Add knowledge to your AI
Turn any PDF, book, or scan into markdown chunks with chunks.md, then feed them straight into your AI. One workflow, four platforms.
From document to AI knowledge
The workflow is simple: drop in a PDF, scan, or photo and convert it with chunks.md to get clean markdown. Then upload that markdown to any AI platform as a knowledge file. chunks.md handles the hard part — OCR and formatting, with several models to choose from. Your AI handles the rest — indexing and retrieval.
Markdown is the ideal format for AI ingestion. No layout noise, no broken tables from copy-paste, no garbled headers from raw PDF extraction. Just clean, structured text your AI can actually read.
Got a whole folder? Combine many documents into one AI-ready file to stay under your platform's file limit. Working from a screenshot? Pull the text out first.
Gemini Gems
Open Gemini → Gems → Create Gem. Name your Gem and write instructions describing what it should know. In the knowledge area, upload files or link Google Drive documents — supports PDF, TXT, MD, and CSV.
Save the Gem and every chat with it will draw on your uploaded files as context. Limit is roughly 10 files per Gem. Drive files stay synced — edits in Drive are reflected automatically.
Upload .md files from chunks.md directly, or paste into a Google Doc and link it.
Perplexity Spaces
Open Perplexity → Spaces and create or open a Space. Click Add Sources to upload files — supports PDF, DOCX, XLSX, CSV, MD, JSON, and TXT.
Pro accounts get 50 files per Space at up to 40 MB each. When querying, toggle between Web, your files, or both. Perplexity retrieves relevant sections and cites them inline.
Upload markdown chunks as .md files for clean retrieval with inline citations.
ChatGPT Projects
Open ChatGPT → Projects and create or open a Project. Use Add files in the sidebar to upload PDFs, DOCX, XLSX, CSV, TXT, or code files.
All chats inside that Project share the uploaded files as persistent context. ChatGPT retrieves relevant portions for each question and can reference which file it used.
For books, split into chapter-level markdown files for better retrieval.
Claude Projects
Open claude.ai → Projects and create or open a Project. Click the + button in the Knowledge panel on the right to upload files or paste text directly.
Supports PDF, DOCX, CSV, TXT, HTML, ODT, RTF, and EPUB — up to 30 MB per file. Claude handles large context well, so you can upload entire book extractions at once.
Paste markdown directly into the Knowledge panel for the fastest setup.
Why markdown?
Markdown strips away layout noise and gives AI models clean, structured text. No broken tables from copy-paste, no garbled headers from PDF extraction. chunks.md gives you markdown that's ready to drop into any of these platforms — no cleanup needed.
Related reading
Combine documents for AI
Merge PDFs, Word docs, and notes into one clean markdown file for ChatGPT, Claude, Gemini, or NotebookLM. Beat the file-count limits — all in your browser, nothing uploaded.
Extract text from a screenshot
Pull editable text out of any screenshot in your browser. Free, no signup, and the image never leaves your device — OCR runs fully on-device, output is clean text or markdown.
What is chunks.md?
PaddleOCR v5, SmolDocling, MangaOCR, DeepSeek-based Unlimited-OCR — 8 OCR models running in your browser. PDFs and scans to markdown, zero uploads.