Error when using Map-based schema

[habuma]

Following the example set in the sample application (at https://github.com/spring-projects-experimental/spring-ai-mcp/blob/main/spring-ai-mcp-sample/src/main/java/org/springframework/ai/mcp/sample/server/McpServerConfig.java#L135), I'm trying to specify the input schema with a Map. Specifically, if I specify the schema with Map.of("entityId", "String"), I get the following error:

{
  "error": {
    "message": "Invalid schema for function 'entity-live': schema must be a JSON Schema of 'type: \"object\"', got 'type: \"None\"'.",
    "type": "invalid_request_error",
    "param": "tools[1].function.parameters",
    "code": "invalid_function_parameters"
  }
}

However, it works fine if I specify the schema with a String, like this:

    private static final String COMMON_INPUT_SCHEMA = """
                                {
                                    "type": "object",
                                    "properties": {
                                        "entityId": {
                                            "type": "string",
                                            "description": "Entity ID"
                                        }
                                    },
                                    "required": ["entityId"]
                                }
                                """;
...
    private McpServer.ToolRegistration themeParkApiEntityScheduleToolRegistration() {
        return new McpServer.ToolRegistration(
                new McpSchema.Tool(
                        "entity-schedule",
                        "Return schedule data about an entity, including hours of operation",
                        COMMON_INPUT_SCHEMA),
                (arguments) -> {
                    String entityId = (String) arguments.get("entityId");
                    var entityScheduleJson = themeParkService.getEntitySchedule(entityId);
                    McpSchema.TextContent content = new McpSchema.TextContent(entityScheduleJson);
                    return new McpSchema.CallToolResult(List.of(content), false);
                });
    }

(The entire project can be found at https://github.com/habuma/spring-ai-examples/tree/main/spring-ai-mcp)

HOWEVER, I can get it to work with a Map-based schema is the Map is defined like this:

Map.of("type", "object", "properties",
                    Map.of("entityId",
                            Map.of("type", "string", "description", "Entity ID")), "required",
                    List.of("entityId"))

Which is, incidentally, the same Map that is produced internally when given the String schema above.

So, my take on this is that one of the two things is true:

• The Spring AI MCP sample is incorrectly showing how to specify a Map-based schema.
• The McpSchema.Tool should somehow wrap the incoming Map with the outer Map as the value of the "properties" key.