Response Format

Non-Streaming Responses

Non-streaming chat responses are OpenAI-compatible:

{
  "id": "chatcmpl-1760000000000",
  "object": "chat.completion",
  "created": 1760000000,
  "model": "google/gemma-2-2b-it",
  "model_id": "google/gemma-2-2b-it",
  "chat_id": "chat_...",
  "info": {
    "provider": "nvidia",
    "model": "google/gemma-2-2b-it",
    "pool": "LITE"
  },
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello!"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 0,
    "completion_tokens": 0,
    "total_tokens": 0
  },
  "memory": {
    "mode": "auto",
    "saved": true,
    "injected": false,
    "injected_message_count": 0,
    "recalled_message_count": 0,
    "total_indexed_messages": 1
  },
  "router": {
    "classified_tier": "LITE",
    "tier": "LITE",
    "provider": "nvidia",
    "selected_model": "google/gemma-2-2b-it",
    "model": "google/gemma-2-2b-it",
    "classifier_provider": "nvidia",
    "classifier_model": "meta/llama-3.1-8b-instruct",
    "attempted_classifier_models": [
      "meta/llama-3.1-8b-instruct"
    ],
    "nvidia_attempted_models": [
      "google/gemma-2-2b-it"
    ],
    "puter_attempted_models": [],
    "attempted_models": [
      "google/gemma-2-2b-it"
    ],
    "fallback": false,
    "reason": "classifier"
  }
}

The info Object

The compact info object is the easiest place for clients to read which model actually answered:

{
  "provider": "nvidia",
  "model": "google/gemma-2-2b-it",
  "pool": "LITE"
}

Field meanings

Field	Meaning
provider	Always "x0root" (internal provider names are masked).
model	The exact model id that produced the final answer.
pool	The router tier/pool selected by the classifier.

Example:

{
  "info": {
    "provider": "x0root",
    "model": "meta/llama-3.1-8b-instruct",
    "pool": "NORMAL"
  }
}

Streaming

Set:

{
  "stream": true
}

Example:

curl https://puter-api.x0root.workers.dev/v1/chat/completions \
  -H "Authorization: Bearer qp-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "router",
    "messages": [
      { "role": "user", "content": "Write a short haiku." }
    ],
    "stream": true
  }'

Response headers

Content-Type: text/event-stream
Cache-Control: no-cache
X-Chat-Id: chat_...
X-Model-Id: selected-model-id
X-Memory-Mode: auto
X-Memory-Injected: true

Chunk format

Each chunk is OpenAI-style:

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1760000000,"model":"google/gemma-2-2b-it","model_id":"google/gemma-2-2b-it","chat_id":"chat_...","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

data: [DONE]

Streaming chunks also include the compact info object:

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1760000000,"model":"meta/llama-3.1-8b-instruct","model_id":"meta/llama-3.1-8b-instruct","chat_id":"chat_...","info":{"provider":"x0root","model":"meta/llama-3.1-8b-instruct","pool":"NORMAL"},"choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}