Dane | Inferock docs

Before you begin

Dane runs on Inferock Serve and is paid from your prepaid balance. Set up Inferock Serve first; it shows your gateway base URL with your Inferock key.

export GATEWAY_BASE_URL='paste-the-gateway-base-url-shown-at-activation'
export INFEROCK_KEY='your-issued-key'

Your first call

curl -sS "${GATEWAY_BASE_URL%/}/v1/chat/completions" \
  -H "Authorization: Bearer ${INFEROCK_KEY}" \
  -H 'Content-Type: application/json' \
  -H 'x-governance-provider: inferock-serve' \
  -d '{"model":"dane","messages":[{"role":"user","content":"Reply with one short sentence."}]}'

A successful request returns HTTP 200 with one answer in choices[0].message.content.

The same call in the Messages format:

curl -sS "${GATEWAY_BASE_URL%/}/v1/messages" \
  -H "x-api-key: ${INFEROCK_KEY}" \
  -H 'Content-Type: application/json' \
  -H 'x-governance-provider: inferock-serve' \
  -d '{"model":"dane","max_tokens":1024,"messages":[{"role":"user","content":"Reply with one short sentence."}]}'

Request fields

Field	What it does
`model`	`dane`.
`messages`	The conversation, in the Chat Completions format.
`max_tokens`	An integer up to 4,096 is accepted. A lower value lowers the balance a call needs to start.
`reasoning_effort`	`low` (recommended), `medium` or `high`. Low effort is the default when none is named.
`stream`	`true` sends the answer as server-sent events.
`tool_choice`	`"none"` is accepted; any other value returns HTTP `400` with `unsupported_parameter`.
`tools`	Function definitions are not supported for Dane; a request that sends `tools` returns HTTP `400` with `unsupported_parameter`.

Other fields, such as temperature or top_p, return HTTP 400 with unsupported_parameter.

The response

Each call returns one answer.

choices[0].message.content holds the answer. choices[0].finish_reason is stop or length.
usage.prompt_tokens and usage.completion_tokens are the tokens billed at Dane's input and output prices.
With stream: true, the answer arrives as server-sent events and the stream ends with data: [DONE]. If the call fails after the stream starts, the last event carries error instead.

Limits

Limit	Value
Context window	1,000,000 tokens
Maximum output	4,096 tokens
Input	Text
Output	Text
Effort levels	Low (recommended), Medium, High

Hard questions can take several minutes.

Dane calls count toward your account's request, token and spend limits like any other call. See Limits.

Errors

Status	`error.code`	Meaning	What to do
`400`	`unsupported_parameter`	The request sent a field Dane does not support.	Remove the field the error names.
`400`	`invalid_request`	The request is invalid.	Fix the request, then try again.
`402`	`insufficient_balance`	Your prepaid balance cannot cover the start of this request.	Add funds, or lower `max_tokens`, then try again.
`404`	`model_not_available`	Dane is not available for your account.	Make sure Inferock Serve is active; contact support if it is.
`413`	`payload_too_large`	The request is too large.	Shorten the conversation, then try again.
`429`	`rate_limited`	Retry after the limit resets.	Wait for the `Retry-After` delay when one is sent, then try again.
`429`	`quota_exceeded`	Retry after the quota resets.	See Limits.
`429`	`spend_cap_exceeded`	Retry after the spend cap resets.	See Limits.
`502`	`answer_unavailable`	The answer could not be produced.	Try once more; if it repeats, contact support with your request ID.
`503`	`service_unavailable`	Inferock Serve is temporarily unavailable.	Try again later.
`503`	`policy_unavailable`	Tenant limit policy is temporarily unavailable.	Try again later.
`503`	`dane_unavailable`	Dane is temporarily unavailable. Please retry shortly.	Try again shortly.
`504`	`answer_timeout`	The answer was not ready within the time limit.	Try again, or set a lower `reasoning_effort`.

Account limit errors (413 with payload_too_large, 429 with rate_limited, quota_exceeded or spend_cap_exceeded, 503 with service_unavailable or policy_unavailable) work as described in Limits.