COMING Q2 2026

Knowledge Distiller

Drop a PDF, image, or presentation into your Drive folder. Get back a high-fidelity markdown knowledge vault file — structured, fidelity-scored, and ready to feed any AI system. No prompts. No dashboard.

Processed in RAM only. Access revoked instantly by unsharing your folder.
WHAT IT PROCESSES

Any document in. Knowledge vault file out.

DROP IN
PDFJPGPNGPPTX
PDFs — reports, research papers, documentation
Images — diagrams, screenshots, scanned pages
Presentations — slide decks in PPTX format
Mix formats freely in one run
Knowledge Distiller~5 min
GET BACK
Knowledge Vault MD file
One structured .md file per input document
Full content extraction
Text, headings, tables, lists — all preserved with hierarchy
Visual descriptions
Images and diagrams described in detail for AI ingestion
Fidelity score per section
Confidence rating so you know where to verify
Inline warnings on low-confidence passages
Flagged with [LOW CONFIDENCE] for your review
AI-ready formatting
Clean markdown optimised for RAG pipelines and knowledge bases
UNDER THE HOOD

What happens after you drop a file

knowledge_distiller.sh

> POLL: Scanning 'Knowledge Distiller' folder in Google Drive...

> DETECT: 3 files found — 1 PDF, 1 PNG, 1 PPTX

> CLASSIFY: Routing by format — PDF (text extraction), PNG (visual description), PPTX (slide decomposition)

> EXTRACT: PDF: extracting structured text, headings, tables, and footnotes...

> DESCRIBE: PNG: generating high-fidelity visual description with layout, color, and content notes...

> DECOMPOSE: PPTX: converting 24 slides — titles, bullets, speaker notes, embedded images...

> SCORE: Calculating fidelity scores per section. 2 low-confidence passages flagged with inline warnings.

> DELIVER: 3 Knowledge Vault MD files uploaded to Drive folder. Email notification sent.

> STATUS: Knowledge Vault MD files delivered to Drive folder. Processing memory purged.

THREE FORMAT PIPELINES

PDF, image, or deck. One agent handles all three.

PDF → MD

PDF Documents

Extracts structured text, headings, tables, and footnotes. Preserves document hierarchy as markdown. Handles multi-column layouts and embedded images.

JPG/PNG → MD

Images & Diagrams

Generates high-fidelity visual descriptions of charts, screenshots, scanned pages, and diagrams. Every image becomes AI-readable context.

PPTX → MD

Presentations

Decomposes slides into titled sections with bullets, speaker notes, and embedded image descriptions. Each slide becomes a navigable markdown block.

Mix formats freely — drop a PDF, an image, and a deck in the same run. Each file gets its own Knowledge Vault MD output with fidelity scoring.

Interested in this agent?

Join the waitlist and we'll notify you when Knowledge Distiller launches. Tell us what you're converting — it helps us prioritise formats and fidelity features.

Join the Waitlist