montevive/autocache v1.0.0

on GitHub

[1.0.0] - 2025-10-08

Added

Core Features

• Intelligent cache-control injection for Anthropic Claude API requests
• Automatic token counting and analysis for optimal cache breakpoint placement
• ROI analytics with detailed cost savings calculations
• Support for multiple cache strategies (conservative, moderate, aggressive)
• Streaming and non-streaming request support
• Response headers with detailed cache metadata (X-Autocache-* headers)

Cache Intelligence

• Smart breakpoint placement using ROI scoring algorithm
• Token minimums enforcement (1024 for most models, 2048 for Haiku)
• Automatic TTL assignment (1h for stable content, 5m for dynamic)
• Support for system prompts, tool definitions, and content blocks

API Endpoints

• POST /v1/messages - Main proxy endpoint with automatic caching
• GET /health - Health check endpoint
• GET /metrics - Metrics and configuration endpoint
• GET /savings - Comprehensive savings analytics and statistics

Configuration

• Environment variable configuration support
• Multiple caching strategies with customizable thresholds
• API key handling via headers or environment variables
• Configurable logging (text/JSON, multiple levels)

Docker Support

• Multi-stage Dockerfile with optimized build
• Docker Compose configuration with health checks
• Non-root user security
• Resource limits and restart policies

Documentation

• Comprehensive README with examples
• Architecture documentation (CLAUDE.md)
• API key handling guide
• n8n integration documentation
• Troubleshooting guides

Testing

• Unit tests for all core components
• Real API integration tests
• Test fixtures and utilities
• n8n workflow testing scripts

Technical Details

• Built with Go 1.23+
• Modular architecture with clean separation of concerns
• Logrus-based structured logging
• Graceful shutdown handling
• HTTPS support for Anthropic API communication