v0.37.2

Mesh-LLM/mesh-llmv0.37.2Mar 19, 2026by michaelneale

AI Summary

Significant performance optimization for Claude Code by disabling attribution headers that were invalidating the KV cache, along with general optimizations for local inference.

Key Highlights

Fixes 90% slowdown in Claude Code by setting CLAUDE_CODE_ATTRIBUTION_HEADER=0
Disables attribution on commits, PRs, telemetry, and progress bars for local inference optimization
Optimized for local model performance

New Features

Performance fix for Claude Code
Attribution disabling for telemetry and progress bars

Full Release Notes

## What's new

- **Fix 90% slowdown in Claude Code**: `CLAUDE_CODE_ATTRIBUTION_HEADER=0` now set by `mesh-llm claude`. Without this, Claude Code prepends a changing header that invalidates the KV cache on every request — massively degrading performance with local models. ([ref](https://unsloth.ai/docs/basics/claude-code))
- Disables attribution on commits/PRs, telemetry, progress bar — all optimized for local inference

### Install (macOS Apple Silicon)

```bash
curl -fsSL https://github.com/michaelneale/decentralized-inference/releases/latest/download/mesh-llm-aarch64-apple-darwin.tar.gz | tar xz && sudo mv mesh-bundle/* /usr/local/bin/
```