[Bug] Different results across KM instances

[nurkmez2]

Context / Scenario

Scenario;

I have 3 different platforms;

• A kernel memory service running in a local docker
• A Kernel memory service tane azure app container
• A dotnet serverless app running with Azure Search Service

All settings are the same for all 3 platform

SearchAsync finds different results.

What happened?

When SearchAsync is called with minRelevance: 0.6 for all 3 platforms above,

• The service running in local docker finds no results.
• But the service running in Azure app container and dotnet serverless app find 10 results.

**When I decrease minRelevance: 0.01,

• All find 10 results.

What can cause this result?

Importance

a fix would make my life easier

Platform, Language, Versions

Windows, C#, last version of KM

Relevant log output

[dluc]

The service has some defaults that you might not have in the serverless, so I would compare these configurations:

• KernelMemory.DataIngestion.TextPartitioning
• KernelMemory.Retrieval.SearchClient
• KernelMemory.Services.AzureAISearch

Do you have 3 different results or 2 different results? Could you provide more details about the differences, and about the configurations used? Are you using hybrid search in any of the 3?

[nurkmez2]

Thanks for the prompt reply
The result only differs if I send the same query to the Kernel Memory Service running Locally (Item -1). The rest have the same result.

1. configurations for kernel memory service running in a local docker

Logs from this service

2. configurations for the kernel memory service running in the Azure app container, the same as the kernel memory service running in a local docker

Just in case some manual entry

3. The dotnet serverless app running with Azure Search Service