OpenAI

This integration connects Sentry with the OpenAI Python SDK. The integration has been confirmed to work with OpenAI 1.13.3.

Once you've installed this SDK, you can use Sentry LLM Monitoring, a Sentry dashboard that helps you understand what's going on with your AI pipelines.

Sentry LLM Monitoring will automatically collect information about prompts, tokens, and models from providers like OpenAI. Learn more about it here.

Install

Install sentry-sdk from PyPI with the openai extra:

pip install --upgrade 'sentry-sdk[openai]'

Configure

If you have the openai package in your dependencies, the OpenAI integration will be enabled automatically when you initialize the Sentry SDK.

An additional dependency, tiktoken, is required if you want to calculate token usage for streaming chat responses.

import sentry_sdk

sentry_sdk.init(
    dsn="https://examplePublicKey@o0.ingest.sentry.io/0",
    # Set traces_sample_rate to 1.0 to capture 100%
    # of transactions for tracing.
    traces_sample_rate=1.0,
    # Set profiles_sample_rate to 1.0 to profile 100%
    # of sampled transactions.
    # We recommend adjusting this value in production.
    profiles_sample_rate=1.0,
)

Verify

Verify that the integration works by creating an AI pipeline. The resulting data should show up in your LLM monitoring dashboard.

import sentry_sdk
from sentry_sdk.ai.monitoring import ai_track
from openai import OpenAI

sentry_sdk.init(...)  # same as above

client = OpenAI(api_key="(your OpenAI key)")

@ai_track("My AI pipeline")
def my_pipeline():
    with sentry_sdk.start_transaction(op="ai-inference", name="The result of the AI inference"):
      print(
          client.chat.completions.create(
              model="gpt-3.5", messages=[{"role": "system", "content": "say hello"}]
          )
          .choices[0]
          .message.content
      )

After running this script, a pipeline will be created in the LLM Monitoring section of the Sentry dashboard. The pipeline will have an associated OpenAI span for the chat.completions.create operation.

It may take a couple of moments for the data to appear in sentry.io.

Behavior

• The OpenAI integration will connect Sentry with all supported OpenAI methods automatically.

• All exceptions leading to an OpenAIException are reported.

• The supported modules are currently chat.completions.create and embeddings.create.

• Sentry considers LLM and tokenizer inputs/outputs as PII and doesn't include PII data by default. If you want to include the data, set send_default_pii=True in the sentry_sdk.init() call. To explicitly exclude prompts and outputs despite send_default_pii=True, configure the integration with include_prompts=False as shown in the Options section below.

Options

By adding OpenAIIntegration to your sentry_sdk.init() call explicitly, you can set options for OpenAIIntegration to change its behavior:

import sentry_sdk
from sentry_sdk.integrations.openai import OpenAIIntegration

sentry_sdk.init(
    # ...
    send_default_pii=True,
    integrations = [
        OpenAIIntegration(
            include_prompts=False,  # LLM/tokenizer inputs/outputs will be not sent to Sentry, despite send_default_pii=True
            tiktoken_encoding_name="cl100k_base",
        ),
    ],
)

You can pass the following keyword arguments to OpenAIIntegration():

• include_prompts:

  Whether LLM and tokenizer inputs and outputs should be sent to Sentry. Sentry considers this data personal identifiable data (PII) by default. If you want to include the data, set send_default_pii=True in the sentry_sdk.init() call. To explicitly exclude prompts and outputs despite send_default_pii=True, configure the integration with include_prompts=False.

  The default is True.

• tiktoken_encoding_name:

  If you want to calculate token usage for streaming chat responses you need to have an additional dependency, tiktoken installed and specify the tiktoken_encoding_name that you use for tokenization. See the OpenAI Cookbook for possible values.

  The default is None.

Supported Versions

• OpenAI: 1.0+
• tiktoken: 0.6.0+
• Python: 3.9+