Finally got real results! #3546
Replies: 8 comments 11 replies
-
|
all fully automous on --continuous |
Beta Was this translation helpful? Give feedback.
-
|
What prompts did you use? |
Beta Was this translation helpful? Give feedback.
-
|
I was using notes.txt and other keywords to make it create notes to eachother, running 4-5 autogpts at a time all working together at one point, all using the same working dir. Sorry been a while since Ive been here. |
Beta Was this translation helpful? Give feedback.
-
|
Would love to see your prompts |
Beta Was this translation helpful? Give feedback.
-
|
I'll release them after testing further too, I hope you understand. It's officially at a company structure with self-reminders. I will release soon though! |
Beta Was this translation helpful? Give feedback.
-
|
A month later is soon, right? Let's hear it. |
Beta Was this translation helpful? Give feedback.
-
|
please give us an idea of what prompts you used , would really appreciate it |
Beta Was this translation helpful? Give feedback.
-
|
I made it a CEO basically. with a very structured prompt ai_settings. |
Beta Was this translation helpful? Give feedback.
-
|
Gotta spill the beans... What version? What plugins? What memory backend? Latest stable is broken badly, I'm reverting back and trying again... Trying with the new 16k model though... |
Beta Was this translation helpful? Give feedback.
-
|
I stopped using it tbh after a while when i was using too many tokens for the output I was getting, found out how to create what I wanted it to do on my own in my own codes, still love autogpt for inspiration though. |
Beta Was this translation helpful? Give feedback.
-
|
How about share a how to. I have seen others use local llm self hosted wrapper to catch open ai and instead use free tokens of own llm or multiple llm based on another intent wrap that analysis intern of request and select best model with refiners allowed from user input over time. How to on your side and I'll put the rest together and add to it. |
Beta Was this translation helpful? Give feedback.
-
|
What worked for me was essentially having multiple instances of autogpt all pointing to the same workspace dir, this dir had notes from each that the mangement AI then proceeded to read on each startup, changing their own notes and getting notes from the management within the main_notes and progress notes and such. Notes were key and keeping it in their prompting was essential too when I changed the core of their prompts during that time. I havent used this in a while so it may not work! I hope you find something amazing. You can also direct it to never duplicate names or use extended names instead of properly replacing files, one was responsible for fixing errors other ai left. |
Beta Was this translation helpful? Give feedback.
-
|
what your profit progress atm? |
Beta Was this translation helpful? Give feedback.
-
|
Very curious what you're having it do? Understand you're not giving specific prompts yet, but mind giving a brief description/overview of the overall end goal/daily goal/purpose? |
Beta Was this translation helpful? Give feedback.
-
|
Check out my Twitter doctorlordbasil it has images of them talking and
such.
…On Sat, May 13, 2023, 3:45 AM Zyler Kade ***@***.***> wrote:
Very curious what you're having it do? Understand you're not giving
specific prompts yet, but mind giving a brief description/overview of the
overall end goal/daily goal/purpose?
—
Reply to this email directly, view it on GitHub
<#3546 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/A6G5RBETLU72PO7SG3TEX23XF432RANCNFSM6AAAAAAXQOD2FQ>
.
You are receiving this because you authored the thread.Message ID:
<Significant-Gravitas/Auto-GPT/repo-discussions/3546/comments/5890704@
github.com>
|
Beta Was this translation helpful? Give feedback.
-
|
I don't see a doctorlordbasil account |
Beta Was this translation helpful? Give feedback.
-
|
I stopped on that, you can just email me at [email protected] or here ! been working on my own stuff, but I can definitely help anyone with prompting AI agents 👯 |
Beta Was this translation helpful? Give feedback.
-
Good job! :) It somethimes even hard to get it to do some stuff despite AI help, especially for those of us who have trouble doing repetitive stuff, i imagine it does not connect them but more in isolation? Its bottleneck seems to be remembering what it did and continually making progress i think. :) But awesome :)) Here if your interessted i wrote about it: #792 It did that for me gpt3.5 long time ago i have 600 in 1 folder tree it created, the problem is not that it cant code and fix the code. The problem clearly is that it forgets what it has already done!!
That picture above is outdated, many in the community seem to focus on it and very very highly informed, like the paper voyager was read a while ago and ja ^^ but its still useful to let people know "hey memory there is something there" ^^ <3 Still almost never get it to finish my complex stuff but it does lots of impressive individual stuff just not combined. Or with a significant overall plan, often loops out seemingly. |
Beta Was this translation helpful? Give feedback.
-
'Organization is everything' so to say. Seems so sometimes. 😄 |
Beta Was this translation helpful? Give feedback.
-
it now produces .exe files constantly! 90% work. gpt4 api
Beta Was this translation helpful? Give feedback.
All reactions