Changelog¶

All notable changes to CraftyCannon are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased ¶

Smart Redaction can now render detected regions as either pixelation or solid black boxes. Settings -> Advanced includes a Use black boxes for auto redaction toggle; leaving it off keeps the existing pixelated behavior.
The editor includes a manual Black Redact destructive tool for dragging a CIA-style black box over known-sensitive regions.
Image upload preparation can strip metadata before upload and transcode images to a selected output format: PNG, JPEG, GIF, or TIFF.
Added release_app.py, a local release helper that builds, signs, notarizes, installs, and restarts CraftyCannon.

Upload-time auto-redaction and the editor's Detect Sensitive action now use the shared redaction render-mode preference.
Image upload preparation now tracks temporary prepared files consistently across redaction, metadata stripping, transcoding, re-upload, local mirroring, post-upload tasks, and secondary S3 copies.
File naming previews now reflect the selected image upload format extension.
Zipline uploads send x-zipline-filename and x-zipline-file-extension headers derived from CraftyCannon's sanitized upload filename.
Updated feature, user-guide, redaction, and README documentation for black-box redaction, image format controls, metadata stripping, and release automation.

Added coverage for black-box redaction placement and temporary PNG rendering.
Added coverage for image metadata stripping, EXIF orientation normalization, upload format transcoding, and Zipline filename headers.

Smart Redaction and the manual blur/pixelate tools applied the filter to a vertically mirrored region, leaving the sensitive content visible. Two separate bugs compounded here:
Detection: Vision reports text observation bounding boxes in bottom-left-origin coordinates (like faces and barcodes), but they were treated as top-left, double-flipping the box.
Rendering: filterRegion drew the filtered patch into the y-up bitmap context using a raw top-left y, mirroring it vertically. Both are fixed; the redaction now lands on the detected/selected region. Verified end-to-end against a rendered email capture, and the processor test now exercises the real PNG-load path (the previous test built its image with a hand-flipped buffer that masked the mirror).

Add GitHub Actions CI that builds and runs the test suite via xcodebuild (XcodeGen-generated project) and verifies the build.sh distribution path on every push and pull request.
Define an explicit shared CraftyCannon scheme in project.yml so the test target is reproducibly discoverable by xcodebuild.

Smart Redaction — a Vision/ImageIO-backed detector that finds sensitive regions in a capture and pixelates them. Detects faces, barcodes, and a wide range of text PII via OCR: email addresses, phone numbers, credit card numbers, IPv4/IPv6 and MAC addresses, URLs/domains, API keys, AWS access keys, GitHub tokens, OpenAI keys, bearer tokens, JWTs, private-key blocks, session cookies, password fields, environment variables, file paths, and usernames/hostnames.
"Detect Sensitive" editor action that runs detection on the current image and applies redactions in place.
Upload redaction policy (Off, Ask before upload, Auto-redact) so captures can be scanned and redacted before they leave the machine.
Redaction settings UI: per-detector toggles, a minimum-confidence slider, a fast-OCR mode, an option to allow raw match previews, and a reset-to-defaults control.
Persisted redaction detector settings and upload policy in runtime preferences.

Link ImageIO.framework (build script and Xcode project) for redaction image handling.

Initial CraftyCannon snapshot: screen capture, image editor, and upload destinations.
Local OCR indexing and search.
OLED theme and Zipline S3 mirroring.
AWS CLI S3 mirror onboarding.
Cloudflare allowlist automation and endpoint validation.
Paste links for Discord targets.