Responses

Request contract

Use string input or bounded typed input items, optional instructions and max_output_tokens. This subset is stateless: omit storage or set store:false. Carry previous text/function items forward explicitly.

Tool correlation

Assistant function_call items may be sent back with function_call_output using the matching call_id. Preserve every call and its result. Missing, duplicate or orphaned results are rejected.

Streaming lifecycle

Events include ordered sequence numbers and native response/output events. Successful output ends in response.completed. Output-bound exhaustion ends in response.incomplete with max_output_tokens; the stream does not use [DONE].

Storage boundary

store:true, previous_response_id, conversation IDs, background execution and native retrieve/delete/cancel semantics are unsupported. store:false disables public native response retrieval; it does not erase required durable safety and accounting records.

Request example

Use your customer key and a published model. Copying an example does not send a request.

curl https://api.notholy.com/v1/responses \
  -H "Authorization: Bearer $NOTHOLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "notholy-pulse",
  "input": "Explain least privilege.",
  "max_output_tokens": 256,
  "store": false
}'

Request fields

Generated from the versioned contract. Refinement rules above still apply.

Response schema