Add AfterLLMCallback to modify LLM stream text post-generation

[achapla]

Introduced an after_llm_cb function to enable developers to modify the LLM-generated text before it is committed to transcription and passed to the before_tts_cb function. This function provides flexibility to adjust or remove specific tokens or phrases from the transcription source, ensuring a cleaner conversation flow without requiring frontend workarounds. The modified text now flows through to before_tts_cb, allowing consistent text handling across the conversation stages.

#1044

[changeset-bot]

⚠️ No Changeset found

Latest commit: 6f68997

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[CLAassistant]

All committers have signed the CLA.

[achapla]

@davidzhao I don't have experience with Python, so please review the parameter types for after_llm_cb.

Note: Modifying the text in the after_llm_cb callback function will pass the modified text to before_tts_cb.

[theomonnom]

Hey, this makes sense, I think we should expose the before_forward_cb argument of the TTSForwarder inside the VoicePipelineAgent instead

[davidzhao]

Hey, this makes sense, I think we should expose the before_forward_cb argument of the TTSForwarder inside the VoicePipelineAgent instead

I think what's committed and what's forwarded should match.. the TTSForwarder would likely change with the new TextStream APIs as well?

in that sense.. perhaps after_llm_cb would make more sense? naming is always hard..

[achapla]

@theomonnom @davidzhao Thanks for the feedback! You have a solid understanding of the inner workings of the framework.
Personally, I prefer adding after_llm_cb as it is clearer and more intuitive than before_forward_cb.
I can also work on making the necessary changes to expose the before_forward_cb callback.

This feature would undoubtedly benefit many developers.