Feature Request: Add `serve` mode compatible with OpenAI Embeddings API

[isaac47]

Description:

Feature Request: Serve embed-anything with OpenAI-compatible API

Description:
It would be great to have a serve mode that allows embed-anything to expose an API endpoint compatible with the OpenAI embeddings format.

Why?
This would allow seamless integration with existing applications that expect OpenAI's API format, making embed-anything a drop-in replacement for local embedding models.

Expected API Format:

POST /v1/embeddings
{
"input": ["text to embed"],
"model": "your_model_name"
}

Response example:
```json
{
    "object": "list",
    "data": [
        {
            "object": "embedding",
            "index": 0,
            "embedding": [0.1, 0.2, 0.3, ...]
        }
    ],
    "model": "your_model_name",
    "usage": {
        "prompt_tokens": 10,
        "total_tokens": 10
    }
}

Would this be feasible?
Looking forward to your thoughts!

[akshayballal95]

That's a great idea. This can be done for batch query embedding but may not be that useful for embedding files directly. But we can make a separate endpoint where you can send the pdf files and get the embeddings too.

[lij55]

some model-as-a-service providers have various LLM modes but lack of embedding models. this features will help help to create a always-available local embedding service.