Genesis Model API¶
Genesis gives applications a single entry point for model inference. Chat, embedding, and reranking endpoints follow OpenAI-compatible conventions. Responses and Messages preserve the request and event formats of their respective ecosystems. To connect, use a Genesis API Key and point the client at the Base URL shown in the Genesis console.
Before you call the API¶
Collect these three values:
Base URL: Copy it from the Use page in the Genesis console. It usually includes the API version path.
API Key: Obtain one from Genesis key management or a service account in the unified account system.
Model ID: Copy one from Model Marketplace, or call
GET /modelsto discover the models available to the current key.
Read Endpoints and Authentication, then choose an endpoint:
Task |
Endpoint |
Documentation |
|---|---|---|
Text or image conversation |
|
|
OpenAI Responses-style call |
|
|
Anthropic Messages-style call |
|
|
Text or multimodal vectorization |
|
|
Candidate-document relevance ranking |
|
|
Dynamic model discovery |
|
The code generator on the Use page covers these endpoints. OCR, image, video, and audio capabilities do not all share a standard protocol. Confirm the call path and parameters in the model details and Debug Playground before using them, then consult the Model API parameter reference.
Compatibility boundaries¶
Compatibility means that a common SDK can connect by changing its base_url and api_key; it does not mean every model accepts every field. Model capabilities, context limits, streaming events, optional parameters, and defaults depend on the selected model and the deployed service. Validate production requests with the target model in the Genesis Debug Playground, and make clients tolerant of unknown response fields.
For retry guidance, see Error Codes. For key issuance, limits, and request logs, see the Genesis guide.