OfficeIMO.Reader.Tool 3.0.0

OfficeIMO.Reader.Tool

OfficeIMO.Reader.Tool exposes OfficeIMO's local document readers as a .NET global tool. It reads files or standard input, converts bounded folders concurrently, emits Markdown or the stable v5 JSON envelope, and materializes embedded assets.

Install

dotnet tool install --global OfficeIMO.Reader.Tool

Read a file or standard input

officeimo-reader read policy.docx --format markdown --output policy.md
officeimo-reader read report.pdf --format json --output report.reader.json --assets report-assets
Get-Content notes.md -Raw | officeimo-reader read - --name notes.md
Get-Content archive.json -Raw | officeimo-reader read - --name archive.json --max-input-bytes 134217728

--output - or an omitted output path writes to standard output. --name gives piped bytes an extension so Reader can choose the intended handler; its default is stdin.txt. The read command bounds files and standard input to 64 MiB by default; use --max-input-bytes for an explicit positive-byte override.

Convert a folder

officeimo-reader folder ./documents `
    --output ./converted `
    --format json `
    --assets ./converted-assets `
    --concurrency 4 `
    --max-files 500 `
    --max-input-bytes 67108864

Folder output preserves relative source paths and adds .md or .reader.json. Each document has the same 64 MiB default input ceiling as read; use --max-input-bytes to override it. Discovery streams filesystem entries, stops at --max-files, skips reparse points, and sorts the selected paths before conversion; --max-total-bytes provides an optional aggregate input ceiling. Use --no-recursive for the top directory only. Output and asset paths are resolved through existing symbolic links or junctions and must remain outside the input tree.

Inspect capabilities

officeimo-reader capabilities
officeimo-reader capabilities --format json

Exit codes

Dependency boundary

The tool uses OfficeIMO.Reader.All and the existing local adapter graph. It has no third-party command parser, process launcher, native binary, model, network client, or hosted provider. Standalone images are projected as metadata/assets without OCR, notebooks are never executed, and SRT/WebVTT files are parsed locally. Audio decoding, Whisper, YouTube downloads, and OCR engines remain explicit host dependencies and are not configured.