Migrate
Point your OpenAI SDK at idapt and keep your code: chat, Responses, embeddings, images, and audio, all through one API.
- https://api.openai.com/v1
+ https://idapt.app/api/openai/v1Authenticate with an idapt API key (Authorization: Bearer uk_…). Create one in Settings → the Gateway tab.
| OpenAI API | In idapt | How |
|---|---|---|
| POST /chat/completions (stream + tools) | Identical endpoint, 200+ models | Automatic |
| POST /responses | Non-streaming Responses bridge (stateless text)stream:true is an explicit 501. Use /chat/completions for streaming. | Automatic |
| Embeddings / images / audio (TTS, STT) | Same-shape endpoints | Automatic |
| Model names | idapt model ids (see /models); gpt-… slugs resolve where unambiguous | Manual |
| auto model / n > 1 / assistants state | Not supported on the compat surfaceUnsupported fields are rejected with a clear error, never silently ignored. | No equivalent |
Settings → Gateway → API keys. Keys are scoped (completions read/write) and can carry a lifetime USD spend cap.
Set your SDK's baseURL to https://idapt.app/api/openai/v1 and the API key to your uk_ key. Everything else stays.
Browse /models for 200+ ids and switch per request. One key reaches all of them.
Private AI runs open models (Llama, Qwen, Gemma, Mistral, Phi and more) on hardware you own, through the idapt daemon: free, never rate-limited, and no third-party AI provider ever sees the prompt. They sit beside the 200+ cloud models in the same chat.
LLM tokens bill against your credits or plan allowance at the prices shown in the app. Platform revenue comes from subscriptions and a 12.5% service fee on credit top-ups.
Yes. /chat/completions streams SSE exactly like OpenAI's, tools included. Only the Responses endpoint is non-streaming for now.