Not getting any response from my local model though openclaw

[misty mauve]

That LM Studio log is the key clue: OpenClaw only ever calls:

POST  ${baseUrl}/chat/completions

So if LM Studio is seeing POST /chat/completions (missing /v1), then the baseUrl OpenClaw is actually using does not include /v1 (or it’s reading a different config than the one you edited).

1) First, verify what OpenClaw thinks your baseUrl is

Run and paste the vllm/lmstudio provider section it prints:

openclaw models status

(That output usually includes the provider baseUrl it will call.)

2) Make sure your config is in the right place (this is the common gotcha)

In current OpenClaw, local OpenAI-compatible providers are configured under:

• models.providers.<name>

not as a top-level vllm: { ... } block.

Use a config like this (JSON5), then restart the gateway:

{
  agents: {
    defaults: {
      model: { primary: "lmstudio/deepseek-coder-v2-lite-instruct" },
    },
  },

  models: {
    mode: "merge",
    providers: {
      lmstudio: {
        baseUrl: "http://127.0.0.1:1234/v1",
        apiKey: "lmstudio",
        api: "openai-completions",
        models: [
          {
            id: "deepseek-coder-v2-lite-instruct",
            name: "DeepSeek Coder V2 Lite (LM Studio)",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },

            // start smaller until it works, then raise
            contextWindow: 16384,
            maxTokens: 2048,
          },
        ],
      },
    },
  },
}

Restart:

openclaw gateway restart

After this, LM Studio should log POST /v1/chat/completions.

Docs that match this shape:

• https://docs.openclaw.ai/gateway/local-models
• https://docs.openclaw.ai/providers/vllm (same idea, different provider name)

3) If it hits the right endpoint but still times out

LM Studio + some models can be flaky with streaming. Try disabling streaming for that model:

agents: {
  defaults: {

    models: {
      "lmstudio/deepseek-coder-v2-lite-instruct": {
        params: { streaming: false },
      },
    },
  },
},

If you paste openclaw models status output, I can tell you immediately whether this is “wrong config location / wrong config file” vs a real LM Studio compatibility issue.

That “Unknown model” is expected with what you pasted: you changed the default model name, but OpenClaw still has zero models defined for the lmstudio provider.

You can see it in your models status:

• Configured models (0) → there is no model catalog entry for deepseek-coder-v2-lite-instruct
• Missing auth - lmstudio → also no provider config / apiKey for lmstudio

So OpenClaw has nothing to route lmstudio/deepseek-coder-v2-lite-instruct to.

Fix (the minimal working config)

Put your LM Studio endpoint under models.providers.lmstudio and define the model in the provider’s models: [...] list. Example:

{
  agents: {
    defaults: {

      model: { primary: "lmstudio/deepseek-coder-v2-lite-instruct" },
    },
  },

  models: {
    mode: "merge",
    providers: {
      lmstudio: {
        baseUrl: "http://127.0.0.1:1234/v1",
        apiKey: "lmstudio",          // any string, LM Studio usually ignores auth
        api: "openai-completions",
        models: [
          {
            id: "deepseek-coder-v2-lite-instruct",
            name: "DeepSeek Coder V2 Lite",

            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },

            // start conservative; you can raise later
            contextWindow: 16384,
            maxTokens: 2048,
          },
        ],
      },
    },
  },
}

Then restart:

openclaw gateway restart

After that:

• openclaw models status should stop showing “Missing auth - lmstudio”
• LM Studio should start seeing POST /v1/chat/completions (not /chat/completions)
• the “Unknown model” error should be gone

One important clue from your status output

This line matters a lot:

source=models.json: ~\.openclaw\agents\main\agent\models.json

It means your model setup may be coming from that per-agent models.json file (created by the wizard), not just ~\.openclaw\openclaw.json.

So if you edit openclaw.json and nothing changes, open and paste (redact keys if you want) the contents of: