Chat Completions - Mibyan Docs

This endpoint follows the OpenAI Chat Completions shape. Streaming returns data-framed completion chunks followed by data: [DONE].

curl https://api.mibyanai.com/v1/chat/completions \
  -H "Authorization: Bearer $MIBYAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"mibyan-4.1","messages":[{"role":"user","content":"Hello"}]}'

curl https://api.mibyanai.com/v1/chat/completions \
  -H "Authorization: Bearer $MIBYAN_API_KEY" \
  -H "Content-Type: application/json" \
  -N \
  -d '{"model":"mibyan-4.1","messages":[{"role":"user","content":"Hello"}],"stream":true}'

The request must contain at least one message. The gateway validates message size, generation limits, sampling values, tools, and structured-output settings before sending the request.

Request fields

Field	Type	Notes
`model`	string, required	Use an ID from List models, currently `mibyan-4.1`
`messages`	array, required	At least one message. Roles: `system`, `developer`, `user`, `assistant`, `tool`. Content can be a string or an array of content parts
`stream`	boolean	Default `false`. When `true`, returns server-sent events
`temperature`	number	From 0 to 2
`max_tokens`	integer	At least 1
`tools`	array	Function tools only, at most 64. Each has a `name` (up to 64 characters), an optional `description` (up to 4,000 characters), and JSON Schema `parameters`
`tool_choice`	string or object	`auto`, `none`, `required`, or an object naming a function
`parallel_tool_calls`	boolean	Allow the model to call several tools in one turn
`response_format`	object	Structured-output settings

Response

A non-streaming call returns a choices array, where each choice has an index, a message, and a finish_reason, plus a usage object with prompt_tokens, completion_tokens, and total_tokens.

Function calling

Describe your functions in tools, let the model decide with tool_choice: "auto", run the function yourself, and send the result back in a tool message.

Behavior to know

Cancelling the HTTP request cancels the upstream generation.
This endpoint does not create asynchronous jobs, so callbacks and webhooks are not accepted.
Possible errors are 400, 401, 403, 404, 413, 429, 502, 503, and 504. See errors and retries.

For field-by-field behavior, see compatibility and streaming.