Support /v1/embeddings endpoint for embedding models

[lemoncat7]

Feature Request

Currently the CLI Proxy API only exposes:

• POST /v1/chat/completions
• POST /v1/completions
• GET /v1/models

The /v1/embeddings endpoint is missing (returns 404 page not found).

Use Case

When using OpenClaw memory-search with provider: "custom-cli-proxy-api-8317", embedding requests fail because the proxy does not support the /v1/embeddings endpoint that OpenClaw expects.

Expected Behavior

Support POST /v1/embeddings with standard OpenAI-compatible request format:

{
  "model": "text-embedding-3-small",
  "input": "hello world"
}

Additional Request

Also, please consider adding kimi-2.6 model support if it is available from the backend providers.

[Sheldontao]

same for gemini:
/v1beta/models/embedding-2:embedContent
not /v1beta/openai/embeddings(current)

[YiHui-Liu]

希望官方能增加对embedding的支持