fix: handle OpenAI.CreateEmbeddingResponse.Usage instead of only Open…

[selbyk]

Problem

OpenAI embeddings.create not reporting prompt_tokens because the embeddings Usage type is different from CompletionsUsage

Changes

handle OpenAI.CreateEmbeddingResponse.Usage instead of only OpenAI.CompletionsUsage

Release info Sub-libraries affected

Bump level

• Major
• Minor
• Patch

Libraries affected

• All of them?
• langfuse
• langfuse-node

Changelog notes

• Fixed token usage reporting for embeddings

---

Important

Fixes token usage reporting for OpenAI embeddings by handling OpenAI.CreateEmbeddingResponse.Usage in parseUsage and wrapMethod.

• Behavior:
  • Fixes token usage reporting for OpenAI embeddings by handling OpenAI.CreateEmbeddingResponse.Usage in parseUsage in parseOpenAI.ts.
  • Updates wrapMethod in traceMethod.ts to support OpenAI.CreateEmbeddingResponse.Usage.
• Functions:
  • Modifies hasCompletionUsage in parseOpenAI.ts to check for OpenAI.CreateEmbeddingResponse.Usage.
  • Updates wrapMethod to handle both OpenAI.CompletionUsage and OpenAI.CreateEmbeddingResponse.Usage for streaming responses.
• Misc:
  • Adjusts completionTokens handling in parseUsage to default to 0 if not present.

^{This description was created by for 7cbb60c. It will automatically update as commits are pushed.}

[vercel]

@selbyk is attempting to deploy a commit to the langfuse Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

Disclaimer: Experimental PR review

PR Summary

Modified OpenAI usage parsing to correctly handle token reporting for both completion and embedding responses, fixing missing prompt tokens in embedding responses.

• Updated type handling in langfuse/src/openai/traceMethod.ts to support both OpenAI.CompletionUsage and OpenAI.CreateEmbeddingResponse.Usage
• Added type guard hasCompletionUsage in langfuse/src/openai/parseOpenAI.ts to handle optional completion_tokens
• Modified usage parsing logic to default completion_tokens to 0 for embedding responses

_{💡 (3/5) Reply to the bot's comments like "Can you suggest a fix for this @greptileai?" or ask follow-up questions!}

_{2 file(s) reviewed, no comment(s)
Edit PR Review Bot Settings | Greptile}

[selbyk]

Closing because PR #486 addresses this issue in a better way.

@hassiebp, could you take a look at my other PR #483 as well when you get a chance? We're patching it in at the moment, but I hate relying on internal patches to libraries

I really appreciate the quick attention to the usage issue (unless you had already happened to grab that ticket off your backlog by coincidence!)

Happy Friday, and have a good weekend++