OpenAI Whisper model is open source and can be run locally. There is a "light" version also to help it run on lower resourced machines, e.g local machine perhaps
Have setup a python package that makes it easy to interact with LLMs over voice
You can set it up on local, and start interacting with LLMs via Microphone and Speaker.
The idea is to abstract away the speech-to-text and text-to-speech parts, so you can focus on just the LLM/Agent/RAG application logic. And it does not tie you down to any specific LLM model or implementation. i.e you can use it to build simple or complex applications using the tech stacks of your choice
Currently it is using AssemblyAI for speech-to-text and ElevenLabs for text-to-speech, though that is easy enough to make configurable in the future
I just kickstarted this off as a fun project after working a bit with Vapi
Has a few issues, and latency could defo be better. Could be good to look at some integrations/setups using frontend/browsers also.
Would be happy to put some more time into it if there is some interest from the community
Package is open source, and is available on GitHub and PyPI.
I kicked of a small PHP package, with the aim of making it easy to work with LLMs, Agents and RAG in PHP.
It's just got a single method/class just now :) (release fast!?)
Wondering if it is something that sound's interesting to the PHP community. Why should Python guys have all the fun!
Would be great to have some sign of interest, and even feature requests, to inspire me to put some more work into it. Have kicked off the package here on github, and it is available on packagist also.
Like I said, it currently just supports one action, which allows you to send a message to OpenAI Chat by doing
$chat = new ChatOpenAI(getenv('OPENAI_API_KEY'));
$response = $chat->sendMessage('Hello, world!');
I know there is the LLPhant library, but this could evolve to be something a little different, perhaps with its own benefits
Any feedback? Feature requests would be great too.
