feat(plugin-api): export the interactive prompt content models

[DABH]

Description

Third-party plugin packages are asked to import plugin-authoring symbols only from the stable nat.plugin_api facade (per docs/source/extend/third-party-plugins.md). PR #2113 exported the interactive response side of that surface: InteractionPrompt, InteractionResponse, and the HumanResponse union with its concrete response models. The prompt side is still missing: the concrete prompt content models behind the HumanPrompt union are only importable from the implementation module nat.data_models.interactive.

Plugin authors do not only consume responses. HITL middleware hooks, user-input callbacks, and front-end integrations construct prompts: any plugin that calls Context.get().user_interaction_manager.prompt_user_input(...) first has to build a HumanPromptText (or a notification, binary, radio, checkbox, or dropdown variant). The in-repo HITL examples import HumanPromptText from nat.data_models.interactive, and the interactive workflows guide constructs it while pointing readers at that same internal module path — exactly the dependency an external package following the facade contract is told not to take. In addition, the binary and multiple-choice prompt models embed required option models, and those same option models are the field types of the already-exported HumanResponseBinary, HumanResponseRadio, HumanResponseCheckbox, and HumanResponseDropdown (selected_option), so typed construction of the exported response models also needs them.

This change is purely additive and completes the prompt-and-response symmetry of the interactive surface:

• Re-export the concrete prompt content models HumanPromptText, HumanPromptNotification, HumanPromptBinary, HumanPromptRadio, HumanPromptCheckbox, and HumanPromptDropdown, together with the embedded option models BinaryHumanPromptOption and MultipleChoiceOption, through nat.plugin_api, and extend __all__.
• Pin the new symbols in EXPECTED_PLUGIN_API_EXPORTS and extend the denied-import patterns in the plugin-authoring docs test (the existing HumanPrompt pattern already covers the HumanPrompt* model imports as substrings, so only the two option-model patterns are new).
• Add a consumer-style test that constructs every prompt variant through nat.plugin_api imports alone and round-trips each one through InteractionPrompt to check the discriminated union resolves the concrete types.
• Document the new surface in docs/source/extend/plugin-api.md, extending the interactive-models bullet and the "HITL middleware" surface-review row (the models follow that row's existing provisional public, trusted plugin status).

Deliberately not exported: the OAuth consent prompt variant stays private (it is underscore-prefixed by design), and the abstract bases (HumanPromptBase, HumanPromptMultipleChoiceBase) stay internal, mirroring how InteractionBase is not exported; the already-exported HumanPrompt union covers the "any prompt" typing need.

Migrating the in-repo consumers (the two HITL examples and the interactive workflows guide) to the facade imports is deliberately left as a follow-up so this change stays purely additive. No tracking issue exists for this yet; happy to file one — also covering that migration — if the team prefers.

Testing

• uv run pytest packages/nvidia_nat_core/tests/nat/test_plugin_api.py — 8 passed (includes the new consumer-style prompt-construction test).
• uv run pytest packages/nvidia_nat_core/tests/nat/builder packages/nvidia_nat_core/tests/nat/middleware packages/nvidia_nat_core/tests/nat/runtime packages/nvidia_nat_core/tests/nat/data_models — 852 passed.
• uv run pre-commit run yapf --files <touched files> and uv run pre-commit run ruff-check --files <touched files> — passed.
• uv run python ci/scripts/copyright.py --verify-apache-v2 — passed.
• uv run python ci/scripts/path_checks.py — passed.
• uv run vale docs/source/extend/plugin-api.md — 0 errors, 0 warnings, 0 suggestions.

By Submitting this PR I confirm:

• I am familiar with the Contributing Guidelines.
• We require that all contributors "sign-off" on their commits. This certifies that the contribution is your original work, or you have rights to submit it under the same license, or a compatible license.
  • Any contribution which contains commits that are not Signed-Off will not be accepted.
• When the PR is ready for review, new or existing tests cover these changes.
• When the PR is ready for review, the documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: c3498629-4a84-4aa8-a63e-7df065cb219b

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}