Fix: Default empty object schema parameters to include properties: {} for OpenAI compatibility

[pedro-inf-custodio]

Description

This PR fixes an openai.BadRequestError that occurs when an MCP tool provides an inputSchema with "type": "object" but omits the required "properties" key.

Problem

When using langchain-mcp-adapter with an OpenAI model (like Azure OpenAI) in LangGraph, any MCP tool that takes no arguments but defines its schema as just "type": "object" will fail validation.

For example, the ecs-mcp-server tool aws_knowledge_aws___list_regions provides this schema:

"parameters": {
  "type": "object"
}

This causes the OpenAI API to return a 400 error, as it strictly requires a properties key for any object schema:

openai.BadRequestError: Error code: 400 - {'error': {'message': "Invalid schema for function 'ecs_mcp_server_aws_knowledge_aws___list_regions': In context=(), object schema missing properties.", 'type': 'invalid_request_error', 'param': 'tools[4].function.parameters', 'code': 'invalid_function_parameters'}}

Solution

This patch updates the tool conversion logic within the adapter. It checks if an inputSchema's parameters are of "type": "object" and lack a "properties" key. If so, it automatically adds an empty "properties": {} field.

This ensures that a schema like:
"parameters": { "type": "object" }

Is correctly converted to the OpenAI-compliant schema:
"parameters": { "type": "object", "properties": {} }

This change resolves the BadRequestError and allows tools with no arguments to be used successfully with OpenAI models.