Overview
Compresr reduces LLM token costs by up to 90% through intelligent context compression.
Compression Types
Compresr offers three types of compression to fit different use cases:
1. Question-Agnostic Compression
General-purpose compression that preserves semantic meaning without requiring a specific question. Ideal for compressing system prompts, documentation, and static contexts.
Via SDK: CompressionClient
2. Question-Specific Compression
Compresses context based on a specific question, keeping only information relevant to that question. Perfect for RAG systems, Q&A applications, and search results.
Via SDK: QSCompressionClient
3. Agentic Compression (Gateway)
Transparent proxy service that automatically compresses context for AI agents and LLM applications. Drop-in replacement for OpenAI/Anthropic endpoints with automatic compression.
Via Context Gateway (proxy)
Quick Start
- Get your API Key
- Install the SDK:
pip install compresr - Start compressing - see SDK docs or Gateway docs
Integration Options
pip install compresr - Full control with question-agnostic and question-specific compressionX-API-Key header