devswha/patina v4.0.0

on GitHub

4.0.0 — surface reset

Patina 4.0.0 trims the public CLI down to the surfaces that are easiest to explain, audit, and maintain.

Changed

• Reworked the README into a shorter landing page focused on demo, quick start, common commands, CI, and core docs.
• Expanded --list-backends into backend diagnostics with kind, selector hints, default models, auth status, and setup notes.
• Kept provider presets on the HTTP backend path so --provider gemini uses the Gemini HTTP API instead of being routed through local-CLI model heuristics.

Added

• Private-asset leak gate for npm/source packaging checks.
• Centralized backend default-model metadata for OpenAI/Codex, Claude, Gemini, and Kimi.

Removed

• Hosted backend skeleton and hosted compare harness.
• Standalone MAX mode and composite scorer.
• Deprecated/low-signal flags including variants, save-run/cache options, suspected-generator warnings, prompt-mode exposure, gate/json aliases, share-card output, one-time star nudge, and deprecated inline API key handling.
• patina init; patina stays zero-config with optional manual .patina.yaml.

Distribution

• patina-cli@4.0.0 is published on npm with provenance.
• patina-humanizer@4.0.0 is published on npm with provenance.

Full changelog: https://github.com/devswha/patina/blob/main/CHANGELOG.md#400--2026-06-04