Is there a callback in useChat that can be used for each chunk that comes in? (like "onChunk" or something) in AI SDK v5?

[rngyn]

I want to use useChat for my application, but I'm needing to process each chunk as they come in. As far as I can tell, useChat does not have an escape hatch for me to intercept the text chunks as they come in and to do something with them. onData only works for "data-..." types, but not for the "text-delta".

Is there a workaround for this? Or can this be an added feature? Right now I'm just using direct streaming and not useChat, but useChat would be easier to use with its built-in API functionalities.

I know there is an onChunk on server-side, but it would be helpful to have something on client-side.

Some insight would be helpful, thank you.

[arnaugomez]

I would love to have this. Especially for streaming tool calls in the client-side.

Currently, the onToolCall callback for executing tool calls in the client-side only gets executed when the tool finishes streaming.

However, there is no such thing as an onToolStream callback where I can partially execute a tool while the tool is streaming.

The workaround is to get to the messages property of the return type of useChat, and use a useEffect hook to respond to the changes in messages.

[arnaugomez]

For example: we're building a toolkit to let AI agents edit rich text documents inside a rich text editor. The tool calls are applied in the client-side, so if I want to stream content into the editor, I need to use this workaround: https://tiptap.dev/docs/content-ai/capabilities/ai-toolkit/guides/tool-streaming

[github-actions[bot]]

This discussion was automatically locked because it has not been updated in over 30 days. If you still have questions about this topic, please ask us at community.vercel.com/ai-sdk