#Tool for connection AI, tools, and packages

1 messages · Page 1 of 1 (latest)

nimble depot
#

I am currently working on a package with the purpose of connecting different kind of tools/AIs.
The idea is, that you have a tool with an input and output, add use the package to connect everything.

For example:
you have

  • speech-to-text (input audio; output text)
  • GPT (input text; output text)
  • text-to-speech (input text; output audio)

If you make use the package to connect them, you can just tell connect the output of one tool with the input of an other tool.
speech-to-text (output) -> text-to-speech (input) -> back to speech-to-text

You can remove the connection (without restart) and make a new one.
For example:
speech-to-text (output) -> (input) GPT (output) ->(input) text-to-speech (output) -> (input speech-to-text

The same tool can used as just the controller to connect everything.

I don't know if that would help vedal with neuro-same, because some people can make in instance with that tool without knowing the neuro-same code (and vedal can use it), some parts can be shut down and started if needed, and it can be split to more as one computer/server.

I don't know if it would help vedal with adding more function, or just make everything harder.

heady ocean
#

@nimble depot is this still a thing?

nimble depot
# heady ocean <@82982442461302784> is this still a thing?

I didn't continue working on it, but I got a working version with tts->chatGPT->stt

I noticed that there exists a better solution for the messages, and working on it alone make no sense if I have nothing to use it about.

I may make something similar at some point again.

gentle mural
#

can I ask how this would compare to the current stack? I don't see much of a use for it anymore, though maybe I'm wrong

heady ocean
#

I'm not sure what's this for either, but I'm writing down anything I see.

heady ocean
nimble depot
# gentle mural can I ask how this would compare to the current stack? I don't see much of a use...

This post is 3 years old.
I guess a lot of things happened in that time.

I had no idea how the stack and communication between the tools worked, because it was not really open knowledge. (And everyone did something different)

The idea was, that you have a block with input types and output types. The block would have a description, too.

For TTS it would be input=text output=done. The block itself would be the tts with microphone.

If you have enough fitting blocks, you can change everything on the fly, because you just connect it differently.

For example, switch the TTS with send-to-chat, because both have text as input, and only output a "done"

That was the idea.
It is probably not something you need or want now.

gentle mural
#

maybe after the supposed rewrite it will be easier for vedal to implement it himself

heady ocean
nimble depot
nimble depot
# heady ocean So modular tools for making an AI streamer? (sorry if I'm dumb)

Yes and no.

You can make an AI streamer with it, because you can make a loop with it.

What I did was making something like an "hey google" with chatGPT.
I put a filter module (input=text output=text) in between, that was cutting everything out before a "hey google".

No worries.
I try to explain it, but am not sure if I explain it good enough.
Feel free to ask.