date

fix

version 2.3.8

Release 2.2.3: Fix [all] Extra for Complete Installation

🔧 Critical Installation Fix:
Fixed abstractcore[all] to truly install ALL modules including development dependencies.

🎯 What Changed:
• [all] extra now includes 12 dependency groups (was missing dev, test, docs)
• Complete coverage: All providers + features + development tools
• Users can now confidently use pip install abstractcore[all] for everything

📦 Complete Dependencies Now Included:
• All Providers (6): openai, anthropic, ollama, lmstudio, huggingface, mlx
• All Features (3): embeddings, processing, server
• All Development (3): dev (pytest, black, mypy, ruff), test (pytest-cov, responses), docs (mkdocs)

✅ Verification:
• All 12 referenced extras exist and are properly defined
• No circular dependencies or missing references
• Comprehensive configuration tested and verified

🚀 Impact:
• No more partial installations or missing development tools
• Single command installs complete AbstractCore environment
• Development-ready installation with testing and documentation tools

Release 2.2.2: LLM-as-a-Judge Production Implementation

🎯 Major Addition: BasicJudge - Production-ready objective evaluation
• Structured assessments with 9 evaluation criteria
• Multiple file support with global assessment synthesis
• Enhanced assessment structure with judge summary and source reference
• CLI with simple 'judge' command (console script entry point)
• Comprehensive documentation and real-world examples

🚀 Ready for Production:
• Context overflow prevention and error handling
• Chain-of-thought reasoning with consistent scoring (temp 0.1)
• Pydantic integration with FeedbackRetry validation
• Full backward compatibility maintained
• Complete documentation in docs/basic-judge.md

📊 Assessment Features:
• 1-5 scoring scale with clear definitions
• Global assessment appears first for multi-file evaluations
• Optional --exclude-global flag for original behavior
• Custom criteria support and reference-based evaluation
• JSON, plain text, and YAML output formats

🔧 Technical Excellence:
• Sequential file processing to avoid context overflow
• Graceful fallbacks and comprehensive error handling
• Production-grade architecture following AbstractCore patterns
• Extensive testing and documentation coverage

Version 2.1.0: Add vector embeddings support

🎯 Features:
- Vector embeddings with SOTA models (EmbeddingGemma, Granite, etc.)
- Smart caching and ONNX optimization
- Semantic search and RAG capabilities
- Event system integration
- Production-ready performance

🔧 Technical:
- Complete embeddings module in abstractllm/embeddings/
- 16 comprehensive tests with real models
- Comprehensive documentation and examples
- Zero breaking changes

🚀 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>