OpenAI-Compatible API
One of SDKMAX's core design goals is that any code already written against the OpenAI API should work by changing only base_url and api_key — no business-logic rewrite required.
- base_url = "https://api.openai.com/v1"
+ base_url = "https://api.sdkmax.com/v1"
- api_key = "sk-openai-xxx"
+ api_key = "sk-sdkmax-xxx"Even when model points at a non-OpenAI model (Claude, Gemini, DeepSeek, ...), the request/response shape still follows the OpenAI protocol — the gateway handles that translation internally.
Supported compatible endpoints
| Endpoint | Method | Notes |
|---|---|---|
/v1/chat/completions | POST | Chat completions, streaming (stream: true) and function/tool calling (model-dependent) |
/v1/completions | POST | Legacy text completions |
/v1/embeddings | POST | Text embeddings |
/v1/images/generations | POST | Text-to-image, see Images API |
/v1/images/edits | POST | Image editing/inpainting, see Images API |
/v1/audio/transcriptions | POST | Speech-to-text (Whisper-compatible) |
/v1/audio/translations | POST | Speech translation |
/v1/audio/speech | POST | Text-to-speech |
/v1/moderations | POST | Content moderation |
/v1/rerank | POST | Reranking (model-dependent) |
/v1/models | GET | List currently available models |
/v1/models/{model} | GET | Get details for a specific model |
Other native protocols
Beyond the OpenAI protocol, the gateway also accepts /v1/messages (Anthropic Claude's native protocol) via an x-api-key header, and Gemini-style native calls (/v1beta/...) via x-goog-api-key / ?key=. If your code is written against the official Anthropic or Google SDKs, you can point it at SDKMAX by only changing the endpoint — no need to rewrite to the OpenAI shape.
Not yet supported / limited
/v1/images/variationsis not supported yet./v1/filesand/v1/fine-tunesare currently placeholders/incomplete — don't rely on them.- There is no dedicated Batch endpoint yet — see Batch API for an interim approach.
If your use case depends on any of the above, check FAQ or confirm the roadmap with the platform team before shipping a dependency on it.
Using the official OpenAI SDK directly
Because the protocol is compatible, you don't need a dedicated SDKMAX SDK — use each language's official OpenAI SDK with base_url set (see Quickstart). The Python / Node.js / Java / Go SDK pages here mainly show SDKMAX-specific usage (model switching, error handling conventions) — the underlying client is still the standard OpenAI one.
Video and other OpenAI-protocol extensions
Video generation (/v1/video/generations, /v1/videos) is a SDKMAX extension built on top of the OpenAI protocol (aligned with OpenAI's own Video/Sora-style API shape) — see Video API.