hey Cyril, this looks awesome honestly. I was pointing out in abstracted agents, but your library looks very good for agentic model building in general. is there a windows version of this, or what would it take to adapt this library?

thought I could contribute in building the full platform itself, then of course adding some agents of my own 🫠

hahah alright, not what I had in mind exactly but fair enough :)

awesome!! should I join this, or follow?

fantastic! count me in if I'd be of any help. would be awesome to work with hf fellows 🙏🙂

then you'd have this nice collection of agents, some portal that users log in to delegate tasks to computer systems 🙃

awesome, I mean hf could be the main provider of the pipeline. i.e. I would write a full agent for the bakery: tts + llm + stt (with the toolbox) and push into a portal: "hf agents" just like hf inference and the end user should run it just like how it works with spaces, the inference pipeline, etc.

the toolbox component looks really good, almost there, but still presents some questions, what tools can you combine and how, perhaps some community work here would do a good favor.

it is adding another layer of abstraction in essence, sure, but I'd say it's a useful layer of abstraction 🙂. should lower the learning curve quite a bit, and enable wider use.

haha yeah

that looks really good! haven’t seen this one before. but I meant something even more packed, perhaps similar to what dspy does with language models, in the aforementioned bakery code this’d be:
import hf
agent = hf.agent(“order_agent”)
order = agent(prompt, phone_call)
print(order)

similar to what hf already excels on: pipelines for inference, fine-tuning, datasets, spaces…agents?

lmk if that sounds good

definitely, sure agentic suite with easy agent add-ons for apps would be very nice imo.

imagine you’re building a phone-order pipeline for some bakery, it would be awesome to have an easy conversational agent add-on