Fix SSE route dependency and align architecture docs

backend/app/workflows/rag_workflow.py

@@ -1,8 +1,13 @@
+"""Define workflow state for rag workflow."""
+
 from typing import TypedDict, List
 from langgraph.graph import StateGraph, END
+# Keep workflow state definitions compact so transitions stay easy to audit.
+
 
 
 class RagState(TypedDict):
+    """Track workflow state for rag state."""
     query: str
     query_embedding: List[float]
     retrieved_docs: List[dict]
@@ -12,14 +17,14 @@ class RagState(TypedDict):
 
 
 def embed_query(state: RagState) -> dict:
-    """将查询转为向量"""
+    """Embed query."""
     from app.services import embedding_service
     embedding = embedding_service.embed_single(state["query"])
     return {"query_embedding": embedding}
 
 
 def retrieve_docs(state: RagState) -> dict:
-    """向量检索"""
+    """Handle retrieve docs."""
     from app.services import milvus_service
     from app.core.config import settings
     docs = milvus_service.search(
@@ -30,7 +35,7 @@ def retrieve_docs(state: RagState) -> dict:
 
 
 def build_context(state: RagState) -> dict:
-    """构建上下文"""
+    """Build context."""
     context_parts = []
     sources = []
 
@@ -46,7 +51,7 @@ def build_context(state: RagState) -> dict:
 
 
 def generate_answer(state: RagState) -> dict:
-    """生成答案"""
+    """Handle generate answer."""
     from app.services import llm_service
     prompt = f"""请根据以下法规内容回答用户问题，并在回答中标注引用的法规条款。
 
@@ -64,7 +69,7 @@ def generate_answer(state: RagState) -> dict:
     return {"answer": answer}
 
 
-# 构建工作流图
+# Keep workflow state definitions compact so transitions stay easy to audit.
 rag_graph = StateGraph(RagState)
 
 rag_graph.add_node("embed", embed_query)
@@ -82,23 +87,23 @@ rag_workflow = rag_graph.compile()
 
 
 async def run_rag_workflow(query: str) -> RagState:
-    """运行RAG工作流"""
+    """Handle run rag workflow."""
     initial_state: RagState = {"query": query}
     result = rag_workflow.invoke(initial_state)
     return result
 
 
 def stream_rag_workflow(query: str):
-    """流式运行RAG工作流"""
+    """Stream rag workflow."""
     from app.services import llm_service
 
-    # 先完成检索阶段
+    # Keep workflow state definitions compact so transitions stay easy to audit.
     state: RagState = {"query": query}
     state.update(embed_query(state))
     state.update(retrieve_docs(state))
     state.update(build_context(state))
 
-    # 流式生成阶段
+    # Keep workflow state definitions compact so transitions stay easy to audit.
     prompt = f"""请根据以下法规内容回答用户问题，并在回答中标注引用的法规条款。
 
 法规内容：