoint Switching: Connect to the gateway and test switching between endpoints.

Verification

To verify that the fix worked, try switching between the configured endpoints and verify that the UI allows you to choose which endpoint to use.

Extra Tips

• Ensure that the openclaw.json file is properly formatted and valid JSON.
• Verify that the API keys and URLs are correct for each endpoint.
• If issues persist, check the openclaw daemon logs for errors.