Changelog

Validation Filter release notes

v1.1.0 - 2026-04-14

Changed

• Migrate Dockerfile from deprecated filter_base to standard python:3.11-slim build
• Migrate CI/CD to shared premium release workflow (GAR premium-filters/ registry)
• Remove legacy JFROG-based ci.yaml
• Bump openfilter dependency to >=0.1.27
• Update Makefile IMAGE to premium-filters/filter-quality-gates
• Add security-scan.yaml with weekly Grype scans

Changed

• Relax openfilter version constraint from exact pin to ~=0.1.0 for cascade rebuild compatibility

v1.0.0 - 2025-11-18

Changed

• BREAKING: Renamed FilterAlHaytham to FilterQualityGate
• BREAKING: Renamed FilterAlHaythamConfig to FilterQualityGateConfig
• BREAKING: Renamed package from filter_al_haytham to filter_quality_gates
• Updated all imports, examples, and documentation to use new naming
• Changed filter IDs from al_haytham to quality_gate in all examples
• Updated HTML report titles from "Al-Haytham" to "Quality Gates"

Changed - Calibration Workflow

• BREAKING: Removed built-in calibration record mode
• Golden file generation now uses OpenFilter's Recorder filter instead
• Simplified offline process: use Recorder to save golden file, then use Stub + Quality Gates to compare
• Updated documentation to reflect new workflow using Recorder filter

Added

• Comprehensive test suite covering config, comparators, filter, and aggregator
• Consolidated documentation into single overview.md file

v0.2.9 - 2025-01-26

Changed

• Updated documentation

v0.2.8 - 2025-08-08

Added

• OCR benchmarking Capability

v0.2.7 - 2025-07-15

Updated

• Migrated from filter_runtime to openfilter

Added

• Internal improvements: make install uses editable mode by default

v0.2.6 - 2024-03-26

Added

• Internal improvements

v0.2.0 - 2025-02-11

Added

• Edge mode support. It works with standalone filter_runtime filters.

v0.1.0 - 2024-11-26

Added

• Initial Release: new Validation Filter for verifying prediction output via frame-by-frame comparison. It reads two json files and outputs a comparison between them, frame by frame and aggregated.

• Benchmarking via JSON Comparison

  • Compares frame-level data between predictions and ground truth
  • Computes metrics for per-frame and per-ROI correctness

• Golden File Generation

  • Supports automatic creation of ground truth JSON (create_golden_file: true)
  • Switches to benchmarking mode if a ground truth file exists

• Offline Mode Support

  • Simulates pipeline behavior using saved prediction JSON (prediction_json_path)

• Configurable Matching Thresholds

  • Tune ROI matching using iou_threshold and classification comparison using acc_threshold

• Debug and Visualization

  • Optional draw_visualization flag to enable drawing during verification
  • Verbose logging via debug: true

• Structured Output Reports

  • Generates per_frame_results.json and per_roi_results.json with detailed benchmark results