Support multimodal embedding generation

[dineshyv]

🚀 Describe the new functionality needed

Needed for supporting RAG on documents with images and searches using image

💡 Why is this needed? What if we don't build it?

RAG will only support text, which is not ideal.

Other thoughts

No response

[mattf]

@dineshyv /v1/inference/embeddings supports string, typed text, and typed image (url and base64) input. is that sufficient to resolve this issue?

https://github.com/meta-llama/llama-stack/blob/main/llama_stack/apis/inference/inference.py#L433 -> https://github.com/meta-llama/llama-stack/blob/main/llama_stack/apis/common/content_types.py#L75 -> https://github.com/meta-llama/llama-stack/blob/main/llama_stack/apis/common/content_types.py#L66 -> https://github.com/meta-llama/llama-stack/blob/main/llama_stack/apis/common/content_types.py#L42

[ashwinb]

Yes this is now possible with the pointers @mattf provided 🎉