pydantic-mcp

Instructions

Implementation Reference

• src/pydantic_mcp/helpers.py:1012-1040 (handler)

  The core implementation of the generate_model_from_json logic, which performs JSON parsing, model inference, and model rendering.

  def generate_model_from_json_report(
      json_input: Any,
      *,
      model_name: str,
  ) -> ToolResponse:
      try:
          payload = json.loads(json_input) if isinstance(json_input, str) else json_input
      except json.JSONDecodeError as exc:
          return make_response(
              diagnostics=[
                  Diagnostic(
                      level="error",
                      message="Input is not valid JSON.",
                      code="invalid_json",
                      context={"error": str(exc)},
                  )
              ],
              result={"ok": False, "code": None, "models": [], "input_kind": "invalid"},
          )
  
      effective_model_name = _sanitize_model_name(model_name)
      models: dict[str, GeneratedModelSpec] = {}
      root_model = _infer_root_model(payload, effective_model_name, models=models)
      models[root_model.name] = root_model
      ordered_models = list(models.values())
      code = _render_generated_models(ordered_models)
  
      return make_response(
          diagnostics=[

• src/pydantic_mcp/tools.py:293-298 (handler)

  The tool entry point that wraps the generation logic.

  def generate_model_from_json(
      json_input: object,
      model_name: str = "GeneratedModel",
  ) -> ToolResponse:
      """Infer candidate Pydantic models from a JSON string or JSON-like payload."""
      return generate_model_from_json_report(json_input, model_name=model_name)

• src/pydantic_mcp/server.py:45-45 (registration)

  Registration of the generate_model_from_json tool in the MCP server setup.

  "generate_model_from_json",