Redis Backend¶

LongCache can persist responses in Redis. In v0.1.1, use one LongCache process per Redis-backed cache: each process owns a local FAISS index, and Redis does not synchronize indexes between workers.

Installation¶

Ensure the redis extra is installed:

uv pip install "longcache[redis]"
# or
pip install "longcache[redis]"

Synchronous Redis¶

from longcache import LongCache, CacheConfig

config = CacheConfig(
    backend="redis",
    redis_url="redis://:secret@redis-cluster.internal:6379/0",
    namespace="production_rag",
    ttl_seconds=7200,
)

cache = LongCache(config)


@cache.cached
def ask(query: str) -> str:
    return call_model(query)

Async support¶

LongCache(CacheConfig(backend="async_redis")) raises a clear error in v0.1.1 because the async backend is not connected to the complete cache/index lifecycle. AsyncRedisCacheBackend remains available for standalone experimental use. Use backend="redis" for the supported public cache path.

Close cache instances with cache.close() or a with LongCache(config) as cache: block to stop background maintenance and release backend resources.

Local Development with Docker¶

Run Redis locally with Docker Compose:

docker compose up -d redis
LONGCACHE_TEST_REDIS_URL=redis://localhost:6379/0 uv run --no-sync pytest tests/test_redis_backend.py -q