Skip to main content
API Reference8 min readUpdated July 2026

Python API Reference

Complete function signatures, ContextCompressor class, and return types.

compress() Function Signature

The primary entry point for quick prompt context reduction:
llmslim/core.py
def compress(
    text: str,
    target_ratio: float = 0.5,
    mode: str = "auto",
    detect_content: bool = True,
    preserve_code: bool = True,
    preserve_entities: bool = True,
) -> CompressionResult:
    """Surgically compress input prompt text while preserving key instruction fidelity.
    
    Args:
        text: Input string (prompt, RAG context, or code).
        target_ratio: Target proportion of tokens to retain (e.g., 0.5 = 50% retained).
        mode: Compression mode ("auto", "markdown", "json", "xml", "code", "chat").
        detect_content: Auto-detect format type automatically.
        preserve_code: Shield syntax, functions, and formatting in code blocks.
        preserve_entities: Protect proper nouns, numbers, and technical terms.
        
    Returns:
        CompressionResult object with telemetry metrics and compressed text.
    """

Frequently Asked Questions

What object is returned by compress()?

A CompressionResult instance containing compressed_text, original_tokens, compressed_tokens, savings_percent, and detailed execution metrics.