代码提交

offline-backend-20260512/backend/packages/harness/deerflow/runtime/scheduler/service.py

@@ -1445,6 +1445,27 @@ class ScheduledTaskService:
             return "\n".join(p for p in parts if p)
         return str(content)
 
+    async def _agent_soul_and_model(
+        self, task: dict[str, Any], model_name: str | None
+    ) -> tuple[str | None, str | None]:
+        """Resolve the task agent's SOUL; fall back to its model when none chosen.
+
+        Used by both template pipelines so the **task's selected agent** shapes
+        content conversion: the SOUL carries the agent's domain instructions,
+        and when the user didn't pick an explicit model the agent's configured
+        model is preferred over the system default. Returns ``(soul, model)``.
+        """
+        agent_ref = task.get("execution_agent_name")
+        agent_id = await self._resolve_agent_id(agent_ref, task["user_id"]) if agent_ref else None
+        soul = load_agent_soul(agent_id) if agent_id else None
+        if not model_name and agent_id:
+            try:
+                agent_cfg = load_agent_config(agent_id)
+                model_name = (agent_cfg.model if agent_cfg else None) or None
+            except Exception:
+                model_name = None
+        return soul, model_name
+
     async def _convert_to_template_data(
         self,
         template_id: str,
@@ -1574,8 +1595,9 @@ class ScheduledTaskService:
         cfg = self._html_config(task)
         model_name = str(cfg.get("model_name") or "").strip() or None
         # 「执行数据抽取」开关**只**控制 HTML 用「抽取数据」还是「模板默认数据」,
-        # 不影响 Markdown 生成等其它正常操作。
-        extract = bool(cfg.get("template_extract"))
+        # 不影响 Markdown 生成等其它正常操作。缺省(未存该键的老任务)视为开启:
+        # 模板有真实数据来源后,默认走「智能体检索 → 转换填充」,显式 False 才关闭。
+        extract = cfg.get("template_extract") is not False
         also_markdown = _task_requires_markdown(task)
         prompt = str(task.get("prompt") or "").strip()
         description = str(task.get("description") or "").strip()
@@ -1626,9 +1648,12 @@ class ScheduledTaskService:
         source = (data_text or "").strip()
         used_default = True
         if do_extract and source:
+            # 转换阶段同样带上任务所选智能体:SOUL 提供领域口径,未显式选模型时
+            # 兜底用该智能体配置的模型(与模板填充模式一致)。
+            soul, conv_model = await self._agent_soul_and_model(task, model_name)
             try:
                 converted = await self._convert_to_template_data(
-                    tpl_id, source, model_name=model_name, soul=None, run_id=run_id
+                    tpl_id, source, model_name=conv_model, soul=soul, run_id=run_id
                 )
                 # 子页面没抽到有效内容的,回退模板原有数据,避免子页面空白。
                 data_obj = template_fill.merge_with_reference(tpl_id, converted)
@@ -1712,16 +1737,8 @@ class ScheduledTaskService:
             tcfg = tcfg if isinstance(tcfg, dict) else {}
 
             # 选中的内置 agent 的 SOUL + 模型(template_config.model_name 优先,其次 agent.model)。
-            agent_ref = task.get("execution_agent_name")
-            agent_id = await self._resolve_agent_id(agent_ref, task["user_id"]) if agent_ref else None
-            soul = load_agent_soul(agent_id) if agent_id else None
             model_name = str(tcfg.get("model_name") or "").strip() or None
-            if not model_name and agent_id:
-                try:
-                    agent_cfg = load_agent_config(agent_id)
-                    model_name = (agent_cfg.model if agent_cfg else None) or None
-                except Exception:
-                    model_name = None
+            soul, model_name = await self._agent_soul_and_model(task, model_name)
 
             converted = await self._convert_to_template_data(
                 template_id, source, model_name=model_name, soul=soul, run_id=run_id