Implement OpenAI Responses API instrumentation and examples#4166
Closed
vasantteja wants to merge 29 commits intoopen-telemetry:mainfrom
Closed
Implement OpenAI Responses API instrumentation and examples#4166vasantteja wants to merge 29 commits intoopen-telemetry:mainfrom
vasantteja wants to merge 29 commits intoopen-telemetry:mainfrom
Conversation
- Added instrumentation for the OpenAI Responses API, including tracing for `Responses.create` and `Responses.stream` methods. - Introduced example scripts demonstrating the usage of the Responses API with OpenTelemetry. - Created a `.env` file for configuration, including API keys and OpenTelemetry settings. - Updated README files to include instructions for running examples and configuring the environment. - Added unit tests for the new Responses API functionality, ensuring proper tracing and metrics collection. This update enhances the observability of OpenAI API interactions within the OpenTelemetry framework.
- Updated the OpenAIInstrumentor to conditionally wrap and unwrap the Responses API methods based on the installed OpenAI package version (>=1.66.0). - Added version checks in the test suite to skip tests if the Responses API is not available, ensuring compatibility with earlier versions of the OpenAI library. - Improved error handling for missing API methods to prevent runtime exceptions.
- Added pylint disable comments to suppress warnings for specific lines in the Responses API example and patch files. - Updated the `responses_create` and `responses_stream` methods with links to relevant OpenAI documentation for better reference. - Improved code formatting for readability by adjusting line breaks and indentation in the patch file.
- Reformatted code in the patch file to enhance readability by adjusting line breaks and indentation. - Ensured consistent style for model retrieval and span name updates in the ResponseStreamWrapper class. - Minor adjustments to import statements for clarity and organization.
- Updated the OpenAI package version in requirements.txt to 1.66.0 for compatibility. - Refactored span name retrieval in the patch.py file to directly format the span name using operation and model attributes, removing the redundant _get_span_name function. - Improved code clarity and consistency in the responses_create and responses_stream methods.
…omments - Added a comment in the `responses_stream` method to clarify the purpose of avoiding duplicate span creation. - Updated span name retrieval to use a default value of 'unknown' for the model attribute if not present, improving robustness. - Refactored the `_record_metrics` function to directly access the request model from attributes, enhancing clarity and consistency.
- Added a new wrapper function `responses_retrieve` to trace the `retrieve` method of the `Responses` class, enhancing observability. - Updated the `OpenAIInstrumentor` to include the new tracing functionality for the `retrieve` method. - Enhanced test coverage for the `retrieve` method, including new test cases for both standard and streaming responses. - Added new YAML cassettes to support the updated tests for the `retrieve` functionality.
- Added a TODO comment in the patch.py file to consider migrating Responses instrumentation to TelemetryHandler once content capture and streaming hooks are available. - Included a reference link to the OpenAI responses.py file for context on the `retrieve` method.
- Introduced a new module `patch_responses.py` to handle tracing for the `Responses` class methods, including `create`, `stream`, and `retrieve`. - Updated the `__init__.py` file to import the new responses patching functions. - Enhanced test coverage with new YAML cassettes for various response scenarios, including standard and streaming responses. - Removed outdated response tracing logic from `patch.py` to streamline the instrumentation process.
…strumentation - Moved the `_record_metrics` function to the `utils.py` file for better organization and accessibility. - Updated the `patch.py` file to import the `_record_metrics` function from `utils.py`, streamlining the code structure. - Enhanced the `responses_retrieve` method to simplify span attribute checks and improve readability. - Added new test cases and YAML cassettes to cover various response scenarios, including streaming and standard responses.
|
This is a relatively large PR, is there any way it could be split up? |
Contributor
Author
|
@JWinermaSplunk Sure I removed the async stuff to make it small. I will remove stream and retrieve to make it more small. FYI I am rewriting this with TelemetryHandler so that we can start using shared utils. I am assuming that might make the pr little bulky so can I remove the examples? |
…ry support - Added `opentelemetry-util-genai` as a dependency for improved telemetry handling. - Refactored response handling in `patch_responses.py` to utilize `TelemetryHandler` for tracing. - Updated `responses_create` and `responses_retrieve` methods to integrate new telemetry features. - Simplified imports and removed unused code in `__init__.py`. - Added extensive test cases and YAML cassettes for various response scenarios, including streaming and standard responses. - Adjusted requirements files to include the new utility package for testing.
…try features - Updated `responses_create` and `responses_retrieve` methods to streamline content capture logic. - Introduced helper functions for extracting input and output messages, and system instructions. - Enhanced telemetry support by integrating content capture based on experimental mode settings. - Added new test cases to validate content capture functionality in various scenarios, including streaming and standard responses. - Created YAML cassettes for testing response handling and content capture behavior. - Simplified imports and removed unused code in `__init__.py` and `patch_responses.py`.
wrisa
reviewed
Feb 12, 2026
…prove optional feature handling.
lmolkova
reviewed
Feb 20, 2026
| invocation, | ||
| result, | ||
| "service_tier", | ||
| GenAIAttributes.GEN_AI_OPENAI_RESPONSE_SERVICE_TIER, |
Member
There was a problem hiding this comment.
this is a deprecated attribute, use OpenAIattributes.OPENAI_RESPONSE_SERVICE_TIER
Member
|
@vasantteja it seems we have some intersection with this PR - #3715 (switching to new semconv for chat completion), would you mind reviewing it? We'll probably have some merge conflicts, but not terrible ones |
Member
MikeGoldsmith
left a comment
MikeGoldsmith
left a comment
There was a problem hiding this comment.
Generally looks pretty good - left some suggestions and would like to reduce size of this PR if possible.
This is another 3000+ lines change PR across many files - PRs of this size are hard to review with confidence. I know this is experimental and wanting to move fast, but huge PRs makes it challenging.
In this case, it's mostly VCR cassettes but I think we could break up into the following parts:
util-genaionly — egshould_capture_content()plus tests- Responses API core (non-streaming)
- Responses API streaming
As pointed out by @lmolkova - this PR is touching a lot of files #3715 does too. eg both modify utils.py, __init__.py, and patch.py.
Maybe we should we focus on one first, then rebase the other one to minimise conflicts?
Other things I noticed:
- pyproject.toml has
openai >= 1.26.0but Responses API requiresopenai >= 1.66.0 - can we separate some of the changes into other files to avoid having to disable pylint's too many lines? eg something like
patch_chat.py
…e response handling in ResponseStreamWrapper with safe instrumentation methods for error logging.
…e request model extraction.
11 tasks
6 tasks
github-project-automation
Bot
moved this from Reviewed PRs that need fixes
to Done
in Python PR digest
Member
|
@vasantteja looks like you may have some lint errors. You can run the precommit to clean them up. |
tammy-baylis-swi
moved this from Done
to Approved PRs that need fixes
in Python PR digest
github-project-automation
Bot
moved this from Approved PRs that need fixes
to Done
in Python PR digest
Member
|
@vasantteja please can you merge main into this branch and resolve conflicts, then run precommit as there are lint errors. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
14 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
This PR adds OpenAI Responses API instrumentation (sync Responses.create) to opentelemetry-instrumentation-openai-v2, using TelemetryHandler. It also adds tests for Responses API create behavior.
Fixes #3436 partly
Type of change
Please delete options that are not relevant.
How Has This Been Tested?
Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration
Does This PR Require a Core Repo Change?
Checklist:
See contributing.md for styleguide, changelog guidelines, and more.