Feature/ability have no memory

[waynehamadi]

Background

Sometimes, all you need is to do a short task. And you don't need to have the memory. In fact the memory currently is very heavy:

• we copy an array of json into the context, just after the question.
• GPT-3.5, forgets its goals sometimes, because of it. Because the goal is at the top of the chat, and the memory is the last thing GPT sees.
• it makes testing less consistent.

Another aspect of this change is:

• constraints unlock creativity: if you know you have "only" 8000 tokens (and 8000 tokens is actually huge when you think about it), you're going to do your best to optimize your agent so that it summarizes its current progress.

Changes

All I did was giving the option to have no_memory:

• currently MEMORY_BACKEND is not in the env.template, so I didn't add it. I am assuming it's not in the env template because we don't want to overwhelm users with options.
• however for advanced users they can put MEMORY_BACKEND="no_memory" and they will have a very consistent agent.

This, along with the change in temperature I pushed in another PR will unlock more consistency for our integration tests.

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