fix: preserve typed array items in nullable anyOf schemas

[4444J99]

Problem

When OpenAPI represents optional typed parameters (e.g. Optional[List[int]]), it produces an anyOf schema:

{
  "anyOf": [
    {"type": "array", "items": {"type": "integer"}},
    {"type": "null"}
  ]
}

get_single_param_type_from_schema correctly resolves the type to "array" and injects it at the top level, but the items definition remains buried inside the anyOf variant. The resulting MCP tool schema looks like:

{
  "anyOf": [
    {"type": "array", "items": {"type": "integer"}},
    {"type": "null"}
  ],
  "type": "array",
  "title": "time_values"
}

LLM clients that validate tool inputSchema against JSON Schema (such as gpt-4o) reject this because type: "array" at the top level lacks a corresponding items definition.

Solution

Add flatten_nullable_anyof() to openapi/utils.py. When a schema contains a simple nullable anyOf (one non-null type + null), it hoists all type-specific fields (items, properties, required, etc.) from the non-null variant to the top level and removes the anyOf key.

Called after type injection for both query parameters and body parameters in convert.py.

Before

{"anyOf": [{"type": "array", "items": {"type": "integer"}}, {"type": "null"}], "type": "array", "title": "time_values"}

After

{"type": "array", "items": {"type": "integer"}, "title": "time_values"}

Scope

• Only flattens simple nullable patterns (exactly one non-null type + null). Complex multi-type unions are left unchanged.
• Handles arrays (items), objects (properties, required, additionalProperties), and any other fields on the non-null variant (e.g. format, enum, minItems, maxItems).
• No changes to non-nullable schemas.

Tests

6 new tests covering:

• flatten_nullable_anyof utility: array items hoisting, no-op on non-nullable, skip on multi-type unions, object properties hoisting
• End-to-end conversion: optional typed arrays in body params and query params

All 123 tests pass (117 existing + 6 new). Coverage 83.29%.

Fixes #57

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.