feat(monologue agent): enable dialogue across multiple tasks

[anyuyay]

Currently, the monologue agent resets its monologue for every new task. We propose that the monologue agent should handle a series of related tasks instead. After testing, we suggest adding an action that marks the start of each task. This would allow the monologue agent to maintain a continuous dialogue across multiple tasks.

With this change, the monologue agent could handle scenarios like this: First user input: "Write a python file to convert a temperature from Celsius to Fahrenheit." Second user input: "Explain what you just wrote."

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 25.00000% with 6 lines in your changes are missing coverage. Please review.

:exclamation: No coverage uploaded for pull request base (main@06aae67). Click here to learn what that means.

Files	Patch %	Lines
agenthub/monologue_agent/utils/monologue.py	20.00%	4 Missing :warning:
agenthub/monologue_agent/utils/prompts.py	33.33%	2 Missing :warning:

:exclamation: Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1570   +/-   ##
=======================================
  Coverage        ?   58.73%           
=======================================
  Files           ?       92           
  Lines           ?     3761           
  Branches        ?        0           
=======================================
  Hits            ?     2209           
  Misses          ?     1552           
  Partials        ?        0           

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[rbren]

@anyuyay thanks for this contribution!

I actually have a more thorough fix in progress here: https://github.com/OpenDevin/OpenDevin/pull/1646

That will ensure all agents have the ability to work across tasks, rather than just monologue

[neubig]

Hey @rbren, just to confirm: is this PR obsolete given that a more thorough fix is in progress with #1646? If so maybe we can close it?

But @anyuyay , thanks a bunch for the contribution either way!

[rbren]

Yup I think this is obsolete now--all agents should automatically have multi-turn enabled now.

Thanks for the inspiration @anyuyay!