Skip to content

Conversation

@rosmur
Copy link

@rosmur rosmur commented Dec 21, 2025

Summary

This PR migrates the project from conda to uv for dependency management. uv is the modern, performant option and has become the de facto standard in the ecosystem now.

Changes

  • Added requires-python = ">=3.9" to pyproject.toml for proper Python version specification
  • Generated uv.lock for reproducible dependency resolution
  • Removed requirements.txt and requirements.in to enable pyproject.toml as the single source of truth
  • Updated README with clear uv installation and usage instructions. Includes fallback instructions for pip users who can generate requirements.txt if needed

Testing

  • Verified uv sync successfully installs all dependencies
  • Tested uv run sharp --help works correctly
  • Confirmed all project dependencies are properly declared in pyproject.toml

- Update README.md to recommend uv instead of conda
- Add requires-python field to pyproject.toml
- Add uv.lock for reproducible dependency resolution
- Users can still use pip with requirements.txt if preferred
- Remove requirements.txt and requirements.in files
- Update README to focus on uv as primary dependency manager
- Add instructions for activating virtual environment
- Document how pip users can generate requirements.txt if needed
- This completes the migration from conda to uv
@rosmur
Copy link
Author

rosmur commented Dec 21, 2025

Migration Notes

Users can now use:

uv sync
uv run sharp --help

@truedat101
Copy link

Oh, I just made my own fork with some uv "helper" instructions under a new section called quickstart, no code changes : https://github.com/IoTone/ml-sharp/blob/main/README.md#quickstart , my rationale for this is similar to what is suggested in the description. I am noticing a trend that a lot of people who want to try out the tools on their own don't have the background on virtualenvs or managing python toolchain (mis) installation to get anything working. I personally don't see a reason to move people off of pip/conda. Is there any reason why uv instructions couldn't live along side conda/pip? A good CI workflow can ensure all the acceptable ways to build the tools will just work.

@rosmur
Copy link
Author

rosmur commented Jan 14, 2026

Yes no reason why both pip and uv can be options. I think its good for the ecosystem to move away from conda though. Hope the maintainers can chime in

@truedat101
Copy link

Yeah, I'm supporting some non-dev devs and getting into anything that requires long installs, paths to be set up, numerous options during install that might not be consistently displayed depending on the platform, it is better to have something that we can get people through without failure or too many edge cases related to the platform.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants