feat(conformance): convert express compiler tests to conformance tests (#2684)

[jacobsimionato]

Summary & Rationale

Converted the A2UI Express compiler, decompiler, round-trip, and prompt generation tests from programmatic Python-specific unit tests into platform-agnostic conformance test suites. This enables future non-Python implementations (such as TypeScript or Kotlin) to achieve full behavioral and semantic parity against a shared specification.

Key Changes

• Conformance Schema: Added conformance/inference_formats/inference_format_schema.json supporting compile, decompile, round_trip, and generate_prompt actions with inlined or file-based inputs and expected outputs, as well as ExpectError negative test verification.
• Format-Independent Round-Trip Suite: Added conformance/inference_formats/round_trip.yaml covering catalog examples and dynamic data model/component compositions across decompile -> compile cycles.
• Express Test Suites:
  • conformance/inference_formats/express/compile.yaml: 30+ test cases covering basic components, data bindings, checks, format strings, actions, standalone functions, templates, surface directives, version targets (v0.9, v0.9.1, v1.0), and negative error scenarios.
  • conformance/inference_formats/express/decompile.yaml: Decompilation tests for RPC functions, local actions, functional expressions, custom checks, surface directives, update data model, and update components.
  • conformance/inference_formats/express/prompt_generation.yaml: Prompt generation test cases for rules, component signatures, pruning, custom rules, and examples transformation.
  • conformance/inference_formats/express/specification_examples.yaml: Conformance tests for all 36 human-authored .a2ui examples mapped against their JSON counterparts.
• Schema Validation: Updated conformance/tests/test_conformance_yaml.py to validate conformance/inference_formats/**/*.yaml against inference_format_schema.json.
• Python SDK Test Harness: Implemented agent_sdks/python/a2ui_agent/tests/conformance/test_inference_formats_conformance.py which executes all inference format conformance test cases.
• Test Streamlining: Removed redundant programmatic tests from test_compiler.py, test_integration.py, test_parser_decompile.py, and removed test_version_compliance.py, keeping only Python-specific unit and concurrency tests.
• Documentation: Updated conformance/README.md with descriptions and usage instructions for the new inference format test suites.

Addressed Review Feedback

• Feedback Item 1: Updated normalize_value in test_inference_formats_conformance.py to construct and return a new dictionary rather than mutating the input dictionary in-place.
• Feedback Items 2 & 3: Retained standard validation checks per review consensus, ensuring malformed payloads without required dictionary structures fail as intended.
• Feedback Item 4: Renamed conformance/inference_formats/express/examples.yaml to specification_examples.yaml and updated all test suite references and conformance/README.md.
• Feedback Item 5: Removed historical migration comments/placeholders from test_integration.py, test_compiler.py, and test_parser_decompile.py.
• Feedback Item 6: Retained remaining tests in test_integration.py as they verify Python SDK-specific API surface (ExpressParser.parse_response, ParsedPart objects, streaming unclosed <a2ui> tags, and A2uiCompilationError exception structures).
• Upstream Sync: Synced branch with upstream/main, resolved merge conflicts in test_compiler.py and test_parser_decompile.py, and updated test harness exception mappings to align with the new A2uiCompilationError hierarchy.

Verification & Testing

• Validated all YAML test files against schemas:
  ./bin/pytest conformance/tests/test_conformance_yaml.py (30/30 passed).
• Ran the new Python conformance test harness:
  PYTHONPATH="agent_sdks/python/a2ui_core/src:agent_sdks/python/a2ui_agent/src" ./bin/pytest agent_sdks/python/a2ui_agent/tests/conformance/test_inference_formats_conformance.py (99/99 passed).
• Ran all Express unit and parser tests:
  PYTHONPATH="agent_sdks/python/a2ui_core/src:agent_sdks/python/a2ui_agent/src" ./bin/pytest agent_sdks/python/a2ui_agent/tests/express/ (57/57 passed).
• Ran standard conformance suite:
  PYTHONPATH="agent_sdks/python/a2ui_core/src:agent_sdks/python/a2ui_agent/src" ./bin/pytest agent_sdks/python/a2ui_agent/tests/conformance/test_conformance.py (254 passed, 31 xfailed).

Closes #2684

TAG=agy
CONV=758b56c3-d6dc-4dc6-96b5-1a0dfa903bac

[gemini-code-assist]

Code Review

This pull request migrates various Express compiler, parser, and integration tests into language-agnostic YAML conformance test suites under conformance/inference_formats/, and introduces a new Python conformance test harness (test_inference_formats_conformance.py) to run them. The feedback highlights potential runtime issues in the test harness: first, the normalize_value helper mutates input dictionaries in-place, which can cause unexpected side-effects across tests; second, defensive checks are needed when calling .get() on keys that might explicitly map to None (such as updateDataModel or createSurface) to prevent AttributeError crashes.

[jacobsimionato]

Synced with upstream/main, resolved conflicts in test_compiler.py and test_parser_decompile.py, aligned exceptions with the new A2uiCompilationError hierarchy, verified all local tests pass, and pushed updates.

[jacobsimionato]

One more comment: I think the inference format conformance test YAML etc should be within the agent/ subfolder which I think already exists. So something like conformance/agent/inference_formats etc.