agent-redact/pi/agent/models.json

{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://llama-inference:8080/v1",
      "api": "openai-completions",
      "apiKey": "llama-cpp",
      "compat": {
        "supportsDeveloperRole": false,
        "supportsReasoningEffort": false,
        "supportsUsageInStreaming": false,
        "maxTokensField": "max_tokens"
      },
      "models": [
        {
          "id": "unsloth/Qwen3.6-27B-MTP-GGUF",
          "name": "Qwen 3.6 27B (local)",
          "reasoning": false,
          "input": ["text", "image"],
          "contextWindow": 114688,
          "maxTokens": 32768,
          "cost": {
            "input": 0,
            "output": 0,
            "cacheRead": 0,
            "cacheWrite": 0
          }
        }
      ]
    }
  }
}