fix(harness): normalize structured content for titles (#1155)

* fix(harness): normalize structured content for titles

Flatten structured LangChain message content before prompting the title model so list/block payloads don't leak Python reprs into generated thread titles.

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

* Apply suggestions from code review

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

---------

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

backend/CLAUDE.md

@@ -158,7 +158,7 @@ Middlewares execute in strict order in `packages/harness/deerflow/agents/lead_ag
 4. **DanglingToolCallMiddleware** - Injects placeholder ToolMessages for AIMessage tool_calls that lack responses (e.g., due to user interruption)
 5. **SummarizationMiddleware** - Context reduction when approaching token limits (optional, if enabled)
 6. **TodoListMiddleware** - Task tracking with `write_todos` tool (optional, if plan_mode)
-7. **TitleMiddleware** - Auto-generates thread title after first complete exchange
+7. **TitleMiddleware** - Auto-generates thread title after first complete exchange and normalizes structured message content before prompting the title model
 8. **MemoryMiddleware** - Queues conversations for async memory update (filters to user + final AI responses)
 9. **ViewImageMiddleware** - Injects base64 image data before LLM call (conditional on vision support)
 10. **SubagentLimitMiddleware** - Truncates excess `task` tool calls from model response to enforce `MAX_CONCURRENT_SUBAGENTS` limit (optional, if subagent_enabled)

backend/docs/AUTO_TITLE_GENERATION.md

@@ -6,12 +6,14 @@
 
 ## 实现方式
 
-使用 `TitleMiddleware` 在 `after_agent` 钩子中：
+使用 `TitleMiddleware` 在 `after_model` 钩子中：
 1. 检测是否是首次对话（1个用户消息 + 1个助手回复）
 2. 检查 state 是否已有 title
 3. 调用 LLM 生成简洁的标题（默认最多6个词）
 4. 将 title 存储到 `ThreadState` 中（会被 checkpointer 持久化）
 
+TitleMiddleware 会先把 LangChain message content 里的结构化 block/list 内容归一化为纯文本，再拼到 title prompt 里，避免把 Python/JSON 的原始 repr 泄漏到标题生成模型。
+
 ## ⚠️ 重要：存储机制
 
 ### Title 存储位置

backend/packages/harness/deerflow/agents/middlewares/title_middleware.py

@@ -21,6 +21,25 @@ class TitleMiddleware(AgentMiddleware[TitleMiddlewareState]):
 
     state_schema = TitleMiddlewareState
 
+    def _normalize_content(self, content: object) -> str:
+        if isinstance(content, str):
+            return content
+
+        if isinstance(content, list):
+            parts = [self._normalize_content(item) for item in content]
+            return "\n".join(part for part in parts if part)
+
+        if isinstance(content, dict):
+            text_value = content.get("text")
+            if isinstance(text_value, str):
+                return text_value
+
+            nested_content = content.get("content")
+            if nested_content is not None:
+                return self._normalize_content(nested_content)
+
+        return ""
+
     def _should_generate_title(self, state: TitleMiddlewareState) -> bool:
         """Check if we should generate a title for this thread."""
         config = get_title_config()
@@ -52,9 +71,8 @@ class TitleMiddleware(AgentMiddleware[TitleMiddlewareState]):
         user_msg_content = next((m.content for m in messages if m.type == "human"), "")
         assistant_msg_content = next((m.content for m in messages if m.type == "ai"), "")
 
-        # Ensure content is string (LangChain messages can have list content)
-        user_msg = str(user_msg_content) if user_msg_content else ""
-        assistant_msg = str(assistant_msg_content) if assistant_msg_content else ""
+        user_msg = self._normalize_content(user_msg_content)
+        assistant_msg = self._normalize_content(assistant_msg_content)
 
         # Use a lightweight model to generate title
         model = create_chat_model(thinking_enabled=False)
@@ -67,8 +85,7 @@ class TitleMiddleware(AgentMiddleware[TitleMiddlewareState]):
 
         try:
             response = await model.ainvoke(prompt)
-            # Ensure response content is string
-            title_content = str(response.content) if response.content else ""
+            title_content = self._normalize_content(response.content)
             title = title_content.strip().strip('"').strip("'")
             # Limit to max characters
             return title[: config.max_chars] if len(title) > config.max_chars else title

backend/tests/test_title_middleware_core_logic.py

@@ -92,6 +92,36 @@ class TestTitleMiddlewareCoreLogic:
         assert "'" not in title
         assert len(title) == 12
 
+    def test_generate_title_normalizes_structured_message_and_response_content(self, monkeypatch):
+        _set_test_title_config(max_chars=20)
+        middleware = TitleMiddleware()
+        fake_model = MagicMock()
+        fake_model.ainvoke = AsyncMock(
+            return_value=MagicMock(content=[{"type": "text", "text": '"结构总结"'}]),
+        )
+        monkeypatch.setattr(
+            "deerflow.agents.middlewares.title_middleware.create_chat_model",
+            lambda **kwargs: fake_model,
+        )
+
+        state = {
+            "messages": [
+                HumanMessage(content=[{"type": "text", "text": "请帮我总结这段代码"}]),
+                AIMessage(content=[{"type": "text", "text": "好的，先看结构"}]),
+            ]
+        }
+
+        title = asyncio.run(middleware._generate_title(state))
+
+        prompt = fake_model.ainvoke.await_args.args[0]
+        assert "请帮我总结这段代码" in prompt
+        assert "好的，先看结构" in prompt
+        # Ensure structured message dict/JSON reprs are not leaking into the prompt.
+        assert "{'type':" not in prompt
+        assert "'type':" not in prompt
+        assert '"type":' not in prompt
+        assert title == "结构总结"
+
     def test_generate_title_fallback_when_model_fails(self, monkeypatch):
         _set_test_title_config(max_chars=20)
         middleware = TitleMiddleware()