Tool Registration with Input and Output schema overrides

[csiemback]

Hello - I'm currently writing a tool using the @mcptool annotation. However the generated input and output schemas do not meet the client's required definition. Specifically - allowing the type to be both required AND nullable. The annotation generation only defines it as a string but I need it to allow null. Example below:

{
   "type":"object",
   "properties":{
      "person":{
         "type":[ "string", "null"  ], <== what I want | what's getting generated ==> "type":"string",
         "description":"Person to send to"
      }
   },
   "required":[
      "person"
   ]
}

I don't think this is possible using the McpTool - but I cannot for the life of me figure out how to define the Tool programmatically while also programmatically defining both the input and output schemas. Totally willing to have both schemas hard coded. Is this even possible? I'd be absolutely surprised if spring ai didn't provide some method of self-registration like this.

Side note - I'm using version 1.1.0 of spring as - but this is greenfield so I'd be fine upgrading to spring boot version 4 and spring ai version 2.0 (I know it's in M1 status).

please help!!

PS: I need some non-ai generated results...I've exhausted that line of searching :)

[myifeng]

"null" is String.

[csiemback]

I'm afraid I don't understand your response. According to the json schema specification - the type can be marked as "type":[ "string", "null" ] meaning the value adhering to the schema must be String OR null. This is a validation rule - not an instance value here.