EU GPT logo
EU GPT

Public preview — This API is in public preview. Endpoints, schemas, and limits may change before general availability.

API

API reference

Interactive OpenAPI 3.1 reference for /v1/responses.

This is the live OpenAPI 3.1 reference for the EU GPT public API. Use the panels below to explore endpoints, see schemas, and try requests against your environment.

For a top-to-bottom prose walkthrough of the same surface, see Create a response.

EU GPT — Public API

v1.0.0
Download openapi.yaml

External-facing API for programmatic access to EU GPT.

Authenticate every request with an API key issued from the EU GPT web app:

Authorization: Bearer eugpt_<token>

Keys are managed in the chat UI (Settings → API Keys). They have no rotating refresh — revoke and reissue when needed.

Base URL

https://chat.eugpt.ai Production

Authentication

apiKey http · bearer

Bearer format eugpt_<token>

EU GPT API key. Issue from the web app; pass as Authorization: Bearer eugpt_<token>.

Endpoints

POST /v1/responses

Create a response

Generate an AI response from a user message. Two modes:

  • Streaming (stream: true, default) — returns Server-Sent Events with OpenAI-compatible event payloads. Each event is delivered as event: message\ndata: <json>\n\n. Stream completion is signalled by a response.completed event.
  • Non-streaming (stream: false) — collects all output and returns a single JSON document.

Conversations are optional. Omit conversation_id for stateless requests; pass an existing UUID to append to a thread that belongs to the authenticated user. Tools (web search, web fetch, calculator, etc.) are auto-approved for API requests — there is no interactive approval gate.

Request body Required

  • modelstring
    Accepted for OpenAI-SDK compatibility but ignored — the router always selects the model. Pass "auto" (the default and only meaningful value); any concrete id you send has no effect. The actual model used is reported in the response model field.
    DefaultautoExampleauto
  • inputstring | ContentItem[]Required
    Either a plain string (single user message) or a list of structured ContentItem objects for multi-modal / file input.
  • streamboolean
    Stream as SSE (true, default) or return a single JSON response (false).
    Defaulttrue
  • instructionsstringNullable
    System prompt for this response (or conversation, if new).
  • conversation_iduuidNullable
    Append to an existing conversation. Omit for a stateless request — an ephemeral conversation is created and returned in the response.
  • project_iduuidNullable
    Project context for RAG retrieval. Caller must be a member of the project.

Responses

200
Either an SSE stream (stream: true) or a single JSON response (stream: false).
application/json · text/event-stream ▾
  • iduuidRequired
  • object"response"Required
  • created_atintegerRequired
    Unix epoch milliseconds (zero if the chain didn't emit).
  • status"completed"Required
  • modelstringRequired
    Concrete model id used (resolved from "auto" when needed).
  • outputobject[]Required
    Structured output items (text blocks, tool calls, etc.) in the same shape as the SSE response.output_item.added events.
  • output_textstringRequired
    Concatenated text deltas — convenience for plain-text consumers.
400
Input message is required or otherwise invalid.
application/json ▾
  • detailstring | object
  • errorobject
401
Missing, expired, or revoked API key.
application/json ▾
  • detailstring | object
  • errorobject
403
License or organization policy denies the request — for example, user is not a member of the requested project, or the organization's plan disallows the operation.
application/json ▾
  • detailstring | object
  • errorobject
404
Referenced conversation or project does not exist.
application/json ▾
  • detailstring | object
  • errorobject
429
Daily or monthly token usage limit exceeded.
application/json ▾
  • detailstring | object
  • errorobject
500
Internal error during response generation.
application/json ▾
  • detailstring | object
  • errorobject

Schemas

CreateResponseRequest

object
  • modelstring
    Accepted for OpenAI-SDK compatibility but ignored — the router always selects the model. Pass "auto" (the default and only meaningful value); any concrete id you send has no effect. The actual model used is reported in the response model field.
    DefaultautoExampleauto
  • inputstring | ContentItem[]Required
    Either a plain string (single user message) or a list of structured ContentItem objects for multi-modal / file input.
  • streamboolean
    Stream as SSE (true, default) or return a single JSON response (false).
    Defaulttrue
  • instructionsstringNullable
    System prompt for this response (or conversation, if new).
  • conversation_iduuidNullable
    Append to an existing conversation. Omit for a stateless request — an ephemeral conversation is created and returned in the response.
  • project_iduuidNullable
    Project context for RAG retrieval. Caller must be a member of the project.

ContentItem

object
  • rolestringNullable
    Optional role hint (e.g. "user").
  • typestringNullable
  • contentInputText | InputFile[]Required

InputContent

InputText | InputFile

InputText

object
  • type"input_text"Required
  • textstringRequired

InputFile

object
  • type"input_file"Required
  • file_iduuidRequired
    Id of a file previously uploaded via the file upload API.

ResponseObject

object
Returned when stream: false.
  • iduuidRequired
  • object"response"Required
  • created_atintegerRequired
    Unix epoch milliseconds (zero if the chain didn't emit).
  • status"completed"Required
  • modelstringRequired
    Concrete model id used (resolved from "auto" when needed).
  • outputobject[]Required
    Structured output items (text blocks, tool calls, etc.) in the same shape as the SSE response.output_item.added events.
  • output_textstringRequired
    Concatenated text deltas — convenience for plain-text consumers.

SseEventStream

text/event-stream
Server-Sent Events stream. Each event is one of the OpenAI-compatible types below; payloads are JSON in the data: line. Event types: - response.created — initial event with response metadata. - response.output_text.delta — incremental text chunks. - response.output_text.done — text part finalised. - response.output_item.added / response.output_item.done — structured items (tool calls, file references). - response.content_part.added — additional content parts. - response.completed — terminal success event. - error — terminal error event with {type, code, message}.

Error

object
  • detailstring | object
  • errorobject
PreviousTool useNextChangelog