> ## Documentation Index
> Fetch the complete documentation index at: https://docs.featherframework.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Caching

> Function and response caching with argument-keyed invalidation, backed by memory or Redis.

```bash .env theme={null}
CACHE_BACKEND=memory       # in-memory: single process, resets on restart
# or
CACHE_BACKEND=redis        # shared across processes, persistent
CACHE_URL=redis://localhost:6379/0
CACHE_DEFAULT_TTL=300      # seconds
```

## Caching function results

Results are cached by the function's arguments, so each distinct call gets its own entry.

```python theme={null}
from feather import cached

@cached(ttl=60)
def get_user_stats(user_id):
    return calculate_stats(user_id)

stats = get_user_stats(123)   # first call: executes
stats = get_user_stats(123)   # second call: cached

get_user_stats.invalidate(user_id=123)   # invalidate when data changes
```

## Caching route responses

```python theme={null}
from feather import cache_response

@api.get('/products')
@cache_response(ttl=300)
def list_products():
    return {'products': Product.query.all()}

# custom cache key from URL params
@api.get('/users/<user_id>')
@cache_response(ttl=60, key='user:{user_id}')
def get_user(user_id):
    return {'user': User.query.get(user_id)}

# skip the cache conditionally
@api.get('/dashboard')
@cache_response(ttl=300, unless=lambda: current_user.is_admin)
def dashboard():
    return {'stats': get_stats()}
```

<Warning>
  `cache_response` varies on the current user by default. On a genuinely public page
  pass `vary_on_user=False`, or every visitor gets their own copy of an identical
  response.
</Warning>

## Direct access

```python theme={null}
from feather import get_cache

cache = get_cache()
cache.set('key', {'data': 'value'}, ttl=60)
value = cache.get('key')      # None if expired or missing
cache.delete('key')
```

<Note>
  The `memory` backend is per-process. Under `gunicorn --workers 4` each worker keeps its
  own cache, so a hit rate looks worse than it is and invalidation reaches only one
  worker. Use `redis` whenever you run more than one process.
</Note>