Less Dependencies

[Mohyoo]

Hi! I liked the idea of your library, fast & simple.
I need a tiny version of it for my AI API chat program.
I'm trying to build it with minimal dependencies so, can you help me or at least show me how?
All I need is to compress user prompts no more.
Thanks in advance.

[metawake]

Hi! Thanks for your interest in the library!
The full library uses spacy for intelligent compression - specifically for:
Named Entity Recognition: Preserving things like "Python 3.11", "OpenAI API", company names
Part-of-Speech tagging: Understanding which words are essential vs. modifiers
Dependency parsing: Keeping sentence structure intact, especially for questions
If you don't need those features and just want to remove common fillers like "please", "I think that", "very", etc. - here's a minimal, zero-dependency version that only uses Python's built-in re module:

`import re

def compress_prompt(text: str) -> str:
"""Compress a user prompt - zero dependencies, ~15-20% token reduction."""
if not text or not text.strip():
return ""

# Remove greetings
text = re.sub(r'^(Hi|Hello|Hey|Greetings),?\s*', '', text, flags=re.IGNORECASE)

# Remove fillers and hedging
text = re.sub(r'\b(I think that|in my opinion|definitely|certainly|probably|possibly)\b', '', text, flags=re.IGNORECASE)
text = re.sub(r'\b(Could you|Can you|I would like to|I want to|Please)\b', '', text, flags=re.IGNORECASE)

# Remove intensity modifiers
text = re.sub(r'\b(very|really|quite|extremely|just|actually|basically)\s+', '', text, flags=re.IGNORECASE)

# Collapse redundant phrases
text = re.sub(r'\bIn order to\b', 'to', text, flags=re.IGNORECASE)
text = re.sub(r'\bgoing to\b', 'will', text, flags=re.IGNORECASE)

# Clean up whitespace and punctuation
text = re.sub(r'\s+', ' ', text)
text = re.sub(r'\s*,\s*,', ',', text)
return text.strip()

`
for instance
compress_prompt("Hi, could you please help me understand how to very quickly learn Python?")
'help me understand how to quickly learn Python?'
Saved: 5 words (42%)

For most chat APIs, this covers ~80% of the compression benefit with zero dependencies. Just copy this function into your project and customize the patterns as needed.
Let me know if you need any help!

[Mohyoo]

Well, that's it? it looks too short.
if your library uses AI or any heavy processing, then well, I think this is enough for me since I wanna keep it lightweight.

Just as an addition, I found a library called stop_words while waiting for your response, and it looks very useful if you ever need it.
Thanks you!