chore: release v0.2.0

Bump metrillm and metrillm-mcp to 0.2.0.
Update CHANGELOG, MCP server.json, and README.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: TheBlueHouse75

CHANGELOG.md

@@ -7,10 +7,29 @@ and this project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.ht
 
 ## [Unreleased]
 
+## [0.2.0] - 2026-03-03
+
 ### Added
 
-- Homebrew formula (`Formula/metrillm.rb`) and maintenance script (`scripts/update-homebrew-formula.sh`).
-- README installation instructions for Homebrew tap usage (`brew tap MetriLLM/metrillm && brew install metrillm`).
+- CPU load measurement during inference — surfaces the MLX vs GGUF/llama.cpp difference (avg + peak %).
+- Bench environment context: thermal pressure, swap delta, battery status detection.
+- LM Studio runtime backend with full CLI, menu, and MCP support.
+- Homebrew formula (`Formula/metrillm.rb`) and tap installation support.
+- DB rescore migration script for batch re-scoring of existing results.
+
+### Changed
+
+- Performance score reweighted from 40/30/30 to 50/20/30 (speed/TTFT/memory).
+- Global score reweighted from 40/60 to 30/70 (hardware/quality).
+- Default Supabase config embedded for zero-config upload.
+
+### Fixed
+
+- LM Studio streaming `evalDuration` now uses token timing for accurate tok/s.
+- CLI exits cleanly after interactive menu quit.
+- Experimental Node TS stripping gated behind env var.
+- Trailing runtime control tokens stripped robustly.
+- Model format derived per-model for LM Studio uploads.
 
 ## [0.1.0] - 2026-03-02
 

README.md

@@ -258,7 +258,7 @@ The tap formula lives in `Formula/metrillm.rb`.
 ./scripts/update-homebrew-formula.sh
 
 # Or pin a specific version
-./scripts/update-homebrew-formula.sh 0.1.1
+./scripts/update-homebrew-formula.sh 0.2.0
 ```
 
 After updating the formula, commit and push so users can install/update with:

mcp/package-lock.json

@@ -1,6 +1,6 @@
 {
   "name": "metrillm-mcp",
-  "version": "0.1.2",
+  "version": "0.2.0",
   "mcpName": "io.github.MetriLLM/metrillm",
   "description": "MCP server for MetriLLM — benchmark local LLMs from any AI coding assistant",
   "type": "module",

mcp/package.json

@@ -7,12 +7,12 @@
     "source": "github",
     "subfolder": "mcp"
   },
-  "version": "0.1.2",
+  "version": "0.2.0",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "metrillm-mcp",
-      "version": "0.1.2",
+      "version": "0.2.0",
       "transport": {
         "type": "stdio"
       },

mcp/server.json

@@ -1,6 +1,6 @@
 {
   "name": "metrillm",
-  "version": "0.1.1",
+  "version": "0.2.0",
   "description": "Benchmark your local LLM models — speed, quality & hardware fitness verdict",
   "type": "module",
   "bin": {