From 5e60e623121f5d251210b59f35dcde268cef09e1 Mon Sep 17 00:00:00 2001 From: wxg0103 <727495428@qq.com> Date: Wed, 21 Oct 2026 15:58:08 +0800 Subject: [PATCH] refactor: enhance ApplicationChatQuerySerializers with detailed node handling and message extraction --- .../serializers/application_chat.py | 100 ++++++++++++++++-- 1 file changed, 94 insertions(+), 6 deletions(-) diff --git a/apps/application/serializers/application_chat.py b/apps/application/serializers/application_chat.py index 28f01096796..d20c7e3b239 100644 --- a/apps/application/serializers/application_chat.py +++ b/apps/application/serializers/application_chat.py @@ -175,18 +175,106 @@ def paragraph_list_to_string(paragraph_list): else "" ) + @staticmethod + def _detail_nodes(details): + """兼容 dict(按节点 id 索引)、list(节点数组)与 v3 execution_details 结构，并展平循环节点内的子节点。""" + + def iter_items(container): + if isinstance(container, dict): + for key, value in container.items(): + if isinstance(value, dict): + yield key, value + else: + for key, value in enumerate(container): + if isinstance(value, dict): + yield key, value + + def visit(key, node): + # v3 处理后结构：节点嵌套在 execution_details 列表中 + execution = node.get("execution_details") + if isinstance(execution, list): + for index, child in enumerate(execution): + if isinstance(child, dict): + yield from visit(index, child) + return + if node.get("type") == "loop-node": + for loop_item in node.get("loop_node_data") or []: + for inner_key, inner in iter_items(loop_item): + yield from visit(inner_key, inner) + return + yield key, node + + for key, node in iter_items(details): + yield from visit(key, node) + + @staticmethod + def _messages_text(messages): + """从 messages 列表(新结构)提取 REASONING 思考与 TEXT 回答。""" + thinking_parts, answer_parts = [], [] + for message in messages or []: + if not isinstance(message, dict): + continue + content = message.get("content") + if not isinstance(content, str) or not content: + continue + # if message.get("type") == "REASONING": + # thinking_parts.append(content) + # el + if message.get("type") == "TEXT": + answer_parts.append(content) + return "\n\n".join(thinking_parts), "\n\n".join(answer_parts) + + @staticmethod + def _node_answer_text(node): + """从节点取值：优先 messages(新结构)的 REASONING/TEXT，回退节点上的 reasoning_content/answer(历史)。""" + reasoning, answer = ApplicationChatQuerySerializers._messages_text(node.get("messages")) + if not reasoning: + reasoning = node.get("reasoning_content") or "" + if not answer: + answer = node.get("answer") or "" + return reasoning, answer + @staticmethod def to_row(row: Dict): details = row.get("details") or {} + node_items = list(ApplicationChatQuerySerializers._detail_nodes(details)) + + # 新结构：详情顶层(或其包装元素)可能携带记录级 messages，作为回答的兜底来源 + record_messages = None + if isinstance(details, dict): + record_messages = details.get("messages") + elif isinstance(details, list) and details and isinstance(details[0], dict): + record_messages = details[0].get("messages") + + # v3 的用户问题/回答记录在节点上，而记录级字段可能为空，这里做兜底填充 + question_list, answer_blocks = [], [] + for key, node in node_items: + node_type = node.get("type") + if node_type in ("start-node", "ai-chat-node", "llm-node", "reply-node", "application-node"): + question = node.get("question") + if isinstance(question, str) and question: + question_list.append(question) + elif isinstance(question, dict) and question.get("content"): + question_list.append(question["content"]) + reasoning, answer = ApplicationChatQuerySerializers._node_answer_text(node) + block = [part for part in (reasoning, answer) if part] + if block: + answer_blocks.append("\n\n".join(block)) + problem_text = row.get("problem_text") or " / ".join(dict.fromkeys(question_list)) + if not answer_blocks: + reasoning, answer = ApplicationChatQuerySerializers._messages_text(record_messages) + answer_text = row.get("answer_text") or ("\n\n".join([part for part in (reasoning, answer) if part])) + else: + answer_text = row.get("answer_text") or answer_blocks[-1] + padding_problem_text = " ".join( - (node.get("answer", "") or "") for key, node in details.items() if node.get("type") == "question-node" + (node.get("answer", "") or "") for key, node in node_items if node.get("type") == "question-node" ) search_dataset_node_list = [ (key, node) - for key, node in details.items() - if node.get("type") == "search-dataset-node" + for key, node in node_items + if node.get("type") in ("search-dataset-node", "search-knowledge-node", "search-document-node") or node.get("step_type") == "search_step" - or node.get("type") == "search-knowledge-node" ] reference_paragraph_len = "\n".join( [ @@ -220,9 +308,9 @@ def to_row(row: Dict): return [ str(row.get("chat_id")), row.get("abstract"), - row.get("problem_text"), + problem_text, padding_problem_text, - row.get("answer_text"), + answer_text, vote_status_map.get(row.get("vote_status")), vote_reason_map.get(row.get("vote_reason")), row.get("vote_other_content"),