Nutter Tools
Nutter Tools

A free toolbox that runs in your browser. A tool for every job.

Your files never leave your device

Images

  • Convert images
  • Compress images
  • Resize images
  • Crop
  • Watermark

PDF & files

  • Merge PDFs
  • Split a PDF
  • Images to PDF
  • PDF to images
  • Remove photo location

Video & audio

  • Compress video
  • Convert video
  • Video to GIF
  • Convert audio
  • Trim audio

Everyday

  • QR codes
  • Passwords
  • Time zones
  • Unit converter
  • Word count
  • Webcam test

Developer

  • JSON formatter
  • Base64
  • Regex tester
  • Hashes
  • JWT decoder

Site

  • About
  • Contact
  • Settings
  • Privacy
  • Terms
  • Site map
  • llms.txt

Everyday

  • Images
  • Video
  • Audio
  • PDF & files
  • Text
  • QR codes
  • Units & math
  • Time & dates
  • Color
  • Passwords & IDs
  • Device tests
  • Sun & moon

For developers

  • JSON
  • Encoding
  • Regex & cron
  • Web & auth
  • Hashing & encryption
  • Config files
  • Network
  • Browser
  • AI
  • Fake data
© Nutter Toolsnutter.tools
  1. Home
  2. AI
  3. RAG chunking simulator
Your files stay on your device

RAG chunking simulator

Split a document with fixed, sentence, paragraph or recursive chunking and preview sizes, boundaries and overlap before you spend money on embeddings.

This tool is in beta

Still being tested. If a result looks wrong, tell us.

Report a problem

Document

Paste text or open a .txt, .md, .markdown, .csv, .json or .html file

Paste a document to see its chunks.

This is a dry run, not an index: nothing is embedded, stored, or sent anywhere, and the token number is the characters divided by 4 heuristic, not a tokenizer count. The sentence scanner treats every period, question mark and exclamation mark as an end, so abbreviations like e.g. split like sentence ends. Overlap carries whole sentences, paragraphs or words (raw characters for the fixed strategy), and a chunk whose single piece is longer than the ceiling is kept whole and flagged rather than cut mid-content. Check the size spread and boundary view here, then commit to the embedding run.

Splitter settings

StrategyWhere the splitter prefers to cut
MeasureEstimated tokens count one token per 4 characters

Chunk size 5 to 100,000 in the chosen measure, overlap below it. Up to 5,000 chunks per run and 200,000 characters of document. Recursive cuts paragraphs first, then sentences, then words.

More AI tools

  • Tokenizer playground
  • Context fit
  • Fine-tune JSONL formatter
All AI tools

Was this tool useful?