v0.9.1

Added

• Data Science Agent - Specialized agent for data analysis and machine learning tasks with Jupyter kernel integration
• Multi-agent Supervisor - Orchestrates between coding and data science agents based on task requirements
• Enhanced grep tool - Smart pattern retry with case-insensitive fallback and partial match variations
• glob tool - Fast pattern-based file searching (e.g., **/*.py, src/**/*.ts)
• Todo persistence - Saves todos to .katalyst/todo.json with automatic loading on startup

Fixed

• Task progress counting - Shows correct "Task 1 of N" in terminal output
• Checkpointer serialization - Resolved singleton manager issues for conversation persistence
• LangGraph imports - Updated to latest module structure
• License format warning - Changed from deprecated TOML table to simple string format
• grep tool max_results - Properly handles None values with sensible defaults

Changed

• Tool improvements:
  • Renamed tools for Unix familiarity: read_file → read, list_files → ls, execute_command → bash, regex_search_inside_files → grep
  • Added respect_gitignore parameter to read tool
  • Removed redundant move_or_rename_path tool (use bash mv/rename instead)
  • Replaced apply_source_code_diff with edit and multiedit tools for simpler file modifications
• Data Science agent folder structure - Uses data/ for datasets and docs/ for documentation
• Test organization - Moved to pytest.ini configuration and removed outdated tests
• Improved prompts - More concise tool descriptions for better LLM performance

Removed

• Obsolete test files - Cleaned up tests for removed features (tool_runner, adaptive planning, etc.)
• Duplicate prompt files - Consolidated tool prompts into single location
• Legacy code - Removed commented-out modules and unused imports