search

run-llama / llama-lab

1.39k stars 168 forks source link

Auto/BabyAGI Programatic access #9

Open Kav-K opened 1 year ago

Kav-K commented 1 year ago

Any plans in the future for enabling "calling" autoGPT/babyAGI through an endpoint or a function so that we can receive results in tidbits that can be used programatically, like doing a .query() call, kinda?

logan-markewich commented 1 year ago

Interesting idea!

So you mean, letting the AGI run off in it's loop in the background, and then pulling in it's status or results as it goes?

At least in llama_agi, all the task history (current and completed) is stored in two GPTListIndex instances, so querying against it would be easy enough.

Kav-K commented 1 year ago

I was thinking more so, a sequential API that allows us to see the response of each step and then send a request to proceed it as well

logan-markewich commented 1 year ago

@Kav-K ah, so like a human-in-the-loop kind of thing?

I was actually thinking of how to add this! Basically, it would also give the user a chance to edit the tasks or objective as the AGI progresses 🧠

Kav-K commented 1 year ago

Yeah! And I feel like it'd be a useful features that would allow a lot of platforms to easily run AutoGPT as a service! :)