v0.9.0 v17.11.0

Name: presidio
Author: data-privacy-stack

medium risk

data-privacy-stack/presidioAug 28, 2026 → Aug 28, 20260 days between releases

What's New

Adds OCR memory bounds, watcher config, and fixes a use-after-free bug

Breaking Changes

Default conflict policy changed from OVERWRITE to SUFFIX.
Deprecated OCR_OUTPUT_DIRECTORY_YEAR_MONTH in favor of OCR_OUTPUT_STRUCTURE=YEAR_MONTH.

New Features

Added --max-ocr-image-mpixels to bound OCR memory usage.
Added configurable watcher output structure (FLAT, YEAR_MONTH, HIERARCHY).
Added configurable watcher conflict policy (SUFFIX, SKIP, OVERWRITE).
Enabled concurrent OCR jobs within a single Python process.

Improvements

Improved memory usage for OCR operations via downsampling.
Improved performance for OCR operations via concurrent job support.

Bug Fixes

Fixed latent use-after-free bug in the pypdfium2 rasterizer.

Upgrade Notes

Migrate watcher configuration by replacing `OCR_OUTPUT_DIRECTORY_YEAR_MONTH` with `OCR_OUTPUT_STRUCTURE=YEAR_MONTH`. Note that the default `OCR_ON_CONFLICT` policy is now `SUFFIX`; explicitly set `OVERWRITE` if overwriting output files is required.

The Two Releases

v0.9.0

AHP Swift Package v0.9.0

Aug 28, 2026· 2 months ago

Implements AHP 0.9.0, introducing an automation catalogue, durable turn errors, and updated session lifecycle management.

Full release notes

v17.11.0

Aug 28, 2026· 2 months ago

This release adds configurable output structures and conflict policies for the watcher, improves memory usage and performance for OCR operations, and fixes a use-after-free bug.

Full release notes