feat: v0.2.0 — test-driven prompt engineering

[sbroenne]

Summary

Repositions pytest-codingagents as a test-driven prompt engineering platform with two new core features:

New: \�b_run\ fixture

• Takes (baseline, treatment, task)\ → (CopilotResult, CopilotResult)\
• Auto-creates isolated \�aseline/\ and \ reatment/\ subdirs under \ mp_path\
• Overrides \working_directory\ on both agents (prevents shared-workspace bugs)
• Runs sequentially (no parallel — agents may have side effects)
• Stashes treatment result for pytest-aitest HTML reporting

New: \optimize_instruction()\ + \InstructionSuggestion\

• Analyzes gap between current instruction and observed agent behavior
• Uses pydantic-ai structured output (default: \openai:gpt-4o-mini)
• Returns \InstructionSuggestion(instruction, reasoning, changes)\
• Designed to drop into \pytest.fail()\ for actionable failure messages
• Exported from top-level _init_.py\

Dependency

• Added \pydantic-ai>=1.0\ as required dependency

Docs

• README + \docs/index.md\ repositioned as 'test-driven prompt engineering'
• New \docs/how-to/optimize.md\ how-to guide
• \�b_run\ section added to \docs/how-to/ab-testing.md\
• \optimize_instruction\ + \InstructionSuggestion\ added to reference API

Tests

• Added 23 new unit tests (\ est_fixtures.py, \ est_optimizer.py)
• All 129 unit tests pass; ruff + pyright clean

Version

0.1.2 → 0.2.0

[github-actions]

Dependency Review

The following issues were found:

• ✅ 0 vulnerable package(s)
• ✅ 0 package(s) with incompatible licenses
• ✅ 0 package(s) with invalid SPDX license definitions
• ⚠️ 1 package(s) with unknown licenses.

See the Details below.

License Issues

pyproject.toml

Package	Version	License	Issue Type
pydantic-ai	>= 1.0	Null	Unknown License

OpenSSF Scorecard

Package	Version	Score	Details
pip/pydantic-ai	>= 1.0	Unknown	Unknown

Scanned Files

• pyproject.toml