chadbailey59

OpenAI and Fireworks vision services

1

I might try to add Fireworks image gen support tomorrow, but we can also merge this as-is if we need to cut a release

This adds another example server (`fly-server.py`) that uses fly.io's API to spin up a new Fly Machine for every new bot, instead of using a python subprocess in the web...

Implement LMNT full-duplex TTS generation

1

[LMNT](https://lmnt.com) has an interesting approach to 'full-duplex' TTS: ```python async def main(): async with Speech() as speech: connection = await speech.synthesize_streaming('lily') t1 = asyncio.create_task(reader_task(connection)) t2 = asyncio.create_task(writer_task(connection)) asyncio.gather([t1, t2]) connection.close()...

Function calling

1

Let's wait to merge until I build a foundational example too.

Re-enable function calling in GPTs, and add a foundational example showing how it works.

Simpler context management for LLMs

2

As [mentioned in Discord](https://discord.com/channels/1239284677165056021/1239289240756551781/1240604191047549029), Pipecat bots don't automatically maintain an LLM context object for you; instead, you have to use a bunch of aggregators and other tools to manage that...

Deployment examples: What do you want to see?

9

We want to build out some code and examples to make it easier to deploy Pipecat bots. Right now, we're looking at the following services: * Hugging Face Spaces *...