v0.58.0

Mesh-LLM/mesh-llmv0.58.0Apr 8, 2026by github-actions[bot]

AI Summary

A maintenance release adding pre-warming of llama.cpp CUDA caches and fixing spurious 503 errors that occurred during slow prefill operations.

Key Highlights

Added warm-caches CI job to pre-warm llama.cpp CUDA cache.
Fixed spurious 503s on slow prefill operations.
Removed tunnel double-proxy to improve performance.

New Features

CUDA cache pre-warming
spurious 503 error fix
tunnel double-proxy removal

Full Release Notes

## What's Changed
* ci: add warm-caches.yml to pre-warm llama.cpp CUDA cache on main by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/211
* Justfile: scope clean-ui Unix variant with [unix] to avoid duplicate on Windows by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/208
* Fix spurious 503s on slow prefill and remove tunnel double-proxy by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/213


**Full Changelog**: https://github.com/michaelneale/mesh-llm/compare/v0.57.0...v0.58.0