OpenAI compatibility

The official OpenAI SDKs work with GEEKIDEA by changing only the base URL and the API key:

client = OpenAI(base_url="https://api.geekidea.ai/v1", api_key="sk-gi-...")

const client = new OpenAI({ baseURL: 'https://api.geekidea.ai/v1', apiKey: 'sk-gi-...' })

Supported endpoints

Endpoint	SDK method	Notes
`POST /v1/chat/completions`	`chat.completions.create`	Streaming and non-streaming
`POST /v1/images/generations`	`images.generate`	`url` and `b64_json` response formats
`POST /v1/images/edits`	`images.edit`	Multipart or JSON reference images; `mask` is not supported yet
`/v1/files`	`files.*`	Images only, used as references
`GET /v1/models`	`models.list`	Adds pricing, capabilities and parameters
`POST /v1/videos/generations`	—	GEEKIDEA extension, always asynchronous
`/v1/generations`	—	GEEKIDEA extension: look up results and billing for any call
`/v1/organization/**`	—	Administration API, with an admin key

Request parameters

All OpenAI request parameters for the supported endpoints are accepted. Unknown fields are rejected rather than ignored: the request fails with 400 unsupported_parameter, and param names the field path, for example messages[0].foo. This includes provider pass-through fields such as provider or extra_body. Model-specific options that OpenAI has no field for go in generation_config, for example an image resolution tier or a video’s audio track. Each model lists the options it accepts in generation_config.

What GEEKIDEA adds to responses

Field	Where	Meaning
`usage.cost`, `usage.currency`, `usage.cost_details`	Every response, the last stream chunk, generations and webhooks	What the call cost, in US dollars
`id`	Image responses	The call’s ID (`img_…`), for looking it up later. OpenAI SDKs ignore extra fields
`error.retryable`, `error.errors`	Errors	Whether retrying makes sense, and every validation error
`x-request-id`	Every response header	Include it when you contact support
`x-geekidea-id`	Image and video response headers	The call’s ID

Model IDs use the author/model form, for example openai/gpt-image-2. See Models.

Get started

Generate

Workflows

Async and webhooks

Account

Supported endpoints

Request parameters

What GEEKIDEA adds to responses

​Supported endpoints

​Request parameters

​What GEEKIDEA adds to responses

Supported endpoints

Request parameters

What GEEKIDEA adds to responses