PyPI page
Home page
Author:
None
License:
MIT
Summary:
Cut your LLM prompt token costs by 30-40% with deterministic, training-free lexical compression. Shrink prompts for OpenAI, Anthropic, and any LLM API while preserving output quality. Includes document-to-markdown reduction for PDFs and Word files, zone-aware compression that protects JSON schemas and output formats, async support, and a built-in 6-metric quality evaluator.
Latest version:
0.6.7
Required dependencies:
bert-score
|
fastapi
|
nltk
|
numpy
|
pymupdf
|
python-docx
|
python-multipart
|
rapidocr-onnxruntime
|
rouge-score
|
sentence-transformers
|
tiktoken
|
uvicorn
Optional dependencies:
build
|
pytest
|
twine
Downloads last day:
14
Downloads last week:
323
Downloads last month:
1,331