Multimodal & Embedding Interfaces
Beyond chat, ZBStream exposes image generation, audio transcription, text-to-speech, embedding, and rerank routes. Results depend on the selected channel having a matching model configured.
Available Routes
| Method | Path | Purpose |
|---|---|---|
| POST | /v1/images/generations | Image generation |
| POST | /v1/audio/transcriptions | Speech-to-text |
| POST | /v1/audio/speech | Text-to-speech |
| POST | /v1/embeddings | Text embeddings |
| POST | /v1/rerank | Result reranking |
Request Requirements
- Build request bodies, file uploads, and response parsing per each endpoint's protocol.
- The model passed must be accessible to the API key and configured on a channel.
- File size, audio formats, embedding dimensions, and image parameters follow the actual limits of the model and channel.
Troubleshooting
First confirm the model via GET /v1/models, then check the path, Content-Type, request body, and key scope. If the failure persists, investigate with the returned HTTP status code and the request records in the usage console.