Mask Span Attributes

The OpenInference Specification defines a set of environment variables you can configure to suit your observability needs. In addition, the OpenInference auto-instrumentors accept a trace config which allows you to set these value in code without having to set environment variables, if that’s what you prefer The possible settings are:

Environment Variable Name	Effect	Type	Default
OPENINFERENCE_HIDE_INPUTS	Hides input value, all input messages & embedding input text	bool	False
OPENINFERENCE_HIDE_OUTPUTS	Hides output value & all output messages	bool	False
OPENINFERENCE_HIDE_INPUT_MESSAGES	Hides all input messages & embedding input text	bool	False
OPENINFERENCE_HIDE_OUTPUT_MESSAGES	Hides all output messages	bool	False
OPENINFERENCE_HIDE_INPUT_IMAGES	Hides images from input messages	bool	False
OPENINFERENCE_HIDE_INPUT_TEXT	Hides text from input messages & input embeddings	bool	False
OPENINFERENCE_HIDE_OUTPUT_TEXT	Hides text from output messages	bool	False
OPENINFERENCE_HIDE_EMBEDDING_VECTORS	Hides returned embedding vectors	bool	False
OPENINFERENCE_HIDE_LLM_INVOCATION_PARAMETERS	Hides LLM invocation parameters	bool	False
OPENINFERENCE_HIDE_LLM_PROMPTS	Hides LLM prompts span attributes	bool	False
OPENINFERENCE_HIDE_LLM_TOOLS	Hides tool definitions advertised to the LLM (`llm.tools.*`)	bool	False
OPENINFERENCE_BASE64_IMAGE_MAX_LENGTH	Limits characters of a base64 encoding of an image	int	32,000

Note: OPENINFERENCE_HIDE_LLM_TOOLS is also applied when OPENINFERENCE_HIDE_INPUTS is enabled, consistent with how OPENINFERENCE_HIDE_INPUT_MESSAGES and OPENINFERENCE_HIDE_LLM_PROMPTS behave. 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

Below is an example of how to set these values in code using our OpenAI Python and JavaScript instrumentors, however, the config is respected by all of our auto-instrumentors.

Python
TypeScript
Go

from openinference.instrumentation import TraceConfig
config = TraceConfig(
    hide_inputs=...,
    hide_outputs=...,
    hide_input_messages=...,
    hide_output_messages=...,
    hide_input_images=...,
    hide_input_text=...,
    hide_output_text=...,
    hide_embedding_vectors=...,
    hide_llm_invocation_parameters=...,
    hide_llm_prompts=...,
    hide_llm_tools=...,
    base64_image_max_length=...,
)

from openinference.instrumentation.openai import OpenAIInstrumentor
OpenAIInstrumentor().instrument(
    tracer_provider=tracer_provider,
    config=config,
)

/**
* Everything left out of here will fallback to
* environment variables then defaults
*/
const traceConfig = {
  hideInputs: true,
  hideLLMTools: true,
}

const instrumentation = new OpenAIInstrumentation({ traceConfig })

Use instrumentation.TraceConfig and pass it to the provider instrumentor via WithTraceConfig. Setting a value in code fully replaces the env-derived config for that instrumentor — matching the Python and JS precedence rules.

import (
    "github.com/openai/openai-go"
    "github.com/openai/openai-go/option"
    "go.opentelemetry.io/otel"

    "github.com/Arize-ai/openinference/go/openinference-instrumentation"
    openaiotel "github.com/Arize-ai/openinference/go/openinference-instrumentation-openai-go"
)

client := openai.NewClient(
    option.WithAPIKey(apiKey),
    option.WithMiddleware(openaiotel.Middleware(
        otel.Tracer("my-app"),
        openaiotel.WithTraceConfig(instrumentation.TraceConfig{
            HideInputs: true,
        }),
    )),
)

The Anthropic Go instrumentor uses the same shape:

option.WithMiddleware(anthropicotel.Middleware(otel.Tracer(...), anthropicotel.WithTraceConfig(instrumentation.TraceConfig{...})))

. See OpenAI Go SDK and Anthropic SDK Go for full setup.

Get Started

Tracing

Evaluation

Datasets & Experiments

Prompts

Settings

Concepts

Resources

Mask Span Attributes