Migrate dependency management to uv

[Syzygy106]

Describe

I'd like to migrate the project to uv for managing dependencies.
Current setup with pip/pip-tools works, but it's noticeably slower and the ecosystem has moved forward. uv has become pretty much the standard now - it's 10-100x faster, gives us proper lock files, and makes the whole dev experience smoother. Plus, it's from the same folks who made ruff, so it's well maintained.

Specification

Here's what I'm thinking:

Move everything to pyproject.toml for dependencies (if not already there)
Add uv.lock for reproducible builds
Update CI workflows to use uv commands instead of pip
Refresh the docs with new setup instructions
Maybe use uv venv for environment management too

The goal is to make this as smooth as possible without breaking existing workflows. We can do it gradually.
Main benefits: faster CI, better dependency resolution, happier developers, and we stay aligned with where the Python ecosystem is heading.

Dependencies

No known

[linear]

APE-2025

[Syzygy106]

I'm working on it

[fubuloubu]

So, broadly agree with this proposal, a few very specific things we like for our setup:

1. Try to avoid uv-specific features in pyproject.toml. This means taking advantage of dependency-groups (including nesting them) and setting up the build system to not use their built-in one (preferring setuptools instead)
2. Migrating linting (e.g. ruff) using "sensible rules". The closest we've gotten to that is represented here
3. Try to minimize changes as much as possible to make it easy to review by taking things in separate PRs e.g.
   a. Drop Python 3.9 (see Deprecate Python 3.9 #2698)
   b. Add 3.14 support (should be pretty easy and can probably combine with (a))
   c. Migrate to ruff
   d. Migrate CI to use uv (unsure if this needs to be before/after (e))
   e. Migrate project management to uv
4. Upgrade the repos in the following sequence:
   a. Non-ape repos e.g. eth-trace, eth-pydantic-types (we did this already somewhat), eip712 (I'm actually doing this one now), etc.
   b. Ape (this repo) as it's very important cornerstone
   c. Ape plugins (any repo starting w/ ape-*)
   d. SDKs and Projects that rely on Ape (silverback, apepay, uniswap-sdk, etc.)

If we can minimize the changes it will make it easier for me to review, and we can bang these out rather quickly!

---

This is probably the best example PR of how I've done it in the past

[Syzygy106]

So, broadly agree with this proposal, a few very specific things we like for our setup:

1. Try to avoid uv-specific features in pyproject.toml. This means taking advantage of dependency-groups (including nesting them) and setting up the build system to not use their built-in one (preferring setuptools instead)
2. Migrating linting (e.g. ruff) using "sensible rules". The closest we've gotten to that is represented here
3. Try to minimize changes as much as possible to make it easy to review by taking things in separate PRs e.g.
   a. Drop Python 3.9 (see Deprecate Python 3.9 #2698)
   b. Add 3.14 support (should be pretty easy and can probably combine with (a))
   c. Migrate to ruff
   d. Migrate CI to use uv (unsure if this needs to be before/after (e))
   e. Migrate project management to uv
4. Upgrade the repos in the following sequence:
   a. Non-ape repos e.g. eth-trace, eth-pydantic-types (we did this already somewhat), eip712 (I'm actually doing this one now), etc.
   b. Ape (this repo) as it's very important cornerstone
   c. Ape plugins (any repo starting w/ ape-*)
   d. SDKs and Projects that rely on Ape (silverback, apepay, uniswap-sdk, etc.)

If we can minimize the changes it will make it easier for me to review, and we can bang these out rather quickly!

This is probably the best example PR of how I've done it in the past

#2722 (comment)