PR batch 4

[nponeccop]

Duplicates

• I have searched the existing issues

Summary 💡

The new merging flow is getting traction.

Clean merge 🌈

• Refactor seed prompt loading: replace text file with Python class for easier maintenance #837
• Fix JSON formatting in prompt.txt #884
• Improve .env File Organization, Readability, and Documentation #980
• Fix flake8 issues pt. 2 (Add E231 & E302 flake8 rules) #1014
• Improved logging on connection fail to a Memory Backend #1011
• Continuous mode limit #1016
• Auto-GPT requires numpy -- added to requirements.txt #1022
• Feature/shell script file #1030
• Linter Autofix for Readme.md  #1031
• Feature/ability have no memory #1032
• remove useless load_variables_method #1034
• Config option for azure_ad to support Managed Identities #1038
• Fix logs path #1043
• Correct link to unit tests in README #1050
• Update serial in Installation step #1053
• Update README.md on log location #1065
• Removed redundant cfg object creation in base memory file #1068
• Typo: in PULL_REQUEST_TEMPLATE.md #1071
• replace deprecated function with current equivalent #1072
• Update test_json_parser.py #1087
• Remove deprecated (404) links, and add alt-text to one image: Update … #1095
• Flags Updates #1096
• Add openai api base support #1097

Conflicts or Red CI from Batch 3:

• migrate to poetry for package management #367
• Added new env variable and speech function for alternative TTS voice #754
• Refactor seed prompt loading: replace text file with Python class for easier maintenance #837
• Add visited website to memory for recalling content without being limited by the website summary. #968

Labeled

• Adding support for openai_organization env variable #289
• Model temperature and token limit can now be set through the AI setup wizard #916
• Persuading AI to always give json in response with GPT-4 #1042
• Add Nix flakes support through direnv #1091
• Migrate to poetry #1102

Motivation 🔦

We want to merge older PRs first, so that PR authors are motivated to fix conflicts once their PR gets into a batch. This happens when I, @nponeccop , approve the PR for the first time. I only check that there are no conflicts, CI is green and the PR only fixes one thing and cannot be sensibly split.

Then the PR gets an additional review from @Torantulino , @p-i- or @richbeales . Unlike my simple review, they review the logic extensively, and subsequently merge.

The merged commits are then tested at the end of the day, and merged into the stable branch. So people wanting to try Auto-GPT should try stable, not master. If master gets broken, we fix it by either accepting out-of-band PRs or fix it ourselves.

[EwoutH]

Thanks for all this amazing work!

Would you consider using Squash and merge method for merging PRs instead of the "Create a merge commit" method? That will result in a way cleaner, and linear git log, since for each PR not multiple commits (including a merge commit) are included but only a single one.

[nponeccop]

@EwoutH Propose in a separate issue, not here

[EwoutH]

Done: #1255

[nponeccop]

Today has been a challenging day. We began by merging issues as usual, but ultimately ended up addressing a long-awaited issue with our repository structure: the introduction of the autogpt module.

This change is significant and has impacted issues from Batch 5 and a few from Batch 4 that were still experiencing Git merge conflicts when we decided to move on. We followed the same carry-over procedure as we did with earlier batches.

In addition, we welcomed some new members to our development channels on Discord, including @0xArty , @BillSchumacher , and @merwanehamadi .