selvage-lab/selvage

v0.3.0 Breaking

This release includes 2 breaking changes for platform teams planning a safe upgrade.

Published 3mo MCP Developer Tools

✓ No known CVEs patched

✓ No known CVEs patched in this version

Affected surfaces

breaking_upgrade auth

Summary

AI summary

Removed legacy model identifiers and date suffixes in Claude model names.

Full changelog

핵심 변경사항

최신 모델 추가: GPT-5.2-Codex (OpenAI), Claude Opus 4.5 (Anthropic), MiniMax M2.1, GLM-4.7 등 최신 AI 모델 지원
모델 설정 단순화: Claude 모델에서 날짜 접미사와 -thinking 변형 제거, thinking 모드 기본 활성화
MCP 프로토콜 보호 강화: stdout 오염 방지를 위한 출력 스트림 관리 개선
고급 추론 모드 활성화: 모든 주요 모델에 high reasoning 설정 적용

세부 구현사항

모델 업그레이드:
- gpt-5, gpt-5-high, gpt-5-mini → gpt-5.2-codex (reasoning_effort: high)
- gemini-2.5-pro/flash → gemini-3-pro/flash
- claude-sonnet-4.5-20250929, claude-sonnet-4.5-20250929-thinking → claude-sonnet-4.5 (thinking 기본)
- claude-opus-4.5-20251101 → claude-opus-4.5 (thinking 기본)
MCP 서버 개선:
- FastMCP show_banner=False 옵션 추가
- warnings를 stderr로 리다이렉트
- BaseConsole 지연 초기화 패턴 도입 (Proxy 패턴)
- 로깅 시스템의 MCP 모드 인식 개선
추론 설정 강화:
- gpt-5.2-codex: reasoning_effort "high"
- minimax-m2.1: reasoning.max_tokens 16000
- glm-4.7: thinking.type "enabled", max_tokens 16000
- claude-*: budget_tokens 32000

Removed legacy model identifiers such as `gpt-5`, `gpt-5-mini`, and `gemini-2.5-flash`.
Claude model names no longer include date suffixes; e.g., `claude-sonnet-4.5-20250929` → `claude-sonnet-4.5`.

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track selvage-lab/selvage

Get notified when new releases ship.

About selvage-lab/selvage

LLM-based code review MCP server with AST-powered smart context extraction. Supports Claude, GPT, Gemini, and 20+ models via OpenRouter.