Module compression 

Smart Context Compression for Tool Outputs

Implements multi-layer semantic compression with RAG retrieval pattern:

1. Semantic Deduplication - Group identical patterns
2. Importance-Weighted Output - Critical=full, Low=counts
3. Hierarchical Summaries - Multi-level detail
4. RAG Pattern - Store full data, return summary with retrieval reference

Structs

CompressedOutput

Compressed output ready for LLM context

CompressionConfig

Configuration for compression behavior

DeduplicatedPattern

A deduplicated pattern representing multiple similar issues

SeveritySummary

Summary counts by severity level

Enums

Severity

Severity levels for importance-weighted filtering

Functions

compress_analysis_output

Compress analyze_project output specifically

compress_tool_output

Main compression function - compresses tool output and stores full data for retrieval