Mark Backman comments

Results 240 comments of


                                            Mark Backman

Daily transcription opens multiple websockets instead of reusing the existing after calling capture_participant_transcription

FYI: There is a duplicate transcription that resides in Daily's server-side code. We're working on a fix.

Daily transcription opens multiple websockets instead of reusing the existing after calling capture_participant_transcription

Do you still see this issue?

how can Livekit transport support video?

We're happy to add better support for LiveKit, but we also need some helping hands; ideally, someone with LiveKit expertise. Is anyone available to help out?

TTS and LLM function calls are not synched

I believe the issue is inherent with how LLMs work. The sequence of events is: 1. User asks to turn on lights 2. LLM calls the tool to "turn_on_light" and...

TTS and LLM function calls are not synched

@itsikbelson-spotlight a few things: - PR #1683 hasn't been merged, so it's not available yet, but it solves a different problem - you suggest that the LLM should speak about...

TTS and LLM function calls are not synched

Also, you point to #1683, but I'm not sure it does what you're thinking. It specifically handles the case where two function calls runs in the same LLM turn. The...

TTS and LLM function calls are not synched

> However, since we're in streaming mode, the LLM outputs tokens whenever it's ready. Correct. > The function call itself can be in-between tokens or after the tokens (for instance...

TTS and LLM function calls are not synched

I see. You are adding `Always respond with text explaining what you are going to do before you call function.` to the system prompt to get the LLM to speak...

add OpenAISTTService

Superseded by: https://github.com/pipecat-ai/pipecat/pull/1183

Potential Memory Leak in with `videotestsrc` `GStreamerPipelineSource`

@aconchillo please take a look when you have some time.