Release v0.0.3

[0.0.3] - 2025-12-29

Added

AI/LLM Agent Integration

• Agent module (clgraph.agent): Build lineage-aware AI agents with LangChain
• Tools module (clgraph.tools): LangChain-compatible tools for lineage queries
  • Lineage tools: trace columns backward/forward, find paths
  • Governance tools: PII detection, impact analysis
  • SQL tools: query generation, validation
  • Schema tools: table/column lookup
  • Context tools: pipeline summary, metadata
• MCP Server (clgraph.mcp): Model Context Protocol server for Claude Desktop integration

Core Features

• to_simplified() method for input/output only lineage graph (filters internal CTEs/subqueries)
• build_subpipeline() convenience method for extracting sub-pipelines
• JSON round-trip serialization: Pipeline.from_json() and Pipeline.from_json_file()
• Template variable support in Pipeline class with template_context parameter
• Validation framework with structured issue reporting (ValidationIssue, add_issue())
• Logging for validation issues at library level (logger: clgraph.validation)
• Enhanced validation for unqualified columns in JOIN conditions
• COUNT(*) resolution to individual columns when schema is known
• Star (*) expansion for cross-query column lineage with EXCEPT/REPLACE support
• API validation mode with auto-generated API dictionary
• __repr__ methods for QueryUnit, QueryUnitGraph, and Pipeline for better debugging

Visualization

• Consolidated visualization functions into library (clgraph.visualizations)
• visualize_pipeline_lineage(), visualize_table_dependencies(), visualize_column_lineage()

Examples

• ClickHouse example with enterprise data pipeline (raw → staging → analytics → marts)
• Enterprise demo with Ollama for local LLM integration

Changed

• Breaking: Renamed package from clpipe to clgraph
• Breaking: Removed GraphVizExporter class (use visualize_* functions from clgraph.visualizations instead)
• Renamed query_lineages to query_graphs for clarity
• Unified column naming for cross-pipeline lineage
• Removed redundant save_metadata, load_metadata, apply_metadata methods (use to_json/from_json instead)
• Filter redundant input star nodes from lineage graph
• Updated minimum sqlglot version to >=28.0.0

Fixed

• Pin Airflow to 2.x for API stability (3.x has breaking changes)
• Handle sqlglot 28.x breaking change in EXCEPT/REPLACE key names
• Exclude star nodes from simplified lineage view
• SELECT queries without destination now treated as virtual result tables ({query_id}_result)
• Sanitize Graphviz node IDs to avoid colon port syntax issues
• Handle Schema objects in multi_query with version fallback
• Fix metadata propagation with two-pass approach

Documentation

• Revamped README with user-focused messaging and updated examples
• Added architecture diagram
• Added illustration and expanded introduction with use cases
• Comprehensive docstrings and output examples in README