Enable metadata tracking with Aim

[roubkar]

Hey, Ruben from Aim here! 👋

First of all thanks for this awesome project!

After experimenting with babyagi I found that integrating Aim can be extremely helpful for prompts exploration and comparison.

How can Aim enhance babyagi observability?

Prompts comparison

Here is an example view where it is possible to explore and compare the outputs for the same objective by different models (gpt-3.5-turbo and text-davinci-003):

Generated tokens count comparison

The same example for generated tokens count:

Log messages for debugging

View the entire session terminal logs

[roubkar]

Hey @yoheinakajima and @francip, any thoughts on this? Would like to hear your feedback on this 🙏

[letsgoduke]

would like to see this as well