Create a chat completion

OpenAI-compatible Chat Completions endpoint. Supports streaming via stream: true, which returns a Server-Sent Events stream of chat.completion.chunk objects.

POST /chat/completions

Create a chat completion

cURL

curl --request POST \
  --url https://api.sailresearch.com/v1/chat/completions \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "zai-org/GLM-5.2-FP8",
  "messages": [
    {
      "role": "system",
      "content": "You are a concise assistant."
    },
    {
      "role": "user",
      "content": "Summarize retrieval-augmented generation in 3 bullets."
    }
  ],
  "max_completion_tokens": 300
}
'

200

{
  "id": "<string>",
  "object": "chat.completion",
  "created": 123,
  "model": "<string>",
  "choices": [
    {
      "index": 123,
      "message": {
        "role": "assistant",
        "content": "<string>",
        "reasoning_content": "<string>",
        "refusal": "<string>"
      },
      "logprobs": null
    }
  ],
  "usage": {
    "prompt_tokens": 123,
    "completion_tokens": 123,
    "total_tokens": 123,
    "prompt_tokens_details": {},
    "completion_tokens_details": {}
  }
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Headers

Idempotency-Key · string

Makes the request retry-safe. Sail stores a reservation keyed by (organization, API key, Idempotency-Key); retrying with the same value returns the previously stored response instead of re-running inference. Keys are capped at 255 characters. See Idempotent Requests for full semantics.

Maximum string length: 255

Body

application/json

model · string · required

messages · object[] · required

Minimum array length: 1

Show child attributes

temperature · number | null

Required range: 0 <= x <= 2

top_p · number | null

Required range: 0 <= x <= 1

max_completion_tokens · integer | null

Required range: x >= 1

response_format · object

Option 1

Option 2

Show child attributes

reasoning_effort · enum<string> | null

Available options: none, minimal, low, medium, high, xhigh

n · enum<integer>

Only n=1 is currently supported.

Available options: 1

modalities · enum<string>[]

Required array length: 1 element

Available options: text

stream · boolean

When true, the response is returned as a Server-Sent Events stream of chat.completion.chunk objects instead of a single JSON response.

stream_options · object

Options that apply when stream is true.

Show child attributes

store · enum<boolean>

Only true is supported.

Available options: true

user · string

Maximum string length: 256

prompt_cache_key · string

Optional routing hint for prompt-prefix cache locality. Requests with the same key are preferentially routed to maximize cache hit rates.

metadata · object

Optional string metadata. completion_window controls scheduling; completion_webhook/webhook_token configure completion webhooks.

Show child attributes

Response

200

application/json

Chat completion. Returns a single JSON object by default, or a Server-Sent Events stream of chat.completion.chunk objects when stream: true (terminated by a final data: [DONE] line).

id · string · required

object · enum<string> · required

Available options: chat.completion

created · integer · required

model · string · required

choices · object[] · required

Minimum array length: 1

Show child attributes

usage · object

Show child attributes