You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There is no reason for any of the values to be optional, since those are all counters. Rather than None, they should be zeros. This change will make tracking usage require much less client code.
Unnecessary optionals:
CompletionUsage.completion_tokens_details
CompletionUsage.prompt_tokens_details
PromptTokensDetails.audio_tokens
PromptTokensDetails.cached_tokens
CompletionTokensDetails.audio_tokens
CompletionTokensDetails.reasoning_tokens
Additional context
Currently, you need to write something like this for type checking to work (I don't think Mypy will work with code catching AttributeError):
reasoning_tokens = (left.completion_tokens_details and left.completion_tokens_details.reasoning_tokens or 0) + (right.completion_tokens_details and right.completion_tokens_details.reasoning_tokens or 0)
This commit addresses the issue outlined in openai#1837 by replacing `None` with `0` for various token fields in the `CompletionTokensDetails` and `PromptTokensDetails` classes. These fields are intended to represent counters, so initializing them with `0` simplifies the code and eliminates unnecessary optional checks.
**Changes made:**
- Replaced `None` with `0` for fields like `audio_tokens`, `reasoning_tokens`, and `cached_tokens` in both `CompletionTokensDetails` and `PromptTokensDetails` classes.
- This update makes the token fields more consistent and eliminates the need for optional handling, as they will always default to `0` if not specified.
**Why `Optional` was not removed:**
- The `Optional` typing has been retained for some fields (like `audio_tokens` and `reasoning_tokens`) because, in some cases, the presence of these tokens may depend on certain conditions. By leaving them as `Optional`, we ensure flexibility in cases where the tokens might not be relevant or provided (e.g., for specific models or types of requests).
This update improves usability by simplifying the tracking of token usage, avoiding unnecessary checks, and ensuring default values of `0` for all relevant counters.
I’ve just opened a PR #2030 to address the request to replace None with 0 for token counters in the CompletionUsage class. The main change ensures that the token fields default to 0, making it easier to track usage and reducing the need for additional checks.
I’d love to hear your feedback on the PR and whether you think any further adjustments are needed. Let me know if there’s anything else you'd like me to change!
Confirm this is a feature request for the Python library and not the underlying OpenAI API.
Describe the feature or improvement you're requesting
https://github.com/openai/openai-python/blob/891e1c17b7fecbae34d1915ba90c15ddece807f9/src/openai/types/completion_usage.py#L27C1-L40C50
There is no reason for any of the values to be optional, since those are all counters. Rather than None, they should be zeros. This change will make tracking usage require much less client code.
Unnecessary optionals:
CompletionUsage.completion_tokens_details
CompletionUsage.prompt_tokens_details
PromptTokensDetails.audio_tokens
PromptTokensDetails.cached_tokens
CompletionTokensDetails.audio_tokens
CompletionTokensDetails.reasoning_tokens
Additional context
Currently, you need to write something like this for type checking to work (I don't think Mypy will work with code catching
AttributeError
):It should be possible and correct to just have:
The text was updated successfully, but these errors were encountered: