Knowledge Pipeline — offline exam dataset studio

Ingest documents, extract strict knowledge records, review them in the staging grid, export a dataset. Everything runs 100 percent in this page and is stored on your device. Internet is only needed the first time the optional libraries load, and only if you use the optional AI extraction.

Records 0

Staged 0

Approved 0

Flagged dupes 0

Storage …

Settings

OpenAI API key — optional, calls the API straight from your browser for AI extraction. Stored only on this device.

Libraries load on demand: PDF reader, Word reader, OCR engine.

1 · Ingest files or text

Source document Section File — txt, md, csv, json, pdf, docx, png, jpg, webp Or paste text

2 · Camera capture

Source label

Point at a book page, capture, and the text is read on-device.

3 · Sidekick commands

Understood: count · export · export csv · export tsv · approve all · clear approved · delete flagged · scan dupes · find words · help

4 · Staging and review grid

Question	Correct answer	Wrong options	Category	Rank	Source	Status

5 · Export downloads

Record schema

Every record holds: question · correctAnswer · distractors as exactly 3 wrong options · category · rank 1 to 5 · context explaining why it matters · source with document, section, method and page · status plus created date. Imports missing any field are repaired or rejected, never half-saved.