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

# Conditional Tracing

> Enable or disable tracing based on environment or runtime conditions

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

```bash theme={null}
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

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

### Disable via init()

```python theme={null}
import pandaprobe

pandaprobe.init(enabled=False)
```

### Environment-based configuration

Common pattern using different configurations per environment:

```python theme={null}
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:

```python theme={null}
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)
```