Adam Sitnik

Triage: it seems that reading a process name on Linux makes some of our tests flaky (#109630, #110643, #111460). @jkotas has suggested in https://github.com/dotnet/runtime/issues/111460#issuecomment-2608999432 that we maybe reading it too...

Every recursive method can fail with StackOverflow when the recursion depth is too deep. For a value like 1000 and async method I am not surprised, having an `[IterationSetup]` may...

> @adamsitnik, I prefer that overload too but this mean we should remove all other overloads and introduce breaking changes. Should we do this? This time I am not 100%...

Why do you register a `Lazy` factory? The DI container won't create it until it's needed for the first time. cc @eerhardt

> cecil should probably switch to TNP for net9-and-above The new `TypeName` API is part of `System.Reflection.Metadata` package, which supports older monikers too:  > Looks like it's an expected...

I am moving the discussion from https://github.com/microsoft/semantic-kernel/pull/11594#discussion_r2049040464 to here so it's easier to discover: > Regular vs. Keyed `Microsoft.Extensions.DependencyInjection` itself and other projects that went over API review process do...

1. Keep Regular and Keyed 2. Focus on `IVectorStoreRecordCollection` first 3. Register both the implementation and abstraction 4. String vs. Fun: have both 5. Have the overloads for most frequently...

A thing that I've just realized during API review: should we also register the collections as `IVectorSearch`? So users can resolve for searching scenarios only?

Closed by https://github.com/microsoft/semantic-kernel/pull/12097

Providing such an information would be very tricky, as we (BDN) use a low level debugger API (ClrMd) to obtain the disassembly. For example over here you can check what...