RagBites

Basic text-chunking strategies for RAG.

Key features

Basic chunking strategies
Fixed-size character, character + overlap, fixed-size word, greedy word, sentence, paragraph, and recursive.
Interactive visualizer
Paste any text, tune each method's parameters with per-card sliders, and watch the chunks update live.
Zero dependencies
Python stdlib only; the web app is a single static page that deploys anywhere.

#	Function	Splits by
1	`basic_chunk`	a fixed number of characters
2	`overlap_chunks`	characters, with overlap between chunks
3	`word_chunks`	a fixed number of words
4	`greedy_word_chunks`	words, greedily filled up to a char budget
5	`sentence_chunks`	whole sentences
6	`paragraph_chunks`	blank-line paragraphs
7	`recursive_chunks`	the largest separator first, recursing down

Run the visualizer locally (Python stdlib, no install):

git clone https://github.com/januaryofmine/rag-bites
cd rag-bites
python -m http.server 8000   # -> http://127.0.0.1:8000

Use the chunkers directly in Python:

import chunk

text = "Your document text..."
print(chunk.recursive_chunks(text, max_chars=150))

File	Role
`chunk.py`	the 7 chunkers, pure Python (reference)
`chunk.js`	1:1 JS port that powers the static demo
`index.html`	the visualizer UI
`app.py`	original Python `/api/chunk` server (kept for reference)