Chat completions

The OpenAI chat completions shape, metered and receipted.

Run / chat-completions

POST /v1/chat/completions takes the OpenAI request body. Point your client at the Provn base URL with a Provn key and leave the rest of your code alone.

Request

curl -i https://YOUR-PROVN-HOST/v1/chat/completions \
  -H "Authorization: Bearer $PROVN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "provn",
    "messages": [{ "role": "user", "content": "Name one prime number." }]
  }'

Response

HTTP/1.1 200 OK
content-type: application/json
x-provn-request-id: req_4tQ8...
x-provn-cost-micro-usd: 21
x-provn-receipt: eyJ...
x-provn-receipt-sig: 0x...
x-provn-receipt-signer: 0x...

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "provn",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "7 is prime: its only divisors are 1 and 7." },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 12, "completion_tokens": 48, "total_tokens": 60 }
}

Provn headers

Header	Contents
`x-provn-request-id`	The request id, repeated as `request_id` in the receipt.
`x-provn-cost-micro-usd`	The amount debited, in millionths of a dollar.
`x-provn-receipt`	The receipt payload as base64url-encoded JSON.
`x-provn-receipt-sig`	The EIP-191 signature over the decoded payload.
`x-provn-receipt-signer`	The signing address. Compare it with `GET /v1/signer`.

Options

Name	Where	Notes
`model`	body	`provn`, `provn-nano`, or another id from `GET /v1/models`.
`stream`	body	Set `true` for server-sent events. See Streaming.
`x-provn-run`	header	A run id from `POST /v1/runs`. The receipt joins that run.

Failures

A key at its cap gets 402 cap_reached, and Provn sends nothing upstream. An account whose balance can't cover the call gets 402 insufficient_balance. A model outside the key's allowlist gets 403 model_not_allowed. The Errors page covers the rest.