Help small LLMs with MCP Tool Calling

[nktltvnv]

Spring AI version: 1.0.0-M8.
LLM: Llama3.2 (3b) running locally with Ollama.

I'm facing an issue with tool calling payload generated by the model. The function input is:

{
  "branchFromThought": "null",
  "branchId": "null",
  "isRevision": "false",
  "needsMoreThoughts": "false",
  "nextThoughtNeeded": "false",
  "revisesThought": "null",
  "thought": "Отрезанный угол",
  "thoughtNumber": "1",
  "totalThoughts": "3"
}

I've connected to the @modelcontextprotocol/server-sequential-thinking which lists following tools:

{
  "result": {
    "tools": [
      {
        "name": "sequentialthinking",
        "description": "A detailed tool for dynamic ...",
        "inputSchema": {
          "type": "object",
          "properties": {
            "thought": {
              "type": "string",
              "description": "Your current thinking step"
            },
            "nextThoughtNeeded": {
              "type": "boolean",
              "description": "Whether another thought step is needed"
            },
            "thoughtNumber": {
              "type": "integer",
              "description": "Current thought number",
              "minimum": 1
            },
            "totalThoughts": {
              "type": "integer",
              "description": "Estimated total thoughts needed",
              "minimum": 1
            },
            "isRevision": {
              "type": "boolean",
              "description": "Whether this revises previous thinking"
            },
            "revisesThought": {
              "type": "integer",
              "description": "Which thought is being reconsidered",
              "minimum": 1
            },
            "branchFromThought": {
              "type": "integer",
              "description": "Branching point thought number",
              "minimum": 1
            },
            "branchId": {
              "type": "string",
              "description": "Branch identifier"
            },
            "needsMoreThoughts": {
              "type": "boolean",
              "description": "If more thoughts are needed"
            }
          },
          "required": [
            "thought",
            "nextThoughtNeeded",
            "thoughtNumber",
            "totalThoughts"
          ]
        }
      }
    ]
  },
  "jsonrpc": "2.0",
  "id": 1
}

As you can see, there is a type mismatch for thoughtNumber param. The MCP server wants an integer but the model generated a String value, so I'm getting an error:

java.lang.IllegalStateException: Error calling tool: [TextContent[audience=null, priority=null, text={
  "error": "Invalid thoughtNumber: must be a number",
  "status": "failed"
}]]

So, the question is if there is a way to intercept the tool calling process, validate and adjust the request payload to fit the schema through existing Spring AI tools which I can't find or I should use prompt engineering practices or develop some workarounds?