baidu-qianfan[patch]: Fix streaming mode of Qianfan

[stanoswald]

Even when streaming is set to true, the original completionWithRetry function waits until all chunks are received before triggering the event and returning the result.
This PR updates the completionWithRetry, _generate and use _streamResponseChunks function to handle streaming chunk from BaiduQianFan.
The refactor also fixed the original error of Invalid "runId" provided to "handleLLMNewToken" callback.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchainjs-docs	✅ Ready (Inspect)	Visit Preview		Sep 10, 2024 8:24pm

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchainjs-api-refs	⬜️ Ignored (Inspect)			Sep 10, 2024 8:24pm

[stanoswald]

@dl102306

[jacoblee93]

Thanks for this @stanoswald!

This seems like a fairly big refactor, and while it all looks reasonable on the surface, I'd love @dl102306's opinion as well since I'm personally unable to test this integration.

Will give them a few more days and if they don't respond, will merge.

[stanoswald]

Thanks for this @stanoswald!

This seems like a fairly big refactor, and while it all looks reasonable on the surface, I'd love @dl102306's opinion as well since I'm personally unable to test this integration.

Will give them a few more days and if they don't respond, will merge.

Sure, I personally have used this branch in production for a while, but only in the usual way of non-streaming and streaming invocation. I'm not sure if this code was mainly maintained by the QianFan platform. As their client, I'll also try to contact them to see if they can review this PR and try to give it a full test to make sure it works well.

[dl102306]

@jacoblee93 @stanoswald Sorry for later response. This is a big changes. I need do some test to check this changes works fine today. Please wait for a day or two, thanks a lot.

[dl102306]

@jacoblee93 @stanoswald please merge these changes. I have tested these changes. It's work well. thanks~

[jacoblee93]

Thanks both of you!

[stanoswald]

Thanks for looking into this guys! I'll make some further fixes for CI later.

[jacoblee93]

@stanoswald this is failing CI:

src/chat_models.ts(343,20): error TS18048: 'finalChunk.generationInfo' is possibly 'undefined'.

Can you fix?

[vercel]

@stanoswald is attempting to deploy a commit to the LangChain Team on Vercel.

A member of the Team first needs to authorize it.

[stanoswald]

@jacoblee93 I just updated the code, it should be fixed now.

[stanoswald]

@jacoblee93 Hello, could you authorize this to let the CI run?

[jacoblee93]

Thank you!