Add Response Model Support for Model Adapters

[cemde]

Description

MASEval's simulators (tool, user, agentic user) relied on hand-rolled JSON extraction
and manual retry loops to coerce LLM outputs into structured formats. This was fragile:
Each simulator reimplemented the same parse-validate-retry logic, and failures from
malformed JSON were a recurring source of flaky benchmark runs.

This PR introduces the instructor library as a
core dependency and threads its response_model support through the entire model adapter
stack. Key changes:

• ModelAdapter.chat() gains a response_model parameter. Pass any Pydantic
  BaseModel class and get a validated instance back in
  ChatResponse.structured_response. All four provider adapters (OpenAI, Anthropic,
  Google GenAI, LiteLLM) implement _structured_chat() using instructor-patched clients.

• Simulators are simplified. LLMSimulator.__call__ now delegates structured output
  handling to instructor via the adapter's response_model support, replacing ~150 lines
  of manual JSON parsing and retry logic with a single chat() call. Each simulator
  declares its expected schema as a class-level _response_model attribute.

• New maseval.core.instructor module provides create_instructor_client() for
  wrapping provider SDK clients, and flatten_model_schema() for generating
  provider-compatible JSON schemas from Pydantic models (used by Tau2's tool parameter
  generation, replacing its own _flatten_schema()).

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Code quality improvement (refactoring, formatting, etc.)

Checklist

Contribution

• I have read the CONTRIBUTING.md guide.
• Commits follow "How to write a good git commit message"

Documentation

• Added/updated docstrings for new/modified functions as instructed CONTRIBUTING.md
• Updated relevant documentation in docs/ (if applicable)
• Tag github issue with this PR (if applicable)

Changelog

• Added entry to CHANGELOG.md under [Unreleased] section
  • Use Added section for new features
  • Use Changed section for modifications to existing functionality
  • Use Fixed section for bug fixes
  • Use Removed section for deprecated/removed features
• OR this is a documentation-only change (no changelog needed)

Example:
- Support for multi-agent tracing (PR:#123)

Architecture (if applicable)

• Core/Interface separation: Changes in maseval/core/ do NOT import from maseval/interface/
• Dependencies: New core dependencies added sparingly; framework integrations go to optional dependencies

Additional Notes

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
maseval
__init__.py
maseval/benchmark/tau2
tau2.py
maseval/benchmark/tau2/domains
base.py
maseval/core
instructor.py					89
model.py
simulator.py					224
maseval/interface/inference
anthropic.py					388-390
google_genai.py					334-336
litellm.py					220-221
openai.py					304-306
Project Total

This report was generated by python-coverage-comment-action