.Net: Feature: Support Vector Embedding generation in AWS Bedrock

[arielcostas]

Support models such as amazon.titan-embed-text-v2:0 and cohere.embed-multilingual-v3 to generate text embeddings via AWS Bedrock.

[arielcostas]

To add some info, since I've been trying to implement it myself (see my fork), and find that the two providers (Amazon and Cohere) have different requests/responses, in that Amazon's models support only embedding a single text string at a time, while cohere supports "batch" requests of multiple texts in one call. Would there be a problem in supporting Amazon's models by sending one request at a time and then returning the responses joined?

[markwallace-microsoft]

@RogerBarreto Have we run into this issue in any other connectors?

[RogerBarreto]

@arielcostas @markwallace-microsoft I have seeing this sometime ago when I was implementing HuggingFace embeddings and the multiple was not supported.

Multiple requests might be a valid approach in the lack of the multiple text support, I would add a XmlDox <remark> clarifying this behavior and attemptiveness to the rate limiting, as a big number of simultaneous embedding requests can be throttled down.