chopratejas/headroom v0.3.7

on GitHub

What's New in v0.3.7

New Features

• any-llm backend — Route requests through 38+ LLM providers (OpenAI, Mistral, Groq, Ollama, etc.) via any-llm
  • Enable with --backend anyllm --anyllm-provider <provider>
  • Install with: pip install 'headroom-ai[anyllm]'
• IntelligentContextManager — Semantic-aware context management with multi-factor importance scoring: recency, semantic similarity, TOIN importance, error indicators, forward references, token density
• LLMLingua-2 Integration — Opt-in ML-based compression using Microsoft's LLMLingua-2 model with content-aware compression rates
• Code-Aware Compression — AST-based, syntax-preserving compression using tree-sitter for Python, JavaScript, TypeScript, Go, Rust, Java, C, C++
• ContentRouter — Intelligent compression orchestrator that auto-routes content to the optimal compressor based on type detection
• Custom Model Configuration — Support for Claude 4.5 (Opus), Claude 4 (Sonnet, Haiku), o3, o3-mini with pattern-based inference for unknown models
• OSS Evaluation Suite — Comprehensive benchmarks for accuracy, latency, and compression quality
• Universal JSON Crush — Improved JSON compression across all transforms

Highlights

• 6 compression algorithms: SmartCrusher, CacheAligner, ContentRouter, CodeCompressor, LLMLingua, IntelligentContext
• Production-ready proxy server with caching, rate limiting, and metrics
• Integrations: LangChain, Agno, MCP, Strands, AWS Bedrock
• 100+ LLM providers via LiteLLM + any-llm

Install

pip install headroom-ai[all]==0.3.7

Full Changelog: v0.3.0...v0.3.7