Reflex Process Management

This skill covers how to compile, run, and reload a Reflex application.

Compiling (Testing the App)

To verify the app compiles without errors, run:

reflex compile --dry

This checks for syntax errors, import issues, and component problems without starting the server. Use this as a quick validation step after making changes.

Running the Server

When instructed to run the Reflex server, always use production mode and redirect output to a log file:

reflex run --env prod --single-port 2>&1 | tee reflex.log

This command starts a long-running server process that does not support hot reload in production mode. Code changes will not be picked up automatically — you must stop and restart the server to apply changes (see Reloading a Running App below).

Using 2>&1 | tee reflex.log captures both stdout and stderr to reflex.log while still printing to the terminal.

Important: Always use --env prod unless the user explicitly requests development mode.

Reloading a Running App

To reload the app without manually stopping and restarting from the terminal, follow these steps:

Step 1: Determine the app port

Read reflex.log to find the port the app is listening on. Look for a line like App running at: http://0.0.0.0:<port>. Do not assume the port is 8000.

Step 2: Find the app process

Using the port from Step 1, locate the listening process (not browser connections):

lsof -i :<port> -sTCP:LISTEN -t

The -sTCP:LISTEN flag is critical — it filters to only the server process that is listening on the port, excluding browser or client connections. Without it, you may kill the user's browser.

If lsof is not available, use:

ss -tlnp | grep :<port>

Or:

fuser <port>/tcp

Step 3: Send an interrupt signal

Send SIGINT (equivalent to Ctrl+C) to gracefully stop the process:

kill -INT $(lsof -i :<port> -sTCP:LISTEN -t)

If the process doesn't stop, escalate to SIGTERM:

kill -TERM $(lsof -i :<port> -sTCP:LISTEN -t)

Step 4: Restart the server

Once the old process has exited, truncate the old log and start the server again:

> reflex.log
reflex run --env prod --single-port 2>&1 | tee reflex.log

Investigating Errors

When the user reports an error, read reflex.log to find and diagnose the issue.