# Models

Route LLM and multimodal requests through SeaRouter with provider credentials hidden server-side.

## Responses API

`POST /api/searouter/llm/v1/responses` is the OpenAI-compatible route for text and reasoning workflows.

Use the standard SeaChat base URL and an API key with `model:invoke` or `searouter:invoke`.

```bash
curl https://seachat.ai/api/searouter/llm/v1/responses \
  -H "Authorization: Bearer $SEACHAT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.5","input":"Summarize the SeaChat API in one paragraph."}'
```

## Multimodal invoke

`POST /api/searouter/v1/invoke/{modality}/{model}/{version}` supports image, video, audio, embedding, rerank, and other modality-specific providers.

Generated media is imported into managed SeaDB artifacts when the provider returns bytes or URLs. Render the returned `outputs[].url` directly and keep `artifactRefId` for later references.

```bash
curl https://seachat.ai/api/searouter/v1/invoke/image/bytedance_seedream_4_0/latest \
  -H "Authorization: Bearer $SEACHAT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"A clean product diagram for an agent API gateway"}'
```

## Discovery

Use `/api/searouter/v1/help`, `/api/searouter/v1/help/modalities`, and `/api/searouter/v1/help/models/search` before selecting models or provider-specific fields.