Responses

curl https://api.noirdoc.de/v1/responses \
  -H "Authorization: Bearer $NOIRDOC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model-id>",
    "instructions": "Answer in three sentences.",
    "input": "What does Mr Müller write to the firm?"
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.noirdoc.de/v1",
    api_key=os.environ["NOIRDOC_API_KEY"],
)

response = client.responses.create(
    model="<model-id>",
    instructions="Answer in three sentences.",
    input="What does Mr Müller write to the firm?",
)
print(response.output_text)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.noirdoc.de/v1",
  apiKey: process.env.NOIRDOC_API_KEY,
});

const response = await client.responses.create({
  model: "<model-id>",
  instructions: "Answer in three sentences.",
  input: "What does Mr Müller write to the firm?",
});
console.log(response.output_text);

What Noirdoc changes

References to stored objects

The providers managed by Noirdoc use one shared account for all organisations. There, the gateway checks every reference in the body:

• previous_response_id, referenced files (file_id) and containers must have been created through your organisation. Otherwise, the gateway answers with 404 object_not_found.
• The gateway rejects conversation, prompt.id, input[] items of type item_reference, input[] items that carry only an id without content, and tools[].vector_store_ids with 403 reference_not_available_on_platform.

With your own provider key (BYOK), these checks do not apply.

Storage at the provider

By default, the Responses API stores input and response at the provider (store). With the providers managed by Noirdoc, the gateway sets store to false unless admins of your organisation have turned on Allow Responses API storage at provider under Models → Shield. New organisations start with the option off; organisations that existed before it was introduced keep it on. As long as it is off, the gateway rejects "store": true, previous_response_id and the sub-routes with {id} with 403 stored_responses_disabled. In that case, send the previous history in input with every request; reasoning models remain usable without storage through include: ["reasoning.encrypted_content"]. With your own provider key (BYOK), the setting does not apply.

Sub-routes

input_tokens and compact require model and run through the catalogue like POST /v1/responses. On compact, Noirdoc Shield inspects instructions and input just like on POST /v1/responses when Shield is active; POST /v1/responses/input_tokens skips Noirdoc Shield at €0.

The calls with {id} carry no model. The gateway forwards them to your organisation’s oldest active provider in the OpenAI format. Your own provider keys (BYOK) take precedence. With providers managed by Noirdoc, you reach only responses that were created through your organisation, and only if the Allow Responses API storage at provider option is on (see Storage at the provider). For all other IDs, the gateway answers with 404 object_not_found.

Stored responses (GET /v1/responses/{id} and input_items) retain one-way Shield placeholders such as [PERSON] (and under legacy masking until 30.11.2026, placeholders are not translated back on retrieval either).

The gateway answers other paths under /v1/responses/ with 404 unsupported_endpoint.

input_tokens generates no model response. The gateway logs the call, skips Shield and bills €0. There is no cost estimate before forwarding. With providers managed by Noirdoc, the following still applies: if the organisation’s credit or the key’s budget is used up, the gateway answers with 402. The gateway bills compact by tokens (plus Shield when active) like POST /v1/responses.

Errors on this endpoint

OpenAI reference

OpenAI’s API reference describes all other request and response fields: Responses (checked on 30 September 2026).