"For Each" - When it is implied that a response or request is/will be iterable, start_agent for each iterable

[jwest75674]

Duplicates

• [X] I have searched the existing issues

Summary 💡

Prompts / Goals that are likely to result in an iterable list of items, treat each item as independent, and perhaps run in parallel with start_agent commands.

With long lists of iterables, AutoGPT's memory seems to choke on the idea of iterating automatically, instead tending to loop around on itself, in my experience. Regardless, in many cases, a list of otherwise iterable items is squished into a single prompt instead of breaking them up.

I propose recognizing the development parlance "For Each" to specifically be a request for list iteration, perhaps with independent agents. In a perfect world, the agents could run in parallel, but I realize that this hopeful wish has major implications, haha.

Examples 🌈

For example:

Goal 1: Identify from the client's website, https://website.com/, the most important services
Goal 2: For each service, create and save to a file a long-form description of the service, including the benefits of the service

In the above example, what is returned is a short description of each service jammed into a single ChatGPT response, one file saved, etc. Similarly, if there are many services, then the result is many short descriptions in one response.

Motivation 🔦

Example uses which run into the list iteration issue are common in the digital marketing space, as an agency owner looking to streamline our ideation needs, here are some examples:

• Key service identification from websites, as shown above.
• From key services, identification of target keywords on a per-service basis
• Same, but identification of missing content pages, given target keywords.
• The creation of ad copy (for example, Google Ads, "Responsive Search Ads" require sets of 15 advertising headlines of 30 characters or less and four descriptions of 90 characters or less, each!
• Most social content requires quantities of content, not just one post, iterating over content blocks to, for example, add keywords, or extract ideas, etc.

Basically, everything in my vertical requires iteration, and AutoGPT would benefit from a specific function to facilitate it!

Great work folks!

*Edit: Grammar

[mikehnta00]

Well said. In addition to this, perhaps there could be a file containing the ai's thoughts only. This way if it finds a duplicate, it can try to contextualize the loop its in and think of an approach to get it out.

It already does this with certain command errors. Say something like a read_file error => search directory command=> file missing => create file=>...

Maybe this could be a solution to have it work its way out. AI has thought =>AI checks for duplicates in the 'thought' file.=> duplicate is true=> Criticism: "This is a duplicate thought, check for a loop and find a different path to solve the problem"=> reads last X num of prompts => comes up with new thought.

[github-actions[bot]]

This issue has automatically been marked as stale because it has not had any activity in the last 50 days. You can unstale it by commenting or removing the label. Otherwise, this issue will be closed in 10 days.

[github-actions[bot]]

This issue was closed automatically because it has been stale for 10 days with no activity.