feat(agent): add modelSettings to agent settings [PC-4672]

[tudormatei1]

Add a modelSettings block to settings in agent.json, parsed into AgentSettings.model_settings. It's a native bag (ModelSettings(BaseCfg), no declared fields); keys are the target model's own parameter names and are forwarded verbatim, so discovery stays the source of truth for the shape. Absent modelSettings parses to None, keeping existing agent.json working.

[Copilot]

Pull request overview

Adds support in the UiPath Python SDK’s agent definition models for a provider-native settings.modelSettings block (a free-form “bag” of model parameters) so agent.json can carry model-specific tuning options without the SDK needing a fixed schema.

Changes:

• Introduce ModelSettings and add AgentSettings.model_settings (aliased to modelSettings) to preserve arbitrary provider parameters.
• Add unit tests validating modelSettings is optional and round-trips verbatim via alias-based dump.
• Bump uipath package version to 2.15.0 and update the lockfile accordingly.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated no comments.

File	Description
packages/uipath/src/uipath/agent/models/agent.py	Adds ModelSettings bag model and wires modelSettings into AgentSettings.
packages/uipath/tests/agent/models/test_agent.py	Adds coverage for absent/present modelSettings behavior and verbatim round-trip.
packages/uipath/pyproject.toml	Bumps package version to 2.15.0.
packages/uipath/uv.lock	Updates locked uipath version entry to 2.15.0.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (3)

packages/uipath/src/uipath/agent/models/agent.py:1427

• ModelSettings is an empty Pydantic model, which makes the public API for a 'native bag' ambiguous: consumers will get a ModelSettings instance (not a dict), and whether arbitrary keys are accepted/preserved depends implicitly on BaseCfg's extra behavior. To make the passthrough contract explicit and stable, consider modeling this as Optional[dict[str, Any]] (most direct) or as a RootModel[dict[str, Any]] / explicitly configuring ModelSettings to allow and preserve extra keys (so future BaseCfg changes won't break this).

class ModelSettings(BaseCfg):
    """Provider-native model settings bag (``settings.modelSettings``).

    Keys are the target model's own parameter names, forwarded verbatim — no fixed
    schema; discovery is the source of truth for the shape.
    """

packages/uipath/src/uipath/agent/models/agent.py:1438

• ModelSettings is an empty Pydantic model, which makes the public API for a 'native bag' ambiguous: consumers will get a ModelSettings instance (not a dict), and whether arbitrary keys are accepted/preserved depends implicitly on BaseCfg's extra behavior. To make the passthrough contract explicit and stable, consider modeling this as Optional[dict[str, Any]] (most direct) or as a RootModel[dict[str, Any]] / explicitly configuring ModelSettings to allow and preserve extra keys (so future BaseCfg changes won't break this).

    model_settings: Optional[ModelSettings] = Field(None, alias="modelSettings")

packages/uipath/tests/agent/models/test_agent.py:4780

• The tests cover non-empty passthrough via alias, but they don't cover the edge case of an explicitly empty bag. Adding a case asserting that modelSettings={} round-trips to {} (and is not coerced to None / omitted) would better lock in the 'verbatim bag' behavior.

    def test_native_bag_survives_verbatim_by_alias(self):
        native = {
            "thinking": {"type": "enabled", "budget_tokens": 2048},
            "output_config": {"effort": "high"},
        }
        settings = AgentSettings.model_validate(
            self._agent_settings(modelSettings=native)
        )
        assert settings.model_dump(by_alias=True)["modelSettings"] == native

[github-actions]

🚨 Heads up: uipath-langchain cross-tests are FAILING 🚨

Your changes may break the uipath-langchain-python integration.

⚠️ These checks are NOT enforced by branch protection rules. Please review the failures before merging.

🔍 Inspect the failed run →

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud