microsoft / kernel-memory

RAG architecture: index and query any data using LLM and natural language, track sources, show citations, asynchronous memory patterns.
https://microsoft.github.io/kernel-memory
MIT License
1.52k stars 293 forks source link

Fix dotnet publish // LLamaSharp conflicts #500

Closed dluc closed 4 months ago

dluc commented 4 months ago

Motivation and Context (Why the change? What's the scenario?)

LLamaSharp doesn't allow to include multiple backends at once in a packaged .NET app, requiring to build once for each backend (CPU, CUDA11, CUDA12, etc). The current code fails to "publish" due to concurring backends.

See also https://github.com/microsoft/kernel-memory/issues/498

High level description (Approach, Design)

Solve the "publish" issue by removing all LLamaSharp backends, leaving it to the end user/developer to choose one.

The need to build multiple times, once per LLamaSharp backend, persists, due to how LLamaSharp is designed. (if anyone knows of a workaround please let us know!)