Refactor print() calls to use structured logging (absl.logging)

[manaspros]

I was reading through the codebase and noticed there are quite a few print() statements scattered throughout (I counted around 200+). While this works for quick debugging, it makes it harder to control output in production environments and doesn't follow Python logging best practices.

Current Issues:

1. No log level control - Can't filter debug vs info vs warning messages
2. Hard to debug in production - No timestamps, no module context, no log levels
3. Inconsistent with project standards - The project already uses absl-py for flags, but logging is inconsistent
4. Can't redirect or suppress output - Print statements always go to stdout

Examples I Found:

In various files

print(f"Agent {self.name} taking action...")
print("Memory retrieved:", memories)
print(f"Debug: {some_variable}")

Proposed Solution:

Replace print statements with absl.logging (already a dependency) or Python's standard logging module:

from absl import logging

Instead of print

logging.info("Agent %s taking action...", self.name)
logging.debug("Memory retrieved: %s", memories)

I'd be happy to work on this systematically across the codebase. I think it's a good way to get familiar with all the modules while improving code quality.

Would the team be open to this? Happy to start with a small PR on a single module to get feedback on the approach before doing the full refactor.

[locross93]

Hi @manaspros, thanks for bringing this up. Since absl-py is already a project dependency, standardizing on absl.logging makes sense. Yes we are open to you or others sending PRs on this, please do so. Thank you for contributing!

We can likely leave print() in examples/ unless they are debugging statements

[manaspros]

Hey! I'd like to work on this.

I found ~250 print() statements in the concordia/ directory. To keep things manageable, I'm planning to break this into a few smaller PRs rather than one massive change.

First PR: I'll start with the core simulation engines:

• concordia/environment/engines/sequential.py (~10 prints)
• concordia/environment/engines/simultaneous.py (~9 prints)

These files have patterns like:

if verbose:
    print(termcolor.colored(f'Event: {event}', 'cyan'))

I'll replace them with:

logging.info('Event: %s', event)

This way we get proper log levels without needing the verbose flag, and it's easier to review.

After this first PR, I can tackle the LLM modules, components, and other files in follow-up PRs if the approach looks good.

Does this sound reasonable? Happy to adjust the approach based on feedback.