AgentGroupChat with Vector database support #10429
-
I want to use the rag pattern with the AgentGroupChat but I don't see any way to hook into my ChatCompletionAgent. I tried to find a way to use Semantic Memory with AgentGroupChat but there does not seem to be a way. Also, using Azure OpenAI File Search or the new VectorDB support would be nice. I have these working with OpenAIAssistentAgent but not with a ChatCompletionAgent. For demonstrative purposes, I want the Program Manager to search a vector database for technology standards used in the organization. I did use a plugin function to write the code the Software Engineer agent created. I am thinking a way to accomplish this is to create a plugin that calls a vector database and returns information (.i.e. Semantic Memory, etc.) Here is the code below: using System.ComponentModel; namespace SemanticKernelCoderAgentConsole
#pragma warning disable SKEXP0110, SKEXP0001 // Rethrow to preserve stack details
} |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
Adding @crickman to help answer. |
Beta Was this translation helpful? Give feedback.
-
Yes, using a plug-in is the way you'd want to integrate RAG/VectorDB inetgrateion with a We have a demo that shows how to do this integration: On line 94, we are registering a vector store text search instance as a plugin: LIne 121 is where it is invoked: For this to work register both a vector store and a vector store text search instance: |
Beta Was this translation helpful? Give feedback.
Yes, using a plug-in is the way you'd want to integrate RAG/VectorDB inetgrateion with a
ChatCompletionAgent
.We have a demo that shows how to do this integration:
On line 94, we are registering a vector store text search instance as a plugin:
https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Demos/VectorStoreRAG/RAGChatService.cs#L94
LIne 121 is where it is invoked:
https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Demos/VectorStoreRAG/RAGChatService.cs#L121
For this to work register both a vector store and a vector store text search instance:
https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Demos/VectorStoreRAG/Program.cs#L69
h…