Conditional Tracing - PandaProbe

PandaProbe supports fully disabling the SDK, making it safe to leave instrumentation in production code that you only want active in certain environments.

Disable via environment variable

export PANDAPROBE_ENABLED=false

When PANDAPROBE_ENABLED=false:

The SDK does not initialize a client
pandaprobe.get_client() returns None
pandaprobe.flush() and pandaprobe.shutdown() silently do nothing
@pandaprobe.trace and @pandaprobe.span decorators pass through transparently (zero overhead)
Wrappers return the original client unchanged
Integration adapters’ instrument() calls are no-ops

This makes it safe to leave PandaProbe instrumentation in all environments. The SDK adds zero overhead when disabled.

Disable via init()

import pandaprobe

pandaprobe.init(enabled=False)

Environment-based configuration

Common pattern using different configurations per environment:

import os
import pandaprobe

env = os.environ.get("APP_ENV", "development")

pandaprobe.init(
    api_key=os.environ.get("PANDAPROBE_API_KEY"),
    project_name="my-project",
    environment=env,
    enabled=env != "testing",
    debug=env == "development",
)

Conditional tracing without disabling

If you want the SDK enabled but want to conditionally trace specific operations:

import pandaprobe

def handle_request(query: str):
    if should_trace(query):
        with pandaprobe.start_trace("agent", input={"messages": [{"role": "user", "content": query}]}) as t:
            return process(query, t)
    else:
        return process(query, None)