documentation

chat completions

Generate text with the familiar OpenAI chat-completions shape.
endpoint

POST /v1/chat/completions

Send an ordered messages array and a Brok language model. The response uses the OpenAI chat-completions shape.

curl https://api.brok.fyi/v1/chat/completions \
  -H "Authorization: Bearer $BROK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"brok-2.3","messages":[{"role":"user","content":"hello"}]}'
request fields

common parameters

  • modelRequired. A language model or stable alias.
  • messagesRequired. System, user, assistant, and tool messages in conversation order.
  • streamOptional boolean. Returns SSE chunks when true.
  • temperatureOptional sampling control.
  • max_tokensOptional maximum generated tokens.
  • toolsOptional OpenAI-compatible function tools.
  • tool_choiceOptional tool selection policy.
response

read the output and usage

const text = data.choices[0].message.content;
const inputTokens = data.usage.prompt_tokens;
const outputTokens = data.usage.completion_tokens;

Save x-brok-request-id. It connects application logs to the exact row in Brok Usage.