furrow 0.1.3

A sleek state-machine parser to segment, extract, and format structured question chains from smashed-together text streams (OCR, handwriting text dumps, transcripts, and LLM outputs).

Furrow

A lightweight, zero-dependency Python package to slice, group, and format messy text streams (like raw handwriting recognition outputs, chaotic OCR blocks, or unformatted LLM dumps) into separate lines without losing any of your data.

Installation

pip install furrow

How to Use It

To format your text cleanly, you need to call the engine's methods sequentially:

from furrow import Plow

# 1. Feed it your raw, smashed-together text block
messy_text = "was1 . i was a girl19.There i with her 500 grapes .7. Amazing!"
engine = Plow(messy_text)

# 2. Run the character boundary scanner (Crucial step!)
engine.run()

# 3. Pull your questions out as a clean list of data blocks
print(engine.collect())
# Output: [{'question_number': '19', 'text': '.There i with her 500 grapes .'}, ...]

# 4. Generate the final text string with line breaks perfectly injected
print(engine.render())
# Output:
# was
# 1. i was a girl
# 19.There i with her 500 grapes .
# 7. Amazing!

The Processing Sequence

Furrow processes your text strings in three distinct, lightweight steps:

• Step 1: engine.run() (The Tokenizer) – Steps through your text character by character to find numbers. It maps out their exact start and end coordinates in the string.
• Step 2: engine.collect() (The Filter) – Checks the distance between the numbers it found and trailing periods. This allows it to figure out the difference between inline data (like 500 grapes) and actual question markers (like 19.).
• Step 3: engine.render() (The Serializer) – Uses the coordinate maps from the previous steps to slice into the original string and drop a clean newline (\n) right before your valid question indices.

Everything that isn't a question (like titles, headers, or instructions) is kept completely safe, unmutated, and untouched.