refactor: llm decouple code executor module (#33400)

Co-authored-by: Byron.wang <byron@dify.ai>

api/dify_graph/nodes/llm/protocols.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from collections.abc import Mapping
 from typing import Any, Protocol
 
 from core.model_manager import ModelInstance
@@ -19,3 +20,11 @@ class ModelFactory(Protocol):
     def init_model_instance(self, provider_name: str, model_name: str) -> ModelInstance:
         """Create a model instance that is ready for schema lookup and invocation."""
         ...
+
+
+class TemplateRenderer(Protocol):
+    """Port for rendering prompt templates used by LLM-compatible nodes."""
+
+    def render_jinja2(self, *, template: str, inputs: Mapping[str, Any]) -> str:
+        """Render the given Jinja2 template into plain text."""
+        ...