Implementation:Confident ai Deepeval TraceManager Configure

Overview

TraceManager Configure is the implementation method that configures the global trace collection behavior in DeepEval. Through the trace_manager.configure() method on the singleton TraceManager instance, developers control sampling rates, data masking functions, environment labels, client integrations, and the global tracing toggle.

API Documentation

Method: trace_manager.configure

Source: deepeval/tracing/tracing.py:L151-178

Import:

from deepeval.tracing import trace_manager

Singleton: trace_manager is a global TraceManager instance defined at deepeval/tracing/tracing.py:L806.

Signature:

trace_manager.configure(
    mask=None,
    environment=None,
    sampling_rate=None,
    confident_api_key=None,
    anthropic_client=None,
    openai_client=None,
    tracing_enabled=None,
) -> None

Parameters

Parameter	Type	Description
mask	Optional[Callable]	A PII redaction function applied to trace data before transmission. Receives data as input and returns the masked version.
environment	Optional[str]	Environment label for trace segregation (e.g., "production", "staging", "development").
sampling_rate	Optional[float]	A value between 0.0 and 1.0 controlling the fraction of traces collected. 0.1 means 10% of traces are collected.
confident_api_key	Optional[str]	Override the default Confident AI API key for trace submission.
anthropic_client	Optional	An Anthropic client instance for automatic instrumentation of Anthropic API calls.
openai_client	Optional	An OpenAI client instance for automatic instrumentation of OpenAI API calls.
tracing_enabled	Optional[bool]	Global toggle to enable or disable all tracing. When False, no traces are collected or transmitted.

Input / Output

• Inputs: Configuration parameters controlling trace collection behavior.
• Outputs: The global TraceManager singleton is configured, and all subsequent trace operations respect the new settings. Returns None.

Usage Example

from deepeval.tracing import trace_manager

def mask_pii(data):
    return data.replace("@", "[REDACTED]")

trace_manager.configure(
    environment="production",
    sampling_rate=0.1,
    mask=mask_pii,
    tracing_enabled=True,
)

Relationships

Principle:Confident_ai_Deepeval_Trace_Configuration

Related Pages

• Environment:Confident_ai_Deepeval_Python_3_9_Runtime
• Environment:Confident_ai_Deepeval_LLM_Provider_Credentials

Metadata

DeepEval Tracing Observability LLM_Evaluation 2026-02-14 09:00 GMT