feat: add Helix integration for optimize_anything

[LakshyAAAgrawal]

Summary

• Add gepa.adapters.helix module with two helpers:
  • make_helix_evaluator(config, target_file, ...) — wraps Helix's run_evaluator as an optimize_anything-compatible evaluator
  • make_helix_batch_evaluator(config, target_file, ...) — wraps Helix's run_evaluators_parallel as a batch_evaluator
• Add helix-evo>=0.1.0 as optional dependency under gepa[helix] (Python >= 3.11)

Usage

from gepa.adapters.helix import make_helix_evaluator
from gepa.optimize_anything import optimize_anything

evaluator = make_helix_evaluator(
    config="helix.toml",
    target_file="src/solver.py",
)

result = optimize_anything(
    seed_candidate=open("src/solver.py").read(),
    evaluator=evaluator,
    objective="Maximize accuracy on the test suite.",
)

cc @KE7

Test plan

• Existing test suite passes (446 passed, 2 skipped)
• ruff check passes
• Manual integration test with a Helix-configured repo

🤖 Generated with Claude Code

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
pyproject.toml	Unsupported file format
pyrightconfig.json	0% smaller
src/gepa/adapters/helix.py	0% smaller

[KE7]

This seems reasonable to me. shutil.copytree works especially since the text artifact will be a single file. git worktrees would have been overkill