aci-cache 0.1.0

Adaptive cache invalidation for Redis. Drop-in wrapper, zero config.

aci-cache

Adaptive cache invalidation for Redis. Drop-in wrapper, zero config.

aci-cache wraps your existing Redis client and adds an intelligent invalidation layer that automatically adjusts its strategy based on real-time write patterns:

• Low write rate → TTL-based expiry (minimal overhead)
• Medium write rate → Batched invalidation (balanced efficiency)
• High write rate → Eager per-write invalidation (maximum consistency)

Quick Start

pip install aci-cache

import redis
from aci_cache import AdaptiveCache

# Wrap your existing Redis client — that's it
r = redis.Redis(host="localhost")
cache = AdaptiveCache(r)

# Use it just like Redis
cache.set("user:123", "alice")
user = cache.get("user:123")

# Check what strategy is active
print(cache.strategy)  # "ttl", "batched", or "eager"

How It Works

The library monitors your write rate in real time using a sliding window. Based on configurable thresholds, it automatically switches between three invalidation strategies:

Strategy	When Active	Behavior
TTL	< 10 writes/sec	Standard Redis TTL. Zero overhead.
Batched	10–50 writes/sec	Buffers keys, flushes every 2.5s via Pub/Sub.
Eager	> 50 writes/sec	Publishes invalidation per write via Pub/Sub.

Multi-Instance Support

When running multiple app instances behind a load balancer, aci-cache coordinates cache invalidation across all instances via Redis Pub/Sub:

# Instance 1 (controller)
cache = AdaptiveCache(redis.Redis(), is_controller=True)

# Instance 2-N (followers)
cache = AdaptiveCache(redis.Redis(), is_controller=False)

Configuration

All settings have sensible defaults. Override via constructor:

cache = AdaptiveCache(
    redis.Redis(),
    high_threshold=100,    # writes/sec → eager
    low_threshold=20,      # writes/sec → ttl
    default_ttl=30,        # seconds
    batch_interval=5.0,    # seconds between batch flushes
    namespace="my-service", # key namespace for multi-service Redis
)

License

MIT