Repository navigation
.Net: Add Bedrock Connector - #7587
Roger Barreto (rogerbarreto) merged 154 commits into
Conversation
…sample program and run
… supported by model
…models functional for all actions supported
…ce fixes in connector
Co-authored-by: Roger Barreto <[email protected]>
…e and IBedrockTextGenerationIOService
|
Apologies, I think there were extra |
|
I still see some Search for should help:
Overall LGTM |
|
The usings I believe you will only be able to solve, removing the |
There was a problem hiding this comment.
Charlize Yeh (@charlizeyeh) Thank you soo much for all changes made. Tomorrow plan to merge this in the feature branch.
95b82dd
into
microsoft:feature-connectors-amazon
|
Roger Barreto (@RogerBarret0) |
|
Charlize Yeh (@charlizeyeh), yes, all connectors feature branches as of now intend to merge in main at some point. One feature that we want to add as part of the connectors is |
|
Charlize Yeh (@charlizeyeh) thanks for your connector! I am looking forward to using it. Does it already support function calling (with Claude)? |
Roger Barreto (@RogerBarret0) Could you please merge this PR into main which has the implementation of |
|
Roger Barreto (@rogerbarreto) Any update on the question above? We'd love to get this out. |
|
Beau Gosse (@Beau-Gosse-dev) yes, we have plan on moving this one forward next sprint 35 (Sep-23 / Oct-06). |
|
Do you have models for TextEmbeddings? Would be nice to have some options for the |
|
Roger Barreto (@rogerbarreto) Thanks for the update. There is a first draft of TextEmbeddings support in our fork, but we've been pulled into other priorities at the moment and wouldn't be able to get it fully ready before your next sprint, so hopefully the first release can go without it and we can support TextEmbeddings in a later release. |
Motivation and Context
There has been a strong demand from developers in the community to be able to utilize Bedrock with the Semantic Kernel. With the wide variety of foundation models offered including specialized models tailored for various domains and use cases, Bedrock's AI capabilities can now be leveraged within the Semantic Kernel's unified ecosystem with this connector.
Description
This pull request implements a connector for Amazon Bedrock, allowing developers to leverage Bedrock's foundation models within the Semantic Kernel framework. Key features of this implementation include:
IChatCompletionServiceandITextGenerationServicefrom the Semantic Kernel.The implementation leverages the following Amazon Bedrock Runtime actions:
InvokeModelAsyncandInvokeModelWithResponseStreamAsyncforGetTextContentsAsyncandGetStreamingTextContentsAsync, respectively.ConverseAsyncandConverseStreamAsyncforGetChatMessageContentsAsyncandGetStreamingChatMessageContentsAsync, respectively.The following base models from various providers are currently connected and available for use under Bedrock:
The kernel builder extension methods support two authentication scenarios:
IAmazonBedrockRuntimeobject: Developers can provide a pre-configuredIAmazonBedrockRuntimeobject, allowing for customized configurations such as specifying custom credentials, endpoints, or other settings.IAmazonBedrockRuntimecreation: If developers do not provide anIAmazonBedrockRuntimeobject, an instance is automatically created and registered in the dependency injection (DI) container using theAWSSDK.Extensions.NETCore.Setuppackage. In this case, theIAmazonBedrockRuntimeobject will be configured with default settings, such as credentials obtained from environment variables, AWS credentials files, or instance profile credentials.Resolve #3094
Resolve #4335
Contribution Checklist