fix: mcp parameter fields are being dropped

[NicoleMGomes]

Disclaimer: Do not include SAP-internal or customer-specific information in this PR (e.g. internal system URLs, customer names, tenant IDs, or confidential configurations). This is a public repository.

Description

mcp_tool_to_langchain in sap_cloud_sdk.agentgateway.converters was silently dropping all JSON Schema property metadata other than type and required. This meant fields like description, enum, format, examples, minimum/maximum, pattern, and others were never forwarded to the Pydantic model or the LangChain tool schema — so the LLM had no visibility into allowed values, constraints, or hints.

The fix introduces a two-bucket mapping strategy:

• Native Pydantic Field kwargs (description, title, examples, deprecated, pattern, minLength/maxLength → min_length/max_length, minimum/maximum → ge/le, exclusiveMinimum/exclusiveMaximum → gt/lt, multipleOf) are forwarded directly as Field(...) keyword arguments with the required camelCase → snake_case translation. These are also enforced at Pydantic validation time.
• Schema-only fields (enum, default, example, const, format, contentEncoding, uniqueItems, items, properties, required, additionalProperties, oneOf, anyOf, allOf) are collected into json_schema_extra so they pass through verbatim into model_json_schema() and reach the LLM.
• Unknown keys (future extensions, x-custom-*) are silently ignored — no crash.

The full set of fields emitted by the MCP builder (type, description, title, enum, default, examples, example, const, maxLength, minimum, pattern, format, contentEncoding, uniqueItems, items, properties, additionalProperties, oneOf/anyOf/allOf) are now all handled.

Related Issue

Closes #

Type of Change

• Bug fix (non-breaking change that fixes an issue)

How to Test

1. Install the package: pip install sap-cloud-sdk[langchain]
2. Construct an MCPTool with a rich input_schema — include description, enum, format, minimum, maxLength, and examples on properties.
3. Call mcp_tool_to_langchain(tool, call_tool, get_user_token).
4. Inspect lc_tool.args_schema.model_json_schema()["properties"] — all metadata fields should be present on each property.
5. Verify Pydantic enforces native constraints: pass a value exceeding maxLength or below minimum and confirm a ValidationError is raised.
6. Run the unit tests: uv run pytest tests/agentgateway/unit/test_converters.py -v

Checklist

• I have read the Contributing Guidelines
• I have verified that my changes solve the issue
• I have added/updated automated tests to cover my changes
• All tests pass locally
• I have verified that my code follows the Code Guidelines
• I have updated documentation (if applicable)
• I have added type hints for all public APIs
• My code does not contain sensitive information (credentials, tokens, etc.)
• I have followed Conventional Commits for commit messages

Additional Notes

• enum is intentionally forwarded via json_schema_extra rather than as a Literal type — it serves as a hint to the LLM, not a Python validation constraint. This matches how MCP tool schemas are intended to be used.
• default from the MCP schema is also in json_schema_extra (not Field(default=...)), because MCP's default is a documentation hint; promoting it to a Python default would silently make required fields optional.
• The examples field is native to Pydantic v2's Field and goes through the native path.
• 45 unit tests cover all paths: each native kwarg, each extra-bucket key, coexistence of multiple fields, unknown key handling, and all pre-existing behaviour (required/optional, nullable, OData underscore stripping, omit_none, invocation forwarding).