Skip to content

v0.3.0

Choose a tag to compare

@ayoub-ibm ayoub-ibm released this 22 Jan 14:29
· 142 commits to main since this release

Bug Fixes

  • release: Resolve semantic-release changelog config deprecation

Chores

  • ci: Exclude attestation files from GitHub releases
  • deps: Bump GitHub Actions in the all-actions group (checkout, setup-python, setup-uv, upload/download-artifact) to latest major versions
  • release: Treat refactor commits as minor version bumps

Documentation

  • deps: Add MkDocs and documentation dependencies
  • readme: Fix PyPI badge to include prereleases
  • readme: Update documentation section with MkDocs links
  • repo: Move community health files to .github and update links

Refactoring

  • core: Remove document caching for stateless operation

Core Modifications:

  • Extractor now returns (models, document) tuple instead of caching
  • Removed all caching logic from DocumentProcessor
  • Pipelines updated to use returned document directly
  • Updated ManyToOneStrategy and OneToOneStrategy

Key Benefits:

  • Stateless architecture – no memory accumulation
  • Predictable memory usage – released after each request
  • Pod-safe for Kubernetes deployments
  • No performance penalty – document still available without re-conversion

Impact:

  • Users calling extractor.extract() directly must update:

    • Old: models = extractor.extract(source, template)
    • New: models, document = extractor.extract(source, template)
  • The run_pipeline() API remains unchanged