AutoGPT
AutoGPT copied to clipboard
More robust settings/onboarding stuff
Background
Felt I need more optionality at start-up: load last settings, load setting file from your history or create new settings. The need came out of running different "missions" with changes to the settings and prompts. Wanted an easy way to recall previous settings
Changes
See new onboarding flow here: https://user-images.githubusercontent.com/1012888/230700981-90509b51-c38a-47d6-96bf-47519c91c6d3.mov
- Created new structure in scripts/data:
scripts/data/ai_settings
scripts/data/ai_settings/last_settings.yaml
scripts/data/ai_settings/history
scripts/data/ai_settings/history/settings_20230407-195943.yaml [every setting that is run is saved in the history dir]
- Refactored
main.py/construct_promptto support 3 onboarding options:
Welcome! Please select one of the following options:
a - Load last used settings
b - Load another settings file
c - Create new settings
- Refactored ai_config to support new i/o stuff
Test Plan
- Manual testing throw each onboarding flow and i/o operation
Change Safety
- [ ] I have added tests to cover my changes
- [ ] I have considered potential risks and mitigations for my changes
Hey love this idea... seems like this would make it easier to share config files as you described 'missions'
This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.
This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.
I would also be interested in seeing this updated/integrated and reviewed, because it's heavily related to the benchmarking issue: #1359
This is a mass message from the AutoGPT core team. Our apologies for the ongoing delay in processing PRs. This is because we are re-architecting the AutoGPT core!
For more details (and for infor on joining our Discord), please refer to: https://github.com/Significant-Gravitas/Auto-GPT/wiki/Architecting
Interesting proposal! Unfortunately we couldn't process it in time, and now the PR is so out of date that merging it would probably be more work than recreating it from the current master.
Even though we can't merge this, we would be happy to have your involvement. Feel free to check in with us on discord for parlay.
Thanks for submitting!