Scaling¶

Tune bridge in the plugin. Tune cache and sessions in Django settings. Defaults are safe for most apps.

Cheatsheet¶

rxconfig.py:

import os
import reflex as rx
from reflex_django.plugins import ReflexDjangoPlugin

config = rx.Config(
    app_name="shop",
    redis_url=os.environ["REDIS_URL"],
    plugins=[
        ReflexDjangoPlugin(config={
            "settings_module": "config.settings",
            "profile": "integrated",
            "bridge": {"mode": "smart"},
        }),
    ],
)

settings.py:

CACHES = {
    "reflex_events": {
        "BACKEND": "django.core.cache.backends.redis.RedisCache",
        "LOCATION": os.environ["REDIS_URL"],
    },
}
RX_EVENT_CACHE = "reflex_events"
RX_EVENT_CACHE_TTL = 60
RX_EVENT_CACHE_FAST_AUTH = False
RX_EVENT_METRICS = True
RX_PERFORMANCE_PRESET = "lean"
SESSION_ENGINE = "django.contrib.sessions.backends.cached_db"

Bridge tuning¶

Need	Setting
Skip middleware on UI-only state	`bridge.mode = "smart"` or `_rx_bridge = "none"` on one class
Full middleware (default)	`bridge.mode = "full"`
Reuse cached auth for auth-only events	`RX_EVENT_CACHE_FAST_AUTH = True`
Pick tier per state/event	`bridge.resolver` or `RX_EVENT_BRIDGE_RESOLVER`
Measure bridge overhead	`RX_EVENT_METRICS = True` or Devtools

class FilterState(rx.State):
    _rx_bridge = "none"

RX_EVENT_CACHE_FAST_AUTH is opt-in because it trades fewer session/auth middleware calls for a small TTL-bound staleness window. Logout invalidates the event cache.

RX_PERFORMANCE_PRESET = "lean" reduces bridge/UI churn by changing mirror and auth-sync defaults only when you have not explicitly overridden them. If you set RX_AUTH_AUTO_SYNC or mirror toggles yourself, your values are preserved.

Custom resolver:

def resolve_bridge_tier(handler_state_cls, event):
    if handler_state_cls.__name__.endswith("FilterState"):
        return "none"
    return "full"

Configure it:

ReflexDjangoPlugin(config={
    "settings_module": "config.settings",
    "bridge": {"resolver": "shop.performance.resolve_bridge_tier"},
})

Upload events are always raised to at least auth_only, even if a resolver returns none.

Event cache¶

Use a shared Django cache when multiple workers process events:

CACHES = {
    "reflex_events": {
        "BACKEND": "django.core.cache.backends.redis.RedisCache",
        "LOCATION": os.environ["REDIS_URL"],
    },
}
RX_EVENT_CACHE = "reflex_events"
RX_EVENT_CACHE_TTL = 60

Set RX_EVENT_CACHE_TTL = 0 to disable event caching. Keep the TTL short when RX_EVENT_CACHE_FAST_AUTH=True.

Metrics¶

RX_EVENT_METRICS=True logs bridge timing at DEBUG. Set RX_EVENT_METRICS_LOGGER = "myapp.performance" to send metrics to a custom logger. For local query counts and tier inspection, use Devtools.

Live updates¶

LiveListMixin uses a process-local broadcaster for model-signal fan-out. For multiple workers, add a shared transport such as Redis pub/sub or Postgres LISTEN/NOTIFY and publish into each worker with live_broadcaster().publish(...).

from reflex_django.live import ModelChange, live_broadcaster

live_broadcaster().publish(ModelChange("shop.product", "updated", 42))

See Live updates.

Production checklist¶

Build SPA in CI with reflex export
Set redis_url when running multiple Reflex workers
Use cache-backed sessions so Django and Reflex agree on login
Proxy WebSocket on /_event with long idle timeout
Proxy uploads on /_upload to the Reflex backend in split deployments
Review Security before production

See Deploy for the build pipeline.