PR batch 3

[nponeccop]

Duplicates

• I have searched the existing issues

Summary 💡

This is a current batch we are merging. It's frozen, no new PRs come in, although we do merge very simple or critical PRs out of band. The new PRs go to #1015

Clean Merge🌈

• Add capability to execute shell commands #463
• command_name null check before calling .lower() #542
• GPT-4 Only Option #733
• Delete ai_settings.yaml #742
• Update .gitignore to ignore potential .idea folder #746
• Add OpenAPI Key Link to README #752
• Fix case where index throws value error. #759
• update gitignore venv/* to *venv/* #770
• browse: (1) apply url validation also to scrape_links(), (2) add unit-tests for scrape_links() #780
• Remove duplicated unnecessary codes #781
• Adds information on how to use the other cache methods available #794
• Improve Dockerfile #798
• More fixes for Azure hosting #802
• refactor debug to debug_mode #805
• Fix flake8 E303,W293,W291,W292,E305 #810
• Fix #803 #825
• check for authorise 'y' without trailing space #827
• fix AttributeError: 'Config' object has no attribute 'debug' #858
• Bugfix for #803 - Debug mode #888
• Bugfix for #840 - Local memory fix #913
• fix Error: 'charmap' codec can't encode character '\u0142' in position 99: character maps to<undefined> #938
• Adds a '--use-memory' flag to choose memory backend at runtime #957
• Added 'MEMORY_BACKEND' to .env.template and updated the README #971

Needs Discussion

• migrate to poetry for package management #367 -- we need to merge Migrate to poetry #1102 which needs discussion
• Update commands.py to support bitcoin commands #749 (piggybacks on Create bitcoin_utils.py to make bitcoin payments available to agents  #750)
• enable milvus as memory backend #801
• Chroma memory #852
• Add the ability to configure a custom DNS server. #931

---

These are the PRs currently classified as unmergeable (by me, I'm not a maintainer), due to various reasons. I will add more PRs as we are preparing the Batch 3.

Major Features

• First draft at adding persistent memory via sqlite3 #124
• Adds vocode integration #148
• Feature/weaviate memory #424
• enable milvus as memory backend #801

Require Benchmark

• Fix JSON string escaping issue #72
• Prompt engineering adjustments #116
• Update prompt.txt #164
• ✨ [Feature] language available as an option #171
• Prompt file update for more specific command identification #172
• initial wolfram alpha integration #437
• Fix user_input prompt - request response should only be in JSON #741
• Add knowledge cutoff date to prompt #758

Need splitting

• devops cleanups #49
• Improve extract_hyperlinks to honor base url #87
• Use selenium for web browsing #121
• Add flake8 workflow, clean up whitespace, and remove unwanted prints #153
• Speak.py adjustment, move console logging to own file, and selenium stuff. (its been a while since my fork was merged) #212
• Refactor ai functions #221
• Fixed Json Error #285
• Code quality workflows #320
• A number of refactorings #325
• 👋 🤩 Convert to proper python module. Working Docker image. #329
• Adds Pre-Commit Hooks, Updates Requirements.txt & Extends .gitignore #388

Motivation 🔦

We need to merge all the simple requests first.

[blankster]

@nponeccop love your "batches" and how you support the maintainer 👏

[Artemonim]

@nponeccop maybe you can ask Maintainer for the Issue Watcher role and use label system for this?
It would greatly improve the navigation of issues and PRs

[nponeccop]

There is already a discussion at #469, you can add your suggestion there.

[nponeccop]

@Torantulino I see some PRs were merged out of band. After finishing with the important PRs can we get back to the 5 remaining PRs in the Batch 2? It's important to merge the PRs FIFO (except the high-priority ones) so that the authors are motivated to do the rebasing/conflict resolution. We need to give them an idea that once their request is accepted, it will get merged in the next batch. If the requests are non-mergeable for some reason we need to clearly indicate the reason we postpone, close the batch and re-schedule the remaining requests to a different batch.

I heard there was a conference call yesterday, what are the decisions taken there regarding the PR flow if any?

[Artemonim]

Can I participate with #889 ? 👉👈😳

[nponeccop]

@Artemonim Done.

For future visitors, the right way to do it is to tag me. I read all Auto-GPT updates in email. Or just ask for a review.

[nponeccop]

The text below, is obviously, courtesy to ChatGPT.

Batch 3 has been closed, and we've moved onto Batch 4 (#1015), which is currently frozen. Any new pull requests will be directed to Batch 5 (#1162).

To keep things organized, we've added labeling to our pull requests. If we're not ready to merge a request yet, we label it as "need-discussion", "enhancement", or "question". Today, our second merger, @p-i-, is finally able to work at full speed.

Any remaining unlabeled issues from Batch 3 will be moved to Batch 4, while any labeled issues will remain here until they are de-labeled. This is because we want to merge all the smaller pull requests first, before moving on to larger architectural improvements and technical debt. Our current version isn't very stable even with its limited set of commands, due to well-known JSON parsing issues, as well as other usability and reliability issues. Therefore, we need to fix these issues first.

We set up CI today and will be increasing the strictness of flake8 every day, so be prepared to fix more linting issues. We now have a stable branch which is updated once a day and is smoke-tested by the maintainers. The master branch is quite volatile, and we encounter daily breakages that we fix at the end of the day before releasing. As a developer, you can help us by testing the intermediate master branch state for breakages. If you're a general user, you can get a general idea of the progress by downloading a versioned GitHub release or cloning a tag.