All notable changes to OmniCommand (omx-cmd) will be documented here.
Format follows Keep a Changelog. This project uses Semantic Versioning.
- First-class batch operations for convert, compress, trim, extract, and resize commands.
- Shell completion generation for bash, zsh, and fish via
omx completion <shell>. - FFmpeg progress parsing for long media jobs with percent, ETA, and output size when available.
- Exported TypeScript interfaces and a stable package library entrypoint.
- Bundled pure-JavaScript DOCX→Markdown/text converter (
convertDocx) that needs no external Pandoc, preserves unicode/emoji/complex scripts, and renders tables as clean aligned GFM pipe tables (via mammoth → rehype/remark, replacing the previous turndown-based HTML conversion). - PDF→Markdown now detects simple aligned tables and renders them as GFM pipe tables (heuristic; scanned or complex layouts still use
--refine). - Image conversion/compression now respects EXIF orientation so camera photos don't come out rotated.
- Gemini OCR and MCP SDK packages are now optional/lazy-loaded so standard offline conversion installs stay lighter.
- Documentation now consistently distinguishes OmniCommand the product,
omx-cmdthe npm package, andomxthe terminal binary. - The root website is documented as a landing/demo page, not a hosted converter UI.
- DOCX→Markdown/text now uses the built-in converter by default (with Pandoc as a fallback); Pandoc is only required for legacy
.doc, PPT/Excel, and RTF. - Image compression with an absolute size target (
200kb,2mb) now actually lands at/under the requested size via a quality search instead of a fixed-quality guess.
- Audio-only compression is now handled correctly. Previously, compressing
.mp3or.wavfiles would incorrectly apply the-c:v libx264video codec, leading to processing failures or corrupted output. Audio files now use dedicated bitrate-targeted compression paths (-b:a). - Improved TXT → PDF output layout. Reduced default margins from 1.0in to 0.75in and added a
1.15line-stretch to ensure generated PDFs fill the page more effectively and are more readable. - CLI integration test race conditions resolved. Fixed a critical testing bug where parallel Vitest workers were intermittently deleting the
dist/folder during integration tests. NODE_ENV=testcommand bypass fixed. Integration tests now correctly execute the CLI logic by explicitly clearing thetestenvironment flag that was causing index commands to be skipped.omx compressnow supportskbtargets in addition tomband%for more granular control over output sizes.- Strict clamping for compression percentages. Percentage targets are now strictly clamped between 1% and 100% to prevent engine crashes on invalid inputs.
- Major Test Suite Expansion. Reached 125 tests (up from ~60) covering extreme edge cases:
- Unicode and special character file names (
문서.pdf,photo (1).png). - Zero-byte and corrupted file handling.
- Directory-as-input protection.
- Transparent PNG → JPEG flattening.
- Double extension handling (
file.test.image.png).
- Unicode and special character file names (
- Comprehensive coverage of global flags (
--json,--dry-run,--quiet,--verbose) across all commands to ensure consistent behavior.
- Image compression now actually reduces file size. Previously
omx compress photo.png to 50%produced a file 17% larger due to hardcodedquality: 50+palette: truesettings that inflated PNGs. PNG is a lossless format and cannot be meaningfully re-compressed — compression now routes PNG to WebP automatically, which is typically 90%+ smaller. --jsonoutput is now clean JSON only. PDF warning messages (two-column layout detection, refine suggestions) were leaking to stdout before the JSON object, breakingJSON.parse()in scripts and CI. All human-readable output is now suppressed when--jsonis active.- Running
omxwith no arguments now exits 0. Previously dumped Commander's internal help and exited with code 1 (treated as an error). Now shows a friendly 4-line example block and exits cleanly. NO_COLORenvironment variable now works. Previously only the--no-colorflag disabled colours. The NO_COLOR standard (NO_COLOR=1 omx ...) is now respected.- MCP server version stays in sync with package version. Was hardcoded to
1.0.1regardless of package version.
omx config get <KEY>— retrieve stored configuration values (onlysetexisted before).- SIGINT/Ctrl+C handler for video operations — cleans up partial output files on cancellation.
--verboseflag is now wired: shows input/output sizes and compression ratio after each operation.- Non-TTY detection — spinners are silenced when stdout is piped (e.g. in CI).
- Format typo suggestions —
omx convert file.pdf to jpgenow suggestsjpg. - Post-compression size warning if output is not smaller than input.
prepublishOnlynow copies bothREADME.mdandLICENSEfrom repo root intocli/before publishing.filesarray inpackage.jsonnow explicitly includesLICENSEandREADME.md.- Keywords expanded from 6 to 18 for better npm discoverability.
repository,homepage, andbugsfields added topackage.json— npm sidebar now shows links.- Landing page GitHub links corrected to
github.com/Rishet11/OmniCommand(were pointing to a 404 URL).
- Internal engine improvements and document format support expansion.
- Added
.docx,.doc,.pptx,.xlsxdocument handling. - Expanded media format whitelist.
- Initial release.
- Five core commands:
convert,compress,trim,extract,resize. - Multi-format engine: FFmpeg (video/audio) + Sharp (images) + Pandoc/pdfjs-dist (documents).
- PDF text extraction with two-column layout detection.
- AI document refinement via Gemini Vision (
--refineflag). - MCP server for agentic integration (convert, compress, trim tools).
omx doctorsystem health check.omx config setfor storing API keys.- Global flags:
--json,--quiet,--overwrite,--dry-run,--no-color.