edfEazyDataFix

Changelog

Every release, grouped by version. Follows Keep-a-Changelog conventions.

v2.0.0Reusable, Reviewable Preparationreleased19 Oct 2026
Added
  • Versioned JSON/YAML PipelineRecipe rules and composable Pipeline workflows
  • Explicit string, integer, number, boolean, datetime and category conversion rules
  • Preserve, reject and quarantine policies with bounded invalid-value evidence
  • Source, recipe and output-bound review plans with selectable processing stages
  • Shared reports and guarded exports across Analysis Ready, ML Ready and Power BI Ready
  • Recipe-bound ML inference artifacts and named multi-table Power BI recipes
  • CLI recipe previews, named tables, validation gates and data/report exports
Changed
  • Existing v1 convenience APIs retain their defaults and return contracts
  • CSV and Excel recipe loading preserves textual identifiers before type inference
  • Transformation evidence now includes counts, dtype changes, source positions and optional full audits
Fixed
  • Duplicate CSV/Excel headings are rejected before pandas can silently rename them
  • Audit coverage supports both pandas 2 and pandas 3 text dtypes
v1.4.1Reproducible Packaging Maintenancereleased21 Sep 2026
Changed
  • Source distributions now include the sample dataset and all packaged test fixtures
  • Added a documented dev extra for the complete quality, test, build and validation toolchain
  • Analysis Ready, ML Ready and Power BI Ready public APIs remain unchanged from v1.4.0
Fixed
  • Replaced the deprecated pandas categorical dtype check with the supported CategoricalDtype API
v1.4.0Transformation-first Readiness Workflowsreleased26 Aug 2026
Added
  • edf.analysis_ready_with_report() — analysis-ready data with scores, changes, warnings and validations
  • edf.ml_ready() — leakage-safe train/test preparation with reusable preprocessing artifacts
  • edf.powerbi_ready() — validated single-table and multi-table Power BI model inputs
  • Key, relationship and cardinality validation with optional continuous date-table generation
Changed
  • Product direction now centres Analysis Ready, ML Ready and Power BI Ready outcomes
  • All v1 public APIs remain compatible under semantic versioning
  • Python 3.10–3.13 support verified for the transformation-first release
v1.0.0Stable Production APIreleased8 Aug 2026
Added
  • edf.run() — unified profile, assessment, controlled cleaning and deterministic EDA workflow
  • Dry-run cleaning, per-column rules and structured before/after change records
  • edf.prepare_with_report() — deterministic preparation diagnostics
  • edf.infer_schema() and edf.validate_contract() — pipeline data contracts
  • Production edf CLI with JSON/YAML configuration, batch processing, JSONL logs and exit codes
  • Stable EazyDataFixError hierarchy and 79 documented public exports
Changed
  • All public v0.5 workflows remain compatible
  • Python 3.10–3.13 support verified for the stable release
v0.9.0Production CLIreleased2026
Added
  • edf command-line entry point
  • JSON and optional YAML workflow configuration
  • Multi-file batch processing, JSON summaries and structured JSONL logs
  • Deterministic exit codes for success, processing failure and configuration failure
v0.8.0Validation and Contractsreleased2026
Added
  • Schema inference with DataContract
  • ContractValidationReport with explicit pipeline pass/fail status
  • Reusable not-null, unique, minimum and maximum quality rules
v0.7.0Preparation and Feature Readinessreleased2026
Added
  • PrepareConfig controls for numeric/date conversion, duplicates, outliers, text and categories
  • PreparationReport with changes, warnings, shapes and data types
  • Threshold-gated conversions and IQR-based outlier capping or dropping
v0.6.0Controlled, Auditable Cleaningreleased2026
Added
  • Configurable missing-value markers and typed per-column cleaning rules
  • Dry-run previews with caller DataFrame preservation
  • Structured CleaningChange audit records
  • Unified edf.run() workflow foundation
v0.5.0Grounded AI Narrativesreleased2026
Added
  • Optional evidence-cited narratives for completed deterministic Agentic EDA workflows
  • Provider-neutral narrative interface and optional OpenAI adapter
  • Citation, numeric, causal-language and workflow-fingerprint guardrails
v0.4.0Notebook Export and Human Approvalreleased2026
Added
  • Ready-to-run deterministic Jupyter Notebook export
  • Human approval checkpoints between planning and execution
  • Dataset fingerprint and dependency validation for approved workflows
v0.3.0Deterministic Agentic EDAreleased2026
Added
  • edf.eda(), edf.plan_eda() and edf.execute_eda()
  • edf.run_agentic_eda() end-to-end deterministic workflow
  • Semantic role detection and traceable priority findings
  • HTML, JSON, Markdown and deterministic PNG report exports
v0.2.1Reports and Polishreleased2026
Added
  • PDF, Excel, CSV and Markdown report exports
  • Console-friendly summary printer
Fixed
  • Assessment handling for empty CSV files
  • Timezone-aware datetime preservation
  • Profiling for all-null columns
v0.1.0Initial Releasereleased2026
Added
  • edf.assess() — dataset quality report
  • edf.fix() — opinionated auto-cleaning pipeline
  • edf.profile() — column-level profiling
  • CSV, Excel and pandas.DataFrame inputs