Add: Auto Confirm For Self-Feedback #3934

[AbTrax]

Background

#3934

Changes

Added Autoconfirm for when self-feedback is used as to #3934

Documentation

Test Plan

PR Quality Checklist

• My pull request is atomic and focuses on a single change.
• I have thoroughly tested my changes with multiple different prompts.
• I have considered potential risks and mitigations for my changes.
• I have documented my changes clearly and comprehensively.
• I have not snuck in any "extra" small tweaks changes

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Comments	Updated (UTC)
docs	⬜️ Ignored (Inspect)	Visit Preview		May 12, 2023 10:41am

[codecov]

Codecov Report

Patch coverage: 2.70% and project coverage change: +0.21 🎉

Comparison is base (bc6f8a2) 60.72% compared to head (f97e454) 60.93%.

❗ Current head f97e454 differs from pull request most recent head 1790800. Consider uploading reports for the commit 1790800 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4000      +/-   ##
==========================================
+ Coverage   60.72%   60.93%   +0.21%     
==========================================
  Files          73       73              
  Lines        3320     3315       -5     
  Branches      475      543      +68     
==========================================
+ Hits         2016     2020       +4     
+ Misses       1164     1156       -8     
+ Partials      140      139       -1     

Impacted Files	Coverage Δ
autogpt/agent/agent.py	46.71% <2.70%> (-1.59%)	⬇️

... and 4 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[waynehamadi]

@AbTrax I am on your branch, running python -m autogpt --debug and I keep typing "s"
And it keeps sending back a very long sentence if it agrees with the next action.
And on top of that it doesn't do the actual action, this means if I keep typing s it will keep doing nothing

Here is what we want:
typing s should do the next action if supervisor agrees, otherwise it should give clear and concise sentence to nudge the agent in the right direction.

Please look at my debug files to see the issue for yourself.
DEBUG.zip

Also you didn't pick the work of @Torantulino in his commit, this commit looks pretty good

[waynehamadi]

@AbTrax I am on your branch, running python -m autogpt --debug and I keep typing "s"
And it keeps sending back a very long sentence if it agrees with the next action.
And on top of that it doesn't do the actual action, this means if I keep typing s it will keep doing nothing

Here is what we want:
typing s should do the next action if supervisor agrees, otherwise it should give clear and concise sentence to nudge the agent in the right direction.

Please look at my debug files to see the issue for yourself.
DEBUG.zip

Also you didn't pick the work of @Torantulino in his PR.

[Boostrix]

if anybody has got a clever idea how to use self-feedback to tell that we're inside a redundant loop, please speak up, I implemented a crude first stab at #3668 by hashing the response hash (thoughts) in conjunction with the upcoming commands + args (all stringified), and it's working nicely to notice that we're about to do something that we have previously done - but that's obviously not the complete story.

We need a way to tell where exactly we're in our top-level plan, to decide if we really want to continue or discard the current iteration.

This (PR #3668 ) is just based on hashing full thoughts + new decision (command + args) and incrementing a counter every time we get to see the same "situation":

PS: See also: #4129

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 58 minutes (more than 100, code: "api-deployments-free-per-day").

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 55 minutes (more than 100, code: "api-deployments-free-per-day").