Topic: router

The ggml/llama.cpp server was updated to fix router eviction race conditions by routing all model loads through a queue, ensuring that no model is evicted prematurely during loading. This update is crucial for improving the stability and reliability of model handling in the server.