Dynamic model selection with Ollama when asking

[Maniga]

Hello everyone,

I'm currently working on a project where I'd like to use multiple models in Ollama for Kernel Memory. Specifically, I want to be able to select which Ollama model is used when querying the Kernel Memory with ask. I added the kernel manager as a .net core web api into my dotnet aspire solution.

Is there a standard or recommended way to implement this type of functionality? Would it be necessary to write a custom MemoryService to enable model selection, or is there an existing feature or pattern that supports this?

Any guidance or examples from those who have faced a similar requirement would be greatly appreciated!

Thank you in advance!

[dluc]

KM supports "Request Context" parameters, which can be used to change configurations during a request, without affecting the deployment or other concurrent requests. To implement what you suggest, you could add a new Request Context param name, and support it in the Ollama generator.

Here's some pointers:

• https://github.com/microsoft/kernel-memory/blob/main/service/Abstractions/Context/RequestContext.cs
• https://github.com/microsoft/kernel-memory/blob/main/service/Abstractions/Constants.cs#L28
• https://github.com/microsoft/kernel-memory/blob/main/service/Abstractions/Context/IContext.cs#L91
• https://github.com/microsoft/kernel-memory/blob/main/examples/209-dotnet-using-context-overrides/Program.cs

To access params stored in the request context, you can inject IContextProvider into OllamaTextGenerator.