Keywords4CV Caching Framework

A high-performance caching engine for Keywords4CV and beyond

A high-performance, thread-safe, and extensible caching engine originally developed for Keywords4CV, but designed for general use in Python applications.

Features

• Flexible Backends: Supports multiple caching backends:
  • In-Memory: Uses cachetools for fast, thread-safe in-memory caching (LRU, TTL).
  • Disk: Uses diskcache for persistent, on-disk caching.
  • Hybrid: Combines in-memory and disk caching for optimal performance.
  • Custom: Easily extend with your own backends.
• Serialization Options: Supports pickle (default) and msgpack (optional, for improved performance) for serializing cached objects.
• Thread and Process Safety: Uses appropriate locking mechanisms to ensure safe concurrent access from multiple threads and processes.
• Memory Management: Includes a MemoryManager to monitor system memory usage and adaptively trim caches to prevent out-of-memory errors.
• Extensible Design: Provides specialized cache managers for vectors (with optional GPU acceleration using torch) and text processing.
• Dynamic Configuration: Uses pydantic for easy configuration and runtime updates.
• Security: Supports optional encryption of cached data using cryptography.
• Metrics and Monitoring: Built-in metrics tracking (hits, misses, etc.) for integration with monitoring systems (e.g., Prometheus).
• Circuit Breaker: Includes a circuit breaker to handle failures gracefully when using external resources (e.g., API calls for synonyms).
• Comprehensive Testing: Includes a thorough test suite with unit tests and performance benchmarks.
• Well-Documented: Provides clear API documentation and usage examples.

Installation

Using pip:

pip install k4cv-caching-framework