Fixed LLama.Web error starting session

[tonib]

This should fix #1080

InferenceOptions.SamplingPipeline now is not required. Is not sent from client. This was the main issue for #1080
Added options to debug SignalR calls at appsettings.Development.json. If you dont have this enabled, SignalR serialization errors are not logged (even with Warning level)
Changed default web model to llava-v1.6-mistral-7b.Q3_K_XS.gguf, previous one (llama-2-7b-chat.Q4_0.gguf) did no exist

[martindevans]

Thanks for looking into this

[Copilot]

Copilot reviewed 1 out of 3 changed files in this pull request and generated 1 comment.

Files not reviewed (2)

• LLama.Web/appsettings.Development.json: Language not supported
• LLama.Web/appsettings.json: Language not supported

[tonib]

I see LLama.Unittest.Native.SafeLlamaModelHandleTests.MetadataValByKey_ReturnsCorrectly failed.
Strange. I runned tests on my machine (Windows) in Debug/Release and they passed.
It seems MetadataValueByKey() in MetadataValByKey_ReturnsCorrectly returns null, but i did not changed anything related to this

[martindevans]

Yeah I don't think that's your bug. I can't reproduce it locally either though :/

[martindevans]

Even though the CI is failing I'll merge this one, I asked a few other maintainers and we're pretty confident the bug isn't in your changes.