model

[unkempt terrace]

Can we use model with custom API? I see only for fallback

[obtuse perch]

Yes! Clawdbot supports custom API endpoints for models (OpenAI-compatible format). You can set a custom baseUrl in your model configuration.

Example config for a custom API:

{
  "models": {
    "my-custom-model": {
      "provider": "openai",
      "model": "your-model-name",
      "baseUrl": "https://your-custom-api.com/v1"
    }
  }
}

This works with any OpenAI-compatible API (LocalAI, LM Studio, etc.).

Key docs:

• Models: https://docs.clawd.bot/concepts/models
• Provider configs: check the docs for your specific provider

What provider/API are you trying to connect? I can help narrow it down! 🦐

[unkempt terrace]

can you check this provider, im trying to integrate, but i think it's not possible or im wrong?

[obtuse perch]

Hey! I'd be happy to check — but I need to know which provider you're trying to integrate! 😊

What service/API are you attempting to connect?

[unkempt terrace]

here
https://jatevo.ai/private-ai/api

[obtuse perch]

Based on my research, Jatevo.ai is a decentralized GPU LLM inference network, but I couldn't find public API documentation for their endpoint format.

For Clawdbot to work with a custom provider, we need to know:

1. API format - Is it OpenAI-compatible (/v1/chat/completions) or Anthropic-compatible (/v1/messages)?
2. Base URL - What endpoint should baseUrl point to?
3. Auth - API key header name (usually Authorization: Bearer <key>)
4. Model IDs - What model names/IDs do they expose?

Can you share:

• Your full API endpoint URL?
• A sample curl request or any documentation they provided?

If it's OpenAI-compatible (most likely for a GPU inference network), the config would look like:

{
  models: {
    mode: "merge",
    providers: {
      jatevo: {
        baseUrl: "https://your-jatevo-endpoint/v1",
        apiKey: "${JATEVO_API_KEY}",
        api: "openai-completions",  // or "anthropic-messages"
        models: [
          { 
            id: "model-name-they-use", 
            name: "Jatevo Model",
            contextWindow: 32768,
            maxTokens: 4096
          }

        ]
      }
    }
  }
}

Let me know what details you have and I'll help you get it working! 🦐

[unkempt terrace]

Here some sample curl

curl -X POST https://inference.jatevo.id/v1/chat/completions
-H "Content-Type: application/json"
-H "Authorization: Bearer YOUR_API_KEY"
-d '{
"model": "glm-4.7-fp8",
"messages": [
{"role": "user", "content": "Hello, how are you?"}
],
"stream": false,
"max_tokens": 1000
}'

[obtuse perch]

Perfect! That's OpenAI-compatible. Here's your config:

{
  agents: {
    defaults: { model: { primary: "jatevo/glm-4.7-fp8" } }
  },
  models: {
    mode: "merge",
    providers: {
      jatevo: {
        baseUrl: "https://inference.jatevo.id/v1",
        apiKey: "${JATEVO_API_KEY}",
        api: "openai-completions",
        models: [
          { 

            id: "glm-4.7-fp8", 
            name: "GLM-4.7 FP8",
            contextWindow: 131072,
            maxTokens: 8192
          }
        ]
      }
    }
  }
}

Set your API key:

export JATEVO_API_KEY="your-actual-api-key"

Or add it to your env file / system env.

The glm-4.7-fp8 model is a Zhipu AI (GLM) model — Jatevo appears to be a decentralized network serving these models. Should work great with the config above! 🦐

[unkempt terrace]

Can i make that as primary model?

[obtuse perch]

Yes! That's exactly what the config does — agents.defaults.model.primary: "jatevo/glm-4.7-fp8" sets it as your primary model.

Steps:

1. Set the env var:

   export JATEVO_API_KEY="your-key-here"
   

2. Apply the config to ~/.clawdbot/clawdbot.json

3. Hot-reload or restart:

   clawdbot gateway call config.apply
   

Then test with:

clawdbot models test jatevo/glm-4.7-fp8 "Hello!"

Let me know if you hit any issues! 🦐

[unkempt terrace]

Is it will work when i send messages from whatsapp?