Fix pydantic serialization warnings #16299
Open
Conversation
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
Three tests that demonstrate the problems caused by the 'del self.field' anti-pattern in Message and Choices classes: 1. test_pydantic_serialization_warnings.py - Shows that serialization generates Pydantic warnings due to deleted fields 2. test_global_filter_pollution.py - Shows that the global warning filter added in 08b2b4f suppresses warnings from user code (action-at-a-distance) 3. test_warnings_leak_in_strict_mode.py - Shows that warnings still leak into pytest output despite the global filter All three tests currently FAIL, demonstrating the bugs exist. They will PASS after applying the fix (Field(exclude=True) instead of del).
Replace the 'del self.field' anti-pattern with Field(exclude=True) in Message and Choices classes. This is the canonical Pydantic approach for excluding fields from serialization. Changes: - Message class: Use Field(exclude=True) for audio, images, annotations, reasoning_content, and thinking_blocks - Choices class: Use Field(exclude=True) for logprobs and provider_specific_fields - Remove all 'del self.field' statements from __init__ methods - Remove global warning filter from transformation.py This eliminates Pydantic serialization warnings at their source without global side effects.
|
@pmorch is attempting to deploy a commit to the CLERKIEAI Team on Vercel. A member of the Team first needs to authorize it. |
|
|
Author
|
About the github action linting errors: This PR only modified:
The mypy errors are in completely different files:
These are all pre-existing mypy errors in the codebase, unrelated to these changes. |
Author
|
Note: This PR does not address the exception handler broadness discussed in "Note on Exception Handling" in #11759 (comment) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Title
Fix pydantic serialization warnings
Relevant issues
Fixes #11759
Pre-Submission checklist
Please complete all items before asking a LiteLLM maintainer to review your PR
tests/litellm/directory, Adding at least 1 test is a hard requirement - see detailsmake test-unitPre-existing test failures:
4 Pre-existing test failures (unrelated to this PR):
All 4 failures are due to missing optional dependencies or missing model cost data, not related to this Pydantic serialization fix.
Type
🐛 Bug Fix
✅ Test
Changes
This PR fixes Pydantic serialization warnings (issue #11759) by using
Field(exclude=True)instead ofdel self.fieldin LiteLLM'sMessageandChoicesclasses, and removes the global warning filter that was masking the issue.The Problem: LiteLLM deletes Pydantic model fields after initialization, which breaks Pydantic's field tracking and generates warnings. A global warning filter was added to suppress these warnings, but it affects all code in the process—including user code.
The Fix:
Field(exclude=True)to fields inlitellm/types/utils.pyand removeddelstatementslitellm/llms/meta_llama/chat/transformation.pySee #11759 (comment) for full context