Tracing

The traced() decorator enables automatic tracing of function calls, inputs, and outputs. It is designed to help you monitor, debug, and audit your code by capturing detailed information about function executions, including arguments, return values, and exceptions.

You can view the traces of an Orchestrator job by going to the Jobs page, click a job, a side panel will open, and they will be available under the Trace tab. These can also be seen in UiPath Maestro when your agent is part of a larger process orchestration.

Usage

Apply the @traced() decorator to any function (sync, async, generator, or async generator) to automatically record its execution as a trace span.

from uipath.tracing import traced

@traced()
def my_function(x, y):
    return x + y

@traced(name="custom_span", run_type="my_type")
async def my_async_function(a, b):
    return a * b

Parameters

Parameter	Type	Description
name	Optional[str]	Custom name for the trace span. Defaults to the function name.
run_type	Optional[str]	Category for the run (e.g., "uipath"). Useful for filtering traces.
span_type	Optional[str]	Custom type for the span. Defaults to function type (sync/async/generator).
input_processor	Optional[Callable[[dict], dict]]	Function to process/transform inputs before recording. Receives a dict of arguments.
output_processor	Optional[Callable[[Any], Any]]	Function to process/transform outputs before recording. Receives the function's return value.
hide_input	bool	If True, input data is redacted in the trace for privacy/security.
hide_output	bool	If True, output data is redacted in the trace for privacy/security.

Input and Output Processors

Processors allow you to mask, redact, or transform sensitive data before it is recorded in the trace. For example:

def mask_inputs(inputs):
    inputs = inputs.copy()
    if 'password' in inputs:
        inputs['password'] = '***REDACTED***'
    return inputs

def anonymize_output(output):
    if isinstance(output, dict) and 'email' in output:
        output = output.copy()
        output['email'] = 'anonymous@example.com'
    return output

@traced(input_processor=mask_inputs, output_processor=anonymize_output)
def login(user, password):
    # ...
    return {"email": user + "@example.com"}

Privacy Controls

• Set hide_input=True to prevent input data from being logged.
• Set hide_output=True to prevent output data from being logged.

@traced(hide_input=True, hide_output=True)
def sensitive_operation(secret):
    ...

Supported Function Types

• Regular functions (sync/async)
• Generator functions (sync/async)

Example with plain python agents

When used with plain python agents please call wait_for_tracers() at the end of the script to ensure all traces are sent, if this is not called the agent could end without sending all the traces.

from uipath.tracing import traced, wait_for_tracers

@traced(name="process_payment", run_type="payment", hide_input=True)
def process_payment(card_number, amount):
    # Sensitive input will not be logged
    return {"status": "success", "amount": amount}

@traced()
def main():
    process_payment()

def main_wait_traces():
    try:
        main()
    finally:
        # this needs to be called after the last `traced` function is done
        # to ensure the trace associated with main is saved
        wait_for_tracers()

if __name__ == "__main__":
    main_wait_traces()

Example with langchain agents

When using uipath-langchain there is no need to call wait_for_tracers our framework will ensure that is called.

@traced()
def my_custom_traced_function(input: str) -> str:
    return { "x": "some-output" }

You can also use @traceable() attribute from langchain, but we recommend using @traced() attribute instead.

@traceable()
# @traced()  ---> do not use both at the same time or it will duplicate spans.
def my_custom_traced_function(input: str) -> str:
    return { "x": "some-output" }