Gibberish output | Friends of the Crustacean 🦞🤝 | Page 1

dreamy quail Apr 26, 2026, 10:11 AM

#

I have hosted meta-llama/Llama-3.3-70B-Instruct model on vLLM on a custom target target machine [Linux server]. The model is served at 192.168.75.81:18789.

The openclaw gateway is located on my windows local machine. Now, I have below config.json file to setup openclaw on my local windows to connect to host model.

{"gateway":{"auth":{"mode":"token","token":""},"mode":"local","port":18789,"bind":"loopback","tailscale":{"mode":"off","resetOnExit":false}},"meta":{"lastTouchedVersion":"2026.4.22","lastTouchedAt":"2026-04-24T10:28:17.046Z"},"agents":{"defaults":{"imageGenerationModel":{"primary":"google/gemini-2.5-flash-preview-05-20"},"model":{"primary":"vllm/meta-llama/Llama-3.3-70B-Instruct"},"models":{"vllm/meta-llama/Llama-3.3-70B-Instruct":{},"google/gemini-3.1-flash-lite-preview":{},"google/gemini-2.5-flash-preview-05-20":{},"google/gemini-2.5-flash-preview-04-17":{}},"workspace":"C:\Users\mcw\.openclaw\workspace"}},"browser":{"enabled":true,"defaultProfile":"openclaw"},"nodeHost":{"browserProxy":{"enabled":true}},"wizard":{"lastRunAt":"2026-04-24T10:28:16.736Z","lastRunVersion":"2026.4.22","lastRunCommand":"configure","lastRunMode":"local"},"auth":{"profiles":{"google:default":{"provider":"google","mode":"api_key"},"vllm:default":{"provider":"vllm","mode":"api_key"}}},"plugins":{"entries":{"google":{"enabled":true,"config":{"webSearch":{"apiKey":""}}},"browser":{"enabled":true},"vllm":{"enabled":true}}},"models":{"mode":"merge","providers":{"vllm":{"baseUrl":"http://192.168.75.81:8000/v1","api":"openai-completions","apiKey":"VLLM_API_KEY","models":[{"id":"meta-llama/Llama-3.3-70B-Instruct","name":"meta-llama/Llama-3.3-70B-Instruct","reasoning":false,"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"contextWindow":16384,"maxTokens":1024}]}}},"tools":{"web":{"search":{"provider":"gemini","enabled":true,"openaiCodex":{}},"fetch":{"enabled":true}}},"skills":{"install":{"nodeManager":"npm"}}}
What is root cause for gibberish output

turbid prismBOT Apr 26, 2026, 10:11 AM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

dreamy quail Apr 26, 2026, 10:16 AM

#

@spark wyvern Answer to above query?

dreamy quail Apr 27, 2026, 5:17 AM

#

@spark wyvern

spark wyvernBOT Apr 27, 2026, 5:18 AM

#

Likely root cause: OpenClaw is not pointing at the vLLM endpoint you described.

You said vLLM is served at:

192.168.75.81:18789

but your config points OpenClaw to:

"baseUrl": "http://192.168.75.81:8000/v1"

So OpenClaw is either hitting the wrong server/port, or not hitting vLLM at all.

Also for LAN/private IP vLLM, add:

#

request: { allowPrivateNetwork: true }

Suggested provider block:

models: {
  providers: {
    vllm: {
      baseUrl: "http://192.168.75.81:18789/v1",
      api: "openai-completions",
      apiKey: "${VLLM_API_KEY}",
      request: { allowPrivateNetwork: true },
      timeoutSeconds: 300,
      models: [

#

        {
          id: "meta-llama/Llama-3.3-70B-Instruct",
          name: "meta-llama/Llama-3.3-70B-Instruct",
          reasoning: false,
          input: ["text"],
          contextWindow: 16384,
          maxTokens: 1024
        }
      ]
    }
  }
}

Then from the Windows machine, test directly:

#Gibberish output