Added CustomOpenaiCallback to ensure exclusive access to nested data.

[LorenzoPaleari]

See #576 for a detailed explanation.

[VinciGit00]

Thank you it looks good, do you have idea of how to implement it for the other providers? I really would like to have it for Mistral and bedrock

[github-actions]

🎉 This PR is included in version 1.19.0-beta.12 🎉

The release is available on:

• v1.19.0-beta.12
• GitHub release

Your semantic-release bot 📦🚀

[github-actions]

🎉 This PR is included in version 1.20.0-beta.1 🎉

The release is available on:

• v1.20.0-beta.1
• GitHub release

Your semantic-release bot 📦🚀

[LorenzoPaleari]

Thank you it looks good, do you have idea of how to implement it for the other providers? I really would like to have it for Mistral and bedrock

I have taken a look at it.
For my understanding it is already working for all LLMs just by using openai_callback, the only downside is that it is not able to tell the cost of models different from OpenAi one.

I tested with Mistral and outputted the number of tokens used correctly, whiteout the price.

To add price we can mimic OpenAICallback class, just adding models names and cost should suffice to make the callback work in any situation.
https://api.python.langchain.com/en/latest/_modules/langchain_community/callbacks/openai_info.html#OpenAICallbackHandler

[VinciGit00]

Ok thx, is it ok if you implement it?

[LorenzoPaleari]

@VinciGit00 Yes, I can do that, will probably do at the end of the week.
Will also try playing around to give a more detailed view of what happens under GraphIterator, to give a more detailed view of where all this used tokens goes. I'm not sure if I can end up with a good result without interfering with function calls, will let you know

[VinciGit00]

Ok thank you so much for the effort

[github-actions]

🎉 This PR is included in version 1.21.0 🎉

The release is available on:

• v1.21.0
• GitHub release

Your semantic-release bot 📦🚀