[stobrien89]

Issue #, if available:
#797

Description

Adds native generation of lazy runtime models and typed model stubs for standalone types packages.

This PR is stacked on #857.

Changes

• Emits models.py and models.pyi.
• Supports structures, errors, unions, enums, defaults, and service renames.
• Resolves cross-namespace targets using full shape IDs.
• Preserves lazy construction, serde behavior, star imports, dir(), and pickling.
• Adds Ruff and strict Pyright validation for generated models.

Follow-up work

• Integrate schema and model generation into the types CLI pipeline.
• Generate package metadata and supporting files.
• Add Java/native parity and large-model tests.
• Add schema and stub partitioning based on performance measurements.

Compatibility

Existing ModelRegistry callers remain supported. New runtime arguments are optional and backward compatible.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

5 open issues

• sys.stdlib_module_names is only available on newer Python versions (introduced in Python 3.10).… · New
• Import grouping depends on stdlib_module_names, which can differ across Python versions/builds,… · New
• These tests hard-require pyright to be installed and runnable as python -m pyright. If the test… · New
• Similarly, this test hard-requires ruff to be installed as python -m ruff and assumes a stable… · New
• This assertion is tightly coupled to the multi-line parenthesized from ... import (...)… · New

What changed in this PR

Adds native generation for lazy runtime Smithy models (models.py) alongside full typing stubs (models.pyi), and updates import rendering to produce grouped/combined imports to support the new generated modules.

Changes:

• Add ModelGenerator to emit lazy runtime model modules and .pyi stubs, including validation of reserved member names.
• Update PythonWriter to render grouped/combined imports and adjust formatting for stub files.
• Extend smithy-core ModelRegistry to support full shape-id target mappings for unambiguous renamed/cross-namespace target resolution.

File	Description
packages/​smithy-python/​tests/​unit/​test_writer.py	Updates import-format assertions and adds coverage for grouped/combined imports.
packages/​smithy-python/​tests/​unit/​test_models.py	Adds end-to-end tests for generated runtime/stub models, plus Ruff/Pyright validation.
packages/​smithy-python/​src/​smithy_python/​writer.py	Implements grouped/combined import rendering and stub-specific spacing.
packages/​smithy-python/​src/​smithy_python/​models.py	Introduces model/stub generation for lazy runtime models.
packages/​smithy-python/​src/​smithy_python/​generation.py	Extends ModelPlan to include registry metadata used by lazy model generation.
packages/​smithy-python/​src/​smithy_python/​files.py	Marks .pyi writers as stubs to adjust formatting behavior.
packages/​smithy-python/​.changes/​next-release/​smithy-python-feature-model-generation.json	Adds changelog entry for model generation feature.
packages/​smithy-core/​tests/​unit/​test_dynamic_models.py	Adds tests for copied name maps and full shape-id target resolution.
packages/​smithy-core/​src/​smithy_core/​_dynamic_models.py	Adds target_names support and improves rename handling during build/target resolution.
packages/​smithy-core/​.changes/​next-release/​smithy-core-enhancement-model-target-names.json	Adds changelog entry for ModelRegistry target-name enhancement.

_{Review effort: Lite}

[Copilot]

🟡 Changes recommended

Unresolved runtime/stub mismatches, public-name collisions, and class-cache replacement can break generated model APIs and serialization.

6 open issues

• Preserve canonical class identity across alias lookups · New
• Prevent synthetic ServiceError name collisions · New
• Prevent generated models from shadowing runtime helpers · New
• Handle explicit input and output traits in stub classification · New
• Align stub and runtime member naming rules · New
• Support Unit union members consistently at runtime · New

5 resolved since last review

• sys.stdlib_module_names is only available on newer Python versions (introduced in Python 3.10).…
• This assertion is tightly coupled to the multi-line parenthesized from ... import (...)…
• Similarly, this test hard-requires ruff to be installed as python -m ruff and assumes a stable…
• These tests hard-require pyright to be installed and runnable as python -m pyright. If the test…
• Import grouping depends on stdlib_module_names, which can differ across Python versions/builds,…

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Unresolved naming regressions, generated-name collisions, and runtime/stub mismatches can break valid models and existing generated clients.

5 open issues

• Preserve leading underscores for digit-prefixed member names · New
• Reserve runtime serialization method names from model members · New
• Escape modeled Unit to avoid collision with prelude Unit · New
• Preserve reserved-name escaping before uppercasing enum members · New
• Align unknown variant stub methods with runtime implementation · New

6 resolved since last review

• Support Unit union members consistently at runtime
• Align stub and runtime member naming rules
• Handle explicit input and output traits in stub classification
• Prevent generated models from shadowing runtime helpers
• Prevent synthetic ServiceError name collisions
• Preserve canonical class identity across alias lookups

_{Review effort: Balanced}

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

No open issues 🎉

5 resolved since last review

• Escape modeled Unit to avoid collision with prelude Unit
• Reserve runtime serialization method names from model members
• Preserve leading underscores for digit-prefixed member names
• Align unknown variant stub methods with runtime implementation
• Preserve reserved-name escaping before uppercasing enum members

_{Review effort: Lite}

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

4 open issues

• Emitting docstrings as a single JSON-escaped string literal can easily produce very long physical… · New
• Within each section, module iteration relies on insertion order (sections[section].items()).… · New
• The 88 line-length cutoff is a hard-coded formatting policy that can drift from the repo’s… · New
• This assertion is now broad enough to allow false positives (e.g., matching an unrelated substring… · New

_{Review effort: Lite}

[Copilot]

🟡 Changes recommended

Renamed union lookups regress for existing callers, and generated error stubs disagree with runtime message signatures and nullability.

2 open issues

• Preserve original variant names and union deserializer aliases · New
• Align modeled error message stubs with runtime constructor and nullability · New

4 resolved since last review

• The 88 line-length cutoff is a hard-coded formatting policy that can drift from the repo’s…
• Within each section, module iteration relies on insertion order (sections[section].items()).…
• Emitting docstrings as a single JSON-escaped string literal can easily produce very long physical…
• This assertion is now broad enough to allow false positives (e.g., matching an unrelated substring…

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Unresolved model-identity collisions and sensitive-message exposure need correction before approval.

2 open issues

• Hide sensitive modeled messages from error repr output · New
• Prevent renamed union aliases from overwriting another namespace's model · New

2 resolved since last review

• Preserve original variant names and union deserializer aliases
• Align modeled error message stubs with runtime constructor and nullability

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Unresolved error classification, alias-cache collisions, naming compatibility, and stub nullability issues affect runtime correctness and the generated API.

4 open issues

• Prevent original-name aliases from overwriting renamed models · New
• Apply error trait fault when building native error classes · New
• Preserve legacy field-name normalization for existing models · New
• Separate field nullability from constructor default handling · New

2 resolved since last review

• Prevent renamed union aliases from overwriting another namespace's model
• Hide sensitive modeled messages from error repr output

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Unresolved class-cache collisions and runtime/stub naming and export mismatches affect supported model APIs.

4 open issues

• Preserve explicit index names for non-union cache keys · New
• Escape annotations to avoid dataclass field collisions · New
• Include underscore-prefixed models in stub all · New
• Escape double-leading-underscore member names consistently · New

1 resolved since last review

• Reserve _smithy_default and repr from generated fields

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Shared runtime and cross-generator naming changes require human compatibility review, and a confirmed runtime correctness issue remains unresolved.

1 open issue

• Escape enum names with leading and trailing underscores · New

4 resolved since last review

• Escape annotations to avoid dataclass field collisions
• Preserve explicit index names for non-union cache keys
• Escape double-leading-underscore member names consistently
• Include underscore-prefixed models in stub all

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Shared runtime and Java naming changes need human compatibility review, with a blocking enum-naming defect still unresolved.

1 open issue

• Escape enum dunder names to preserve members and deserialization · New

1 resolved since last review

• Escape enum names with leading and trailing underscores

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Shared runtime naming and alias-resolution changes affect both generators, and full Java/native compatibility remains unverified.

No open issues 🎉

1 resolved since last review

• Escape enum dunder names to preserve members and deserialization

_{Review effort: Balanced}