Python migration to CondaPkg.jl and PythonCall.jl + claude skill python-dep-manager

[odunbar]

Purpose

• for v1.1, automerge broke as conda tried to default install 3.13.
  • In trying to consolidate the management it lead to further breakages and i discoverd that an much cleaner system appears to be moving from Conda+PyCall to CondaPkg+PythonCall. Here, we don't need dependencies on the Julia SciKitLearn.jl and we can use memory (in the form of a toml file) to store the python versioning, rather than having to call Conda.add( package = "x.y.z") in all the differnt CI piplines with pycall builds
  • I created a skill to maintain this setting in future.
• Closes PDMats inappropriate patch -> breaks GaussianProcesses.jl -> Breaks CES.jl #432

Content

• remove Conda/PyCall, add CondaPkg/PythonCall
• add CondaPkg.toml (which will hold the conda versions fixed to their current working status)
• add .claude/skills/python-dep-manager/SKILL.md
• As now we can call python directly with pythoncall, we use SciKitLearn from python not the wrapping package SciKitLearn.jl. Therefore we also migrate SKLJL->SKLPy with a deprecation/migration notice for users.
• installation docs have been updated,
• buildkite and github actions are now much cleaner
• docs here

MISC

• Pin PDMats to 0.11.36 to stop issues with GP ldiv! ambiguity.
• Another random Manopt error occured. as we double until output_dim it seems like the latest manopt does not handle this, so added an exit case if k==output_dim, though still tries for k==output_dim-1
• docs deployment wasn't tagged to main, updated devbranch=main in docs/make.jl

Example output

During tests we see the standard build output:

...
CalibrateEmulateSample
│      CondaPkg Found dependencies: /home/odunbar/.julia/packages/CondaPkg/lKlVY/CondaPkg.toml
│      CondaPkg Found dependencies: /home/odunbar/.julia/packages/PythonCall/JksWe/CondaPkg.toml
│      CondaPkg Found dev dependencies: /home/odunbar/Dropbox/Caltech/CESjl/tmp/CalibrateEmulateSample.jl/CondaPkg.toml
│      CondaPkg Resolving changes
│               + libstdcxx
│               + libstdcxx-ng
│               + openssl
│               + python
│               + scikit-learn
│               + scipy
│      CondaPkg Initialising pixi
│               │ /home/odunbar/.julia/artifacts/4ae17cee0bd922f66b2c72bf2f01c22481a5ec19/bin/pixi
│               │ init
│               │ --format pixi
│               └ /tmp/jl_XcgVeG/.CondaPkg
│  ✔ Created /tmp/jl_XcgVeG/.CondaPkg/pixi.toml
│      CondaPkg Wrote /tmp/jl_XcgVeG/.CondaPkg/pixi.toml
│               │ [dependencies]
│               │ scikit-learn = "=1.5.1"
│               │ openssl = ">=3, <3.6"
│               │ libstdcxx = ">=3.4,<15.0"
│               │ libstdcxx-ng = ">=3.4,<15.0"
│               │ scipy = "=1.14.1"
│               │ 
│               │     [dependencies.python]
│               │     channel = "conda-forge"
│               │     build = "*cp*"
│               │     version = ">=3.10,!=3.14.0,!=3.14.1,<4, =3.11"
│               │ 
│               │ [workspace]
│               │ name = ".CondaPkg"
│               │ platforms = ["linux-64"]
│               │ channels = ["conda-forge"]
│               │ channel-priority = "strict"
│               └ description = "automatically generated by CondaPkg.jl"
│      CondaPkg Installing packages
│               │ /home/odunbar/.julia/artifacts/4ae17cee0bd922f66b2c72bf2f01c22481a5ec19/bin/pixi
│               │ install
│               └ --manifest-path /tmp/jl_XcgVeG/.CondaPkg/pixi.toml
│  ✔ The default environment has been installed.

---

• I have read and checked the items on the review checklist.

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.97%. Comparing base (74f8e44) to head (7444439).

Files with missing lines	Patch %	Lines
src/Utilities/likelihood_informed.jl	40.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #431      +/-   ##
==========================================
- Coverage   95.12%   94.97%   -0.15%     
==========================================
  Files          12       12              
  Lines        2400     2408       +8     
==========================================
+ Hits         2283     2287       +4     
- Misses        117      121       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[odunbar]

Merging as tests resolved and documentation building with new Pyhton environments