Starstruk Logo MarkStarstruk

v0.59.0

Mesh-LLM/mesh-llmv0.59.0Apr 11, 2026by github-actions[bot]

AI Summary

A feature release that shrinks the UI bundle by 58%, adds image and PDF chat support, and hardens GGUF parser security and symlink safety.

Key Highlights

  • Shrank UI bundle by 58% using simple layout (replacing elkjs).
  • Added image and PDF support to Chat.
  • Hardened GGUF parser, identity key I/O, and symlink safety.
  • Added inference smoke test as a synced CI flow.

New Features

  • UI bundle size reduction
  • image and PDF chat support
  • GGUF parser hardening
  • inference smoke tests

Full Release Notes

## What's Changed
* Shrink UI bundle 58% by replacing elkjs with simple layout by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/218
* chore(react): add a quick recipe to speed up UI dev using public mesh by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/221
* Harden GGUF parser, identity key I/O, and symlink safety by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/216
* Remove unused Blackboard MCP code paths to eliminate Rust warnings by @i386 in https://github.com/michaelneale/mesh-llm/pull/224
* display: rename misleading VRAM labels to capacity by @Bortlesboat in https://github.com/michaelneale/mesh-llm/pull/222
* Fix MLX ref resolution, shard expansion, and search ordering by @i386 in https://github.com/michaelneale/mesh-llm/pull/223
* Add unit tests for error body parsing and blob-URL retry suppression by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/215
* ci: reorganize mesh-llm builds for 50% speedups using warm build caches by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/209
* Chat: image & PDF support, interrupt-on-queue, stop UX by @michaelneale in https://github.com/michaelneale/mesh-llm/pull/228
* fix: strengthen cache-warming job to fail on resource exhaustion by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/230
* Enforce owner-attested config sync with trust policy and reliability fixes by @i386 in https://github.com/michaelneale/mesh-llm/pull/194
* model resolution consistent across models show, models download, and serve --model by @i386 in https://github.com/michaelneale/mesh-llm/pull/227
* fix(warm-caches): fix the syntax issue for env in the warm job by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/236
* feature(smoke-test): add inference smoke test as synced CI flow by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/237
* fix(docs): remove stale docs + small updates by @ndizazzo in https://github.com/michaelneale/mesh-llm/pull/239
* Add MoE ranking publishing workflow and ship llama-moe-analyze in release bundles by @i386 in https://github.com/michaelneale/mesh-llm/pull/244
* [codex] Fix release publishing for tagged builds by @i386 in https://github.com/michaelneale/mesh-llm/pull/246
* [codex] Scope owner and trust flags to the commands that use them by @i386 in https://github.com/michaelneale/mesh-llm/pull/245
* [codex] Keep Linux release binaries portable across runners by @i386 in https://github.com/michaelneale/mesh-llm/pull/249


**Full Changelog**: https://github.com/michaelneale/mesh-llm/compare/v0.58.0...v0.59.0