v0.6.0

Highlights

• Added native Anthropic /v1/messages passthrough for Claude-family models when GitHub Copilot upstream supports it.
• Refactored backend routing into model-aware planning and runtime fallback, so each request surface only tries APIs the target model can actually use.
• Improved compatibility around structured outputs, Anthropic document handling, image and file validation, and OpenAI/Anthropic translation edge cases.
• Expanded automated coverage with live capability probes, proxy smoke tests, and focused backend-plan tests.
• Refreshed the English and Chinese READMEs to reflect GitHub's newer first-party Anthropic support and clarify this project's role as a local OpenAI/Anthropic-compatible Copilot-backed proxy for Claude Code, Codex, SDKs, and custom tooling.

Verification

• bun run lint
• bun test
• bun run build