CompactPrompt MCP server

CompactPrompt

CompactPrompt shortens the text you send to an AI model while preserving its meaning. The result costs less to run, returns faster, and is less likely to exceed the model's input limit. The common case is a single function call, and no background in machine learning is required to use it.

Background

An AI model reads an input — the prompt — and returns a response. Providers charge according to the amount of text processed, measured in tokens (each token is roughly three-quarters of a word), and every model has a maximum input size. A long prompt that combines instructions, documents, tables, and examples therefore costs more, responds more slowly, and may not fit at all.

CompactPrompt reduces the size of a prompt while retaining the information that matters, so you keep the substance and discard the overhead.

Getting started

Install the library:

pip install compactprompt

Shorten a prompt:

from compactprompt import CompactPrompt

result = CompactPrompt.compact(
    "Please could you very kindly go ahead and provide a really concise "
    "summary of the quarterly report."
)

print(result.text)
print(f"{result.ratio:.1f}x smaller "
      f"({result.tokens_before} -> {result.tokens_after} tokens)")

Output:

a really concise summary of the quarterly report.
1.7x smaller (22 -> 13 tokens)

The filler — "Please could you very kindly go ahead and" — is removed, and the meaning is unchanged.

What it does

CompactPrompt provides several methods for reducing the size of a prompt. They can be used individually or together. Each is described below in plain terms, followed by an optional technical note.

Trimming low-value wording

Removes words that carry little meaning, such as conversational filler, and keeps the words that do. This is lossy: the removed words are not recoverable, but the result is ready to use as it is.

from compactprompt import CompactPrompt

# Remove approximately 40% of the tokens
result = CompactPrompt.compact(prompt, ratio=0.4)

# Or target a specific size
result = CompactPrompt.compact(prompt, budget=64)

Technical detail

Each word receives an information score that combines how rare it is in general (static self-information) with how predictable it is in context (dynamic self-information from a small language model). Low-scoring words are removed. Whole grammatical phrases are removed together, using spaCy, so the result remains readable, and names and numbers are protected. This implements the fusion rule from the CompactPrompt paper.

Reversible shortening of repeated phrases

When a phrase recurs, it is replaced with a short placeholder, and a key records what each placeholder stands for. This is lossless: the exact original can be restored at any time.

import compactprompt as cp

doc = "operating cash flow rose. operating cash flow fell. operating cash flow held."
abbr = cp.abbreviate(doc, n=3)

print(abbr.text)        # '@0 rose. @0 fell. @0 held.'
print(abbr.dictionary)  # {'@0': 'operating cash flow'}
print(abbr.restore())   # the exact original

Retain abbr.dictionary so the placeholders can be expanded again later.

Shortened. The full README is on GitHub.