You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jul 4, 2024. It is now read-only.
Hi! I was finally successful in running your plugin with Auto-GPT stable 0.3.1. Long story short, I struggled in getting the right content for the settings template and most of all, I had to set the EOL format to LF instead of CRLF. I remember reading that somewhere before, but I spent a long time looking for bugs that aren't there while distracted by this simple detail... Oh and I had to remove the 'MEMORY_INDEX' from the cmd. I don't even see how that could ever work, as it's not a known command.
So I've got clones running left and right 👍but they never show as agents. As in, autogpt agents. You may be well aware of this, as you also state you're working on cross-communication, but do you want to have a discussion on that? I've been looking at the code long enough by now to come up with ways to make that work. One thing is, autogpt seems adamant to message an agent using its name as the key, which should be an integer, not string. I changed the code for it to work with string keys/indices all the same, but that's not the issue. How can we make autogpt clones show up and be manageable like agents?
Just reaching out to offer my collaborative efforts on this...
The text was updated successfully, but these errors were encountered:
I'll make a note about the EOL format in the upcoming release readme. I've also updated how I pass .env variables to the clones/agents, so there's no need to add MEMORY_INDEX to the cmd directly.
I have interactivity working via the Telegram Plugin, but it's still 1:1 with the human user - they're not yet cross-talking. There are ways to do it via redis and rabbitmq, but I've been spending most free time helping the AutoGPT core team.
That said I plan on publishing a new release and some howtos soon.
@Yeti770 - this is my "Hello world' ai_settings.yaml file
ai_settings.yaml
ai_goals:
- Create a clone with name="C3PO" and goals="Create a file called hello-world.txt and write your name to it."
- Exitai_name: Dollyai_role: An AI that can clone itselfapi_budget: 0.0
Hi! I was finally successful in running your plugin with Auto-GPT stable 0.3.1. Long story short, I struggled in getting the right content for the settings template and most of all, I had to set the EOL format to LF instead of CRLF. I remember reading that somewhere before, but I spent a long time looking for bugs that aren't there while distracted by this simple detail... Oh and I had to remove the 'MEMORY_INDEX' from the cmd. I don't even see how that could ever work, as it's not a known command.
So I've got clones running left and right 👍but they never show as agents. As in, autogpt agents. You may be well aware of this, as you also state you're working on cross-communication, but do you want to have a discussion on that? I've been looking at the code long enough by now to come up with ways to make that work. One thing is, autogpt seems adamant to message an agent using its name as the key, which should be an integer, not string. I changed the code for it to work with string keys/indices all the same, but that's not the issue. How can we make autogpt clones show up and be manageable like agents?
Just reaching out to offer my collaborative efforts on this...
The text was updated successfully, but these errors were encountered: