AsyncInMemoryBackend

Async-native in-memory backend using asyncio.Lock. Use this in async frameworks (FastAPI, async LangGraph, async LLM middleware) to avoid blocking the event loop.

When to Use

Framework	Backend to use
FastAPI / Starlette	AsyncInMemoryBackend
Async LangGraph	AsyncInMemoryBackend
Standard Python scripts	InMemoryBackend (sync, faster)
Shared across processes	RedisBackend (async via redis.asyncio)

---

Usage

from omnicache_ai import AsyncInMemoryBackend

backend = AsyncInMemoryBackend(max_size=10_000)

await backend.set("key", b"value", ttl=60)
value = await backend.get("key")   # b"value"
await backend.delete("key")
exists = await backend.exists("key")  # False
await backend.clear()
await backend.close()

In an async context

import asyncio
from omnicache_ai import AsyncInMemoryBackend

async def main():
    backend = AsyncInMemoryBackend()
    await backend.set("result", b"cached-response", ttl=300)
    data = await backend.get("result")
    print(data)

asyncio.run(main())

---

Internals

AsyncInMemoryBackend wraps InMemoryBackend (LRU + per-entry TTL) behind an asyncio.Lock. All reads and writes acquire the lock before delegating to the synchronous implementation. This ensures correctness in concurrent async code without spawning threads.

---

API Reference

Parameter	Type	Default	Description
max_size	int	10_000	Max entries before LRU eviction

Implements AsyncCacheBackend protocol — all methods are async.