One unified LLM API for a growing model catalog

What Basktre normalizes

• One API key and wallet across supported model providers.
• A messages-based request shape for text model calls.
• Streaming and non-streaming endpoints.
• Manual model tags or automatic routing.
• Input and output token pricing per model.

Switch models in one field

{
  "model": "claude-haiku-4-5",
  "messages": [
    {
      "role": "user",
      "content": "Return a concise release-note summary."
    }
  ],
  "max_tokens": 300,
  "temperature": 0.2
}

Replace the model tag with another supported tag, or use auto for cost-aware selection.

Streaming request

curl https://api.basktre.in/api/v1/models/stream \
  -H "api-key: your_basktre_key" \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"Hello"}]}'

Compatibility boundaries

Basktre uses its own api-key header and model call endpoints, so it should not be described as a drop-in OpenAI base-URL replacement. The public documentation currently covers text messages, token limits, temperature, top-p, stop sequences and streaming. It does not currently document normalized tools, JSON mode, image generation, audio input, speech output or voice APIs.

Migration checklist

1. Map authentication to the Basktre api-key header.
2. Use the documented call or stream endpoint.
3. Map your current model name to a supported Basktre tag.
4. Test provider-specific parameters and error handling rather than assuming they are normalized.
5. Keep image and voice workloads on a service that explicitly supports them.

See the exact request fields and examples in the API documentation.