Aaron C Wacker PRO

Such good news thanks! With this we can now create AI pipelines with much greater simplicity to make models interchangeable service parts. I think for cutting edge techniques like MoE gating networks, Self Reward and Comparison across models, Memory across AI pipelines, etc this becomes the differentiator to make it all much easier. I hope that by operating key models like GPT-4o, Claude 3.5 Sonnet, Gemma, Llama, and other front runners in this open pattern unlocks better more powerful AI coding patterns.

It uses my openai key and org id and is hard to run in an open fashion due to usage. It uses the billed model.

You can use whisper-1 for now and that pattern works great. The speech wav stream recorder is not in the code for openai yet. I use a streamlit recorder in order to get speech in which is working but I am looking for a better speech in/out technique. The audio to text is used as well and is how the video modality inputs its transcript for additive data input with the image slices from video. One thing I also did not see yet was the image generator inside the client api. That would be nice to add as well and also the speech synthesis.

Oh wow that is so cool! I'll have to read up on this. Thanks for sharing source.

I saw the a16z moniker in the code, is that one of Marc Andreessen Horowitz's group inventions? I think they were investing in Roblox and the tech behind the multiplayer aspects with now multiagent systems is pretty neat.

Convex looks interesting too w combo of vector search, db, ts, AI and realtime services with a great stack of client libraries. I was able to double click into it with github perms and get started.

Its pretty amazing to me that HF can host something so cool. Your Dockerfile implementations are simply amazing.

As the root of the proto-indo-European language tree I thought all of our European languages came from it but the Sanskrit Heritage Dictionary has around 180k words, so it might be something smaller like token representations counting inter-combinations of symbols. If you count compound words and infleccted forms then you could probably say Sanskrit has maybe millions of words but probably not billions.. Looks like a language model hallucination to me :)