AnswerToSpeech

[ZanSara]

Proposed changes:

• Add two new primitives, AudioAnswer and GeneratedAudioAnswer, that contains references to audio files into their answer and context fields.
• Add a new node called AnswerToSpeech that can convert regular Answers into a GeneratedAudioAnswer.

Open questions:

• What is your opinion on the new primitives? Is this a direction we would like to take to support multimodal input-output in Haystack?
  • Note, this is a stub: if we like hierarchical primitives, I will do some refinement and come up with a clean hierarchy.
• Shall I add a short tutorial on how to use this node?

Status (please check what you already did):

• First draft (up for discussions & feedback)
• Final code
• Enable actions on your fork
• Added tests
• Updated documentation: Create answer to speech documentation haystack-website#339

[julian-risch]

Looks very promising! I left a few comments that we could discuss and I would like to see a small example pipeline/tutorial. Before we can merge this we would need to add docstrings and also test cases so there is still some work to do but so far it looks great! 👍

[julian-risch]

LGTM! 👍 This is still a bit experimental I'd say and our documentation could mention that. It will be good for us to see how the community uses this feature. We could even ask for feedback on Haystack slack explicitly. Maybe a community member wants to showcase a use case, who knows? Let's make it as simple as possible for the community to try this out. I left a few comments regarding doc strings and I'll add the needs-documentation to make @agnieszka-m aware of this upcoming new pipeline node.

[julian-risch]

LGTM! 👍