Configuration Specification

Environment variables, privacy controls, and data masking

In some situations, you may need to modify the observability level of your tracing. For instance, you may want to keep sensitive information from being logged for security reasons, or you may want to limit the size of the base64 encoded images logged to reduce payload size.

Environment Variables

The OpenInference Specification defines a set of environment variables you can configure to suit your observability needs:

Environment Variable Name	Effect	Type	Default
OPENINFERENCE_HIDE_LLM_INVOCATION_PARAMETERS	Hides LLM invocation parameters (independent of input/output hiding)	bool	False
OPENINFERENCE_HIDE_INPUTS	Hides input.value and all input messages (input messages are hidden if either HIDE_INPUTS OR HIDE_INPUT_MESSAGES is true)	bool	False
OPENINFERENCE_HIDE_OUTPUTS	Hides output.value and all output messages (output messages are hidden if either HIDE_OUTPUTS OR HIDE_OUTPUT_MESSAGES is true)	bool	False
OPENINFERENCE_HIDE_INPUT_MESSAGES	Hides all input messages (independent of HIDE_INPUTS)	bool	False
OPENINFERENCE_HIDE_OUTPUT_MESSAGES	Hides all output messages (independent of HIDE_OUTPUTS)	bool	False
OPENINFERENCE_HIDE_INPUT_IMAGES	Hides images from input messages (only applies when input messages are not already hidden)	bool	False
OPENINFERENCE_HIDE_INPUT_TEXT	Hides text from input messages (only applies when input messages are not already hidden)	bool	False
OPENINFERENCE_HIDE_PROMPTS	Hides LLM prompts (completions API)	bool	False
OPENINFERENCE_HIDE_OUTPUT_TEXT	Hides text from output messages (only applies when output messages are not already hidden)	bool	False
OPENINFERENCE_HIDE_CHOICES	Hides LLM choices (completions API outputs)	bool	False
OPENINFERENCE_HIDE_EMBEDDING_VECTORS	Deprecated: use OPENINFERENCE_HIDE_EMBEDDINGS_VECTORS	bool	False
OPENINFERENCE_HIDE_EMBEDDINGS_VECTORS	Replaces embedding.embeddings.*.embedding.vector values with "__REDACTED__"	bool	False
OPENINFERENCE_HIDE_EMBEDDINGS_TEXT	Replaces embedding.embeddings.*.embedding.text values with "__REDACTED__"	bool	False
OPENINFERENCE_BASE64_IMAGE_MAX_LENGTH	Limits characters of a base64 encoding of an image	int	32,000

Redacted Content

When content is hidden due to privacy configuration settings, the value "__REDACTED__" is used as a placeholder. This constant value allows consumers of the trace data to identify that content was intentionally hidden rather than missing or empty.

Usage

To set up this configuration you can either:

• Set environment variables as specified above
• Define the configuration in code as shown below
• Do nothing and fall back to the default values
• Use a combination of the three, the order of precedence is:
  • Values set in the TraceConfig in code
  • Environment variables
  • Default values

Python Configuration

If you are working in Python, and want to set up a configuration different than the default you can define the configuration in code as shown below, passing it to the instrument() method of your instrumentator:

from openinference.instrumentation import TraceConfig

config = TraceConfig(
    hide_llm_invocation_parameters=False,
    hide_inputs=False,
    hide_outputs=False,
    hide_input_messages=False,
    hide_output_messages=False,
    hide_input_images=False,
    hide_input_text=False,
    hide_output_text=False,
    hide_embeddings_vectors=False,
    hide_embeddings_text=False,
    base64_image_max_length=32000,
    hide_prompts=False,  # Hides LLM prompts (completions API)
    hide_choices=False,  # Hides LLM choices (completions API outputs)
)

from openinference.instrumentation.openai import OpenAIInstrumentor

OpenAIInstrumentor().instrument(
    tracer_provider=tracer_provider,
    config=config,
)

JavaScript Configuration

If you are working in JavaScript, and want to set up a configuration different than the default you can define the configuration as shown below and pass it into any OpenInference instrumentation:

import { OpenAIInstrumentation } from "@arizeai/openinference-instrumentation-openai"

/**
 * Everything left out of here will fallback to
 * environment variables then defaults
 */
const traceConfig = { 
  hideInputs: true,
  hideOutputs: false,
  hideInputMessages: false,
  hideOutputMessages: false,
  hideInputImages: false,
  hideInputText: false,
  hideOutputText: false,
  hideEmbeddingsVectors: false,
  hideEmbeddingsText: false,
  base64ImageMaxLength: 32000
}

const instrumentation = new OpenAIInstrumentation({ traceConfig })

Privacy Best Practices

When working with sensitive data, always configure privacy controls before instrumenting your application to prevent accidental exposure of personal information.

Common Privacy Scenarios

Hide all user inputs and outputs:

config = TraceConfig(
    hide_inputs=True,
    hide_outputs=True,
)

Hide only message content, keep metadata:

config = TraceConfig(
    hide_input_messages=True,
    hide_output_messages=True,
)

Hide embeddings but keep model parameters:

config = TraceConfig(
    hide_embeddings_vectors=True,
    hide_embeddings_text=True,
)

Limit image data size:

config = TraceConfig(
    base64_image_max_length=10000,  # Smaller limit for bandwidth
)