Skip to main content
GENGEN provides OpenAI-compatible Chat Completions and GPT Responses endpoints for supported LLM models. Existing server-side applications can use the OpenAI Python or JavaScript SDK by changing the API key, base URL, and model ID.

Compatibility base URL

For Chat Completions the SDK appends /chat/completions, producing this request URL:
Use a GENGEN workspace API key beginning with gengen_live_.

Choose an endpoint

Check the model catalog and current prices before sending a billable request. Create a workspace API key after signing in. Endpoint and tool support vary by model; this base URL does not imply support for every OpenAI SDK operation.
GPT models support client.responses.create through /v1/responses. See GPT Responses and tools for exact model IDs, tool support, full conversation replay, streaming, and availability boundaries. Native Files, containers, vector stores and stored/background Responses are available through Hosted tools and resources. Assistants, Batch and shared upstream resources remain outside this interface.

Python

Install the OpenAI Python SDK:
Configure the client with your GENGEN API key and compatibility base URL:

JavaScript

Install the OpenAI JavaScript SDK:
Configure baseURL when creating the client:

Direct HTTP endpoint

Applications that do not use an OpenAI SDK can call the canonical GENGEN endpoint directly:
Both URLs use the same GENGEN API key, model IDs, billing, request body, and response format. Streaming Chat Completions are also supported through the compatibility URL. See Create a chat completion for supported parameters, models, streaming behavior, and response schemas.

Function calling

All listed Seed, DeepSeek, GLM, and Gemini LLMs accept function tools through Chat Completions, including SSE streaming. Preserve the full assistant message when returning tool results, including provider reasoning and Gemini thought signatures. See Function calling for the common execution loop. See Gemini tools for Code execution, Computer use, and Gemini conversation replay requirements.