Changelog
Release notes, updates, and what's coming next.
v2.1.1
Latest2026-03-15- featDetect & Heal 闭环作为核心卖点重新包装 — 业界首个自愈式代码审查
- feat`ocr heal` 现在支持所有 8 个 LLM Provider（GLM、DeepSeek、Ollama 等）
- featCLI 参数别名：`--level` = `--sla`，`--api-key` = `--ai-remote-key`
- docs新增 Detect & Heal 完整工作流文档 + CI/CD 自愈集成指南
- docsProvider 列表更新，标注免费 Provider（GLM、Ollama）
- fix修复版本号同步问题
v2.1.0
2026-03-15- featL3 Deep Scan — Remote LLM analysis with 8 provider support (OpenAI, GLM, ZAI, DeepSeek, Together AI, Fireworks, Anthropic, OpenAI-compatible)
- featMulti-Language AI Detectors — Go, Java, Kotlin, Python language-specific AI defect detection
- featAI Auto-Fix — `ocr heal` command for automatic code fixing with dry-run mode
- featIDE Rules Generation — Auto-generate rules for Cursor, Copilot, and Augment
- featUniversal Provider Adapter — OpenAI-compatible adapter for any LLM service
- featProvider preset configuration (auto-fill baseUrl)
- featConfig file support (ocr.config.json)
- featEnhanced CLI with --provider, --api-key, --model, --api-base options
- fixLanguage-specific detector severity type fix
- fixCLI workspace dependency resolution
v1.2.0
2026-03-14- featGitLab CI Component support — one-line integration
- featL2 deep analysis with hallucination detection for npm packages
- featSelf-scan mode — validate OCR itself
- fixImproved logic gap detection accuracy
- docsComprehensive documentation site with quick start guide
v1.1.0
2026-03-01- featAI Self-Heal Loop — generates fix prompts for Copilot/Cursor
- featCustom threshold configuration via .ocrrc.yml
- featJSON output format for CI integration
- fixReduced false positives in hallucination detection
v1.0.0
2026-02-15- featInitial release — Hallucination detection, Logic gap analysis, Quality Score
- featGitHub Actions integration
- featLocal Ollama model support
- featCLI with npx support
- docsREADME with installation and usage guide
Coming Soon
- •VS Code extension — real-time quality feedback in editor
- •Team dashboard — centralized reports and trend analysis
- •Custom rules engine — define your own quality checks