Thin client concept
Scan to searchable PDF
The launcher is just an entry layer: pick a file, choose a preset, inspect the dependency snapshot, then hand the actual work to the local engine.
Output: searchable PDF
Needs: poppler + tesseract
Risk: long OCR jobs
Drop a PDF here or choose a file
This surface is the main affordance: it keeps the first move obvious and keeps all the complexity in the system panel, where a user can see it.
Command preview
Ready to OCR a scanFlow anatomy
What the launcher exposes, not what it hides
1
File selection and quick validation.
2
Preset choice with dependency hints.
3
Run, preview, export, repeat.
Dependency status
Capability-led startup checks
qpdf
Split, merge, rotate, protect.
level 0
poppler
Text extract, render, fonts.
level 1
tesseract
OCR write-back and language packs.
level 2
ghostscript
Compression and size reduction.
optional
Run queue
Recent and scheduled actions
OCR batch
12 files, quiet mode.
Unlock + extract
Waiting for password.
Doctor snapshot
Fresh capability probe.
Notes
Why the launcher stays thin
Keep the first step obvious
People should see doctor, file selection, and the first preset without hunting.
Expose capability, not clutter
Show the dependency snapshot and hand off the work to the engine.