Release pi-commandcode-provider 0.6.1 · patlux/pi-commandcode-provider

Summary

Selectable thinking levels (minimal … xhigh) for meta/muse-spark-1.1, meta/muse-spark-1.2, and meta/muse-spark-1.2-contributor via a catalog override; /thinking and Shift+Tab no longer stay locked on off.
Cache-first startup: the cached model catalog registers immediately and refreshes in the background, so pi no longer waits for a network round-trip on every launch (#63).
Register the commandcode-custom API in the @earendil-works/pi-ai/compat registry so sibling extensions (background agents, memory workers) streaming with the active Command Code model work on plain pi (#68).
Preserve host developer messages on the /alpha/generate transport instead of dropping them (#48).
/commandcode-quota shows the monthly renewal date and remaining days (#62).
Catalog synced to command-code@1.40.1: adds Qwen/Qwen3.8-Flash, z-ai/glm-5.3-flash, claude-fable-5-1, deepseek/deepseek-v4-flash-fast, tencent/hy4-preview; removes stealth/ox-alpha and minimax/minimax-m3-free (#64, #65).
Display pricing refreshed for all 62 models; expired Claude Sonnet 5 and Gemini 3.7 Flash promotions ended.
Catalog tests assert structural invariants instead of pinned model ids, so the daily catalog sync no longer fails on upstream changes (#67).
refresh-model-catalog agent skill and Windows fix for the catalog scripts (#66).

npm test (214 unit tests, pi-local end-to-end against a mock Command Code API), typecheck, format:check, git diff --check
npm audit --audit-level=moderate: 0 vulnerabilities
npm run check:commandcode-catalog: current (command-code@1.40.1)
Live E2E: Go profile PASS (/alpha/generate transport), GOAT profile PASS (Provider API transport, vision via gpt-5.6-luna)
Muse Spark override verified live on Go and GOAT: --thinking off/high/xhigh sends reasoning_effort undefined/"high"/"xhigh"
Manual model-metadata.yml run on main after the merges: green, no drift
CI green on release commit 008765f (#73)