LEXO
မြန်မာ Download

LEXO 08 Changelog

v0.4.1 AGPL-3.0

Changelog

Every release, newest first. This page is generated from the repository's CHANGELOG.md, so it always matches the code.

  1. 0.4.1

    Latest

    Release on GitHub

    Fixed

    • Cropping rotated PDF pages now trims the edge shown in the preview instead of mirroring the crop on pages rotated 90 or 270 degrees.

  2. 0.4.0

    Release on GitHub

    Added

    • Batch PDF OCR is available in both the CLI and desktop app. The new lexo ocr-batch command accepts PDF files or folders and exports one UTF-8 TXT file per source; existing outputs are skipped for resumable runs and can be replaced with --overwrite.
    • The desktop batch workflow supports multi-file or folder selection, output and OCR options, progress, cancellation, Google sign-in recovery, and the Ctrl+Alt+O shortcut.

  3. 0.3.1

    Release on GitHub

    Fixed

    • Splitting a spread PDF whose pages carry a rotation flag no longer drops the far half as a blank page, and re-splitting an already-split PDF no longer produces duplicate halves or loses content. Each half is now taken from the displayed page and anchored to the current cropbox.

  4. 0.3.0

    Release on GitHub

    Added

    • The page preview can be zoomed and panned.
    • Scrolling over the page preview moves between pages.
    • Run page ranges accept all, odd, and even keywords, with quick-select buttons for each.

    Fixed

    • Cropping a PDF page a second time no longer drifts. The new crop is now anchored to the current cropbox, so it insets within the already-cropped region instead of shifting back toward the original page origin.
    • The crop box stays visible when switching between page previews.

  5. 0.2.1

    Release on GitHub

    Fixed

    • PDFs whose text layer is legacy non-Unicode "glyph soup" (e.g. some older Burmese printing fonts encode their text as Private Use Area glyph ids) no longer come back blank. Such a layer is detected as unusable and the page is OCR'd instead of returning meaningless characters.
    • An expired or revoked Google sign-in is now surfaced clearly. The token is validated once before a run, so it fails with a single "sign in again" message instead of every page failing with a cryptic error; is_authenticated no longer reports an unusable token as signed in.

    Added

    • When OCR needs Google sign-in, the desktop app now offers to sign in in place and retries the run on success.

  6. 0.2.0

    Release on GitHub

    Added

    • Per-page OCR results now appear in the GUI as each page finishes, instead of only after the whole run completes.
    • Page edits (delete, rotate, crop, split, append, reorder) run in the background with a busy indicator, so the window stays responsive on large documents. Run/OCR, Save, and page operations are disabled while an edit runs.

    Fixed

    • OCR no longer crashes the app on large multi-page runs. A single Google Drive HTTP client was shared across the engine's concurrent workers; since httplib2 is not thread-safe this corrupted its state and faulted the process (Windows heap corruption). Each worker now gets its own client.
    • OCR returned empty text (or failed to upload with HTTP 413) for PDFs with very large pages. Pages now render as resolution-capped JPEGs that stay within Google Drive's upload and OCR limits.
    • lexo info and --version now report the actual package version (the reported value had drifted from the published version).

    Changed

    • Hash a PDF once per OCR run instead of once per render batch (much faster on large files).
    • Skip rebuilding the whole thumbnail strip when an OCR run finishes; per-page status badges already update live.

  7. 0.1.1

    Release on GitHub

    Changed

    • Use SPDX license expression for verified PyPI metadata.
    • Expand PyPI classifiers: add environment, intended audience, and per-version Python tags.
    • Add Changelog and Releases links to project URLs.
    • Fix README logo URL to absolute path so it renders on PyPI.

  8. 0.1.0

    Release on GitHub

    First public release. Lexo is a complete, from-scratch rebuild of the previous OCR Text Extractor into a local-first desktop document OCR tool. The legacy Tkinter/CustomTkinter application has been removed.

    Added

    • Document-centric engine with a hexagonal design: a UI-agnostic core behind swappable ports, driven identically by the CLI and the GUI.
    • PDF operations: extract page ranges, split, crop, rotate, merge, and split two-up spreads.
    • Smart OCR routing: digital PDF pages use their embedded text layer; only scanned pages are OCR'd (--force-ocr overrides).
    • OCR via Google Docs OCR (free, best Burmese accuracy), behind a pluggable provider port so other engines can slot in.
    • Burmese-aware text handling: NFC normalization and zero-width-space-safe cleaning.
    • Exports: plain text (default), Markdown (with YAML frontmatter), and JSONL.
    • A PySide6 desktop GUI (lexo gui): a top toolbar driving the open -> organize -> run -> proofread -> export flow, a left pages strip with per-page status and reordering, and a page preview beside an editable text pane. A visual crop/split editor for PDFs and images, copy-to-clipboard, and toast confirmations.
    • Typer CLI: info, check-update, extract, ocr, pdf operations, login, logout, and gui.
    • Bundled Noto Sans Myanmar font (SIL OFL) for consistent Burmese rendering.

    Changed

    • Project renamed from ocr-text-extractor to lexo; installed with uv tool install lexo.
    • OAuth scope narrowed to drive.file; tokens are stored in the OS keychain, replacing the previous pickle-based token storage.

    Removed

    • The legacy v1 modules: the Tkinter GUI, the ad-hoc OCR processor, and pickle token storage.
CHANGELOG.md on GitHub