|
|
@@ -0,0 +1,691 @@
|
|
|
+## 简化版LangGraph状态监控方案
|
|
|
+
|
|
|
+### 🎯 **核心理念**
|
|
|
+像ChatGPT一样，只告诉用户"AI正在思考"，不显示百分比进度，专注于当前执行状态。
|
|
|
+
|
|
|
+### 📡 **单一API设计**
|
|
|
+
|
|
|
+**接口**: `GET /api/status/{thread_id}`
|
|
|
+
|
|
|
+**响应格式**:
|
|
|
+```json
|
|
|
+{
|
|
|
+ "status": "running|completed",
|
|
|
+ "name": "AI思考中",
|
|
|
+ "icon": "🤖",
|
|
|
+ "timestamp": "2025-07-30T15:30:45.123Z"
|
|
|
+}
|
|
|
+```
|
|
|
+
|
|
|
+### 🔄 **状态映射**
|
|
|
+基于您的LangGraph流程图节点：
|
|
|
+
|
|
|
+| 节点 | 显示名称 | 图标 |
|
|
|
+|------|----------|------|
|
|
|
+| `__start__` | 开始 | 🚀 |
|
|
|
+| `agent` | AI思考中 | 🤖 |
|
|
|
+| `prepare_tool_input` | 准备工具 | 🔧 |
|
|
|
+| `tools` | 执行查询 | ⚙️ |
|
|
|
+| `update_state_after_tool` | 处理结果 | 🔄 |
|
|
|
+| `format_final_response` | 生成回答 | 📝 |
|
|
|
+| `__end__` | 完成 | ✅ |
|
|
|
+
|
|
|
+### 🏗️ **实现逻辑**
|
|
|
+1. 从Redis获取最新checkpoint
|
|
|
+2. 读取`suggested_next_step`字段
|
|
|
+3. 如果为空或`__end__` → 返回completed
|
|
|
+4. 否则 → 返回running + 对应状态名称
|
|
|
+
|
|
|
+### 💡 **使用方式**
|
|
|
+
|
|
|
+**客户端轮询**:
|
|
|
+```bash
|
|
|
+# Postman测试
|
|
|
+GET http://localhost:5000/api/status/wang1:20250729235038043
|
|
|
+
|
|
|
+# 返回示例1 (执行中)
|
|
|
+{
|
|
|
+ "status": "running",
|
|
|
+ "name": "AI思考中",
|
|
|
+ "icon": "🤖",
|
|
|
+ "timestamp": "2025-07-30T15:30:45.123Z"
|
|
|
+}
|
|
|
+
|
|
|
+# 返回示例2 (完成)
|
|
|
+{
|
|
|
+ "status": "completed",
|
|
|
+ "name": "完成",
|
|
|
+ "icon": "✅",
|
|
|
+ "timestamp": "2025-07-30T15:31:20.456Z"
|
|
|
+}
|
|
|
+```
|
|
|
+
|
|
|
+**客户端实现**:
|
|
|
+```javascript
|
|
|
+// 每秒轮询，直到status=completed就停止
|
|
|
+setInterval(() => {
|
|
|
+ fetch('/api/status/thread_id')
|
|
|
+ .then(r => r.json())
|
|
|
+ .then(data => {
|
|
|
+ console.log(`${data.icon} ${data.name}`);
|
|
|
+ if (data.status === 'completed') {
|
|
|
+ // 停止轮询
|
|
|
+ }
|
|
|
+ });
|
|
|
+}, 1000);
|
|
|
+```
|
|
|
+
|
|
|
+### ✨ **方案优势**
|
|
|
+- **极简**: 只有4个字段，一个API
|
|
|
+- **准确**: 基于真实的checkpoint状态
|
|
|
+- **直观**: 像ChatGPT一样的用户体验
|
|
|
+- **轻量**: 最小化网络传输和服务器负载
|
|
|
+- **可靠**: 直接读取LangGraph的执行状态
|
|
|
+
|
|
|
+这就是您要的最简化版本 - 一个API，几个字段，专注核心功能！
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+**简化LangGraph状态API** 的代码：
|
|
|
+
|
|
|
+```python
|
|
|
+
|
|
|
+from flask import Flask, jsonify
|
|
|
+import redis
|
|
|
+import json
|
|
|
+from datetime import datetime
|
|
|
+
|
|
|
+app = Flask(__name__)
|
|
|
+redis_client = redis.Redis(host='localhost', port=6379, db=0, decode_responses=True)
|
|
|
+
|
|
|
+
|
|
|
+# 简单的节点状态映射
|
|
|
+
|
|
|
+NODE_STATUS = {
|
|
|
+ "__start__": {"name": "开始", "icon": "🚀"},
|
|
|
+ "agent": {"name": "AI思考中", "icon": "🤖"},
|
|
|
+ "prepare_tool_input": {"name": "准备工具", "icon": "🔧"},
|
|
|
+ "tools": {"name": "执行查询", "icon": "⚙️"},
|
|
|
+ "update_state_after_tool": {"name": "处理结果", "icon": "🔄"},
|
|
|
+ "format_final_response": {"name": "生成回答", "icon": "📝"},
|
|
|
+ "__end__": {"name": "完成", "icon": "✅"}
|
|
|
+}
|
|
|
+
|
|
|
+@app.route('/api/status/<thread_id>')
|
|
|
+def get_status(thread_id):
|
|
|
+ try:
|
|
|
+ # 查找最新checkpoint
|
|
|
+ pattern = f"checkpoint:{thread_id}:*"
|
|
|
+ keys = redis_client.keys(pattern)
|
|
|
+
|
|
|
+ if not keys:
|
|
|
+ return jsonify({"error": "未找到执行线程"}), 404
|
|
|
+
|
|
|
+ # 获取最新checkpoint
|
|
|
+ latest_key = sorted(keys)[-1]
|
|
|
+ checkpoint_data = redis_client.get(latest_key)
|
|
|
+
|
|
|
+ if not checkpoint_data:
|
|
|
+ return jsonify({"error": "无法读取数据"}), 500
|
|
|
+
|
|
|
+ checkpoint = json.loads(checkpoint_data)
|
|
|
+ suggested_next_step = checkpoint.get('suggested_next_step', '')
|
|
|
+
|
|
|
+ # 判断状态
|
|
|
+ if not suggested_next_step or suggested_next_step == "__end__":
|
|
|
+ # 完成了
|
|
|
+ return jsonify({
|
|
|
+ "status": "completed",
|
|
|
+ "name": "完成",
|
|
|
+ "icon": "✅",
|
|
|
+ "timestamp": datetime.now().isoformat()
|
|
|
+ })
|
|
|
+ else:
|
|
|
+ # 正在执行
|
|
|
+ node_info = NODE_STATUS.get(suggested_next_step, {
|
|
|
+ "name": "执行中",
|
|
|
+ "icon": "⚙️"
|
|
|
+ })
|
|
|
+
|
|
|
+ return jsonify({
|
|
|
+ "status": "running",
|
|
|
+ "name": node_info["name"],
|
|
|
+ "icon": node_info["icon"],
|
|
|
+ "timestamp": datetime.now().isoformat()
|
|
|
+ })
|
|
|
+
|
|
|
+except Exception as e:
|
|
|
+ return jsonify({"error": str(e)}), 500
|
|
|
+
|
|
|
+if __name__ == '__main__':
|
|
|
+ app.run(debug=True, host='0.0.0.0', port=5000)
|
|
|
+```
|
|
|
+
|
|
|
+# 两种获取Checkpoint的方法
|
|
|
+
|
|
|
+### 🔥 **方法一：直接Redis访问**
|
|
|
+
|
|
|
+就是我们之前使用的方法，直接查询Redis键：
|
|
|
+
|
|
|
+```python
|
|
|
+# 直接Redis查询
|
|
|
+pattern = f"checkpoint:{thread_id}:*"
|
|
|
+keys = redis_client.keys(pattern)
|
|
|
+latest_key = sorted(keys)[-1]
|
|
|
+checkpoint_data = redis_client.get(latest_key)
|
|
|
+```
|
|
|
+
|
|
|
+### 🚀 **方法二：使用LangGraph API**
|
|
|
+
|
|
|
+使用官方的Redis Checkpoint Saver API：## 📊 **两种方法对比**
|
|
|
+
|
|
|
+### **方法一：直接Redis访问**
|
|
|
+
|
|
|
+✅ **优点**：
|
|
|
+
|
|
|
+- 更直接，性能稍快
|
|
|
+- 完全控制数据格式
|
|
|
+- 不依赖LangGraph版本
|
|
|
+
|
|
|
+❌ **缺点**：
|
|
|
+
|
|
|
+- 需要了解Redis键名格式
|
|
|
+
|
|
|
+ - ```
|
|
|
+ checkpoint:wang1:20250729235038043:__empty__:1f06c943-fd28-6539-801f-41a94c25bd43
|
|
|
+ ```
|
|
|
+
|
|
|
+- 硬编码依赖具体实现
|
|
|
+
|
|
|
+- 升级时可能出问题
|
|
|
+
|
|
|
+### **方法二：LangGraph API** ⭐ **推荐**
|
|
|
+
|
|
|
+✅ **优点**：
|
|
|
+
|
|
|
+- 官方API，更稳定可靠
|
|
|
+- 自动处理索引和版本兼容
|
|
|
+- 标准化的CheckpointTuple对象
|
|
|
+- 支持异步操作
|
|
|
+
|
|
|
+❌ **缺点**：
|
|
|
+
|
|
|
+- 需要安装`langgraph-checkpoint-redis`包
|
|
|
+- 稍微多一点的API调用开销
|
|
|
+
|
|
|
+## 🔧 **关键API使用**
|
|
|
+
|
|
|
+基于搜索结果，官方Redis Checkpointer的核心方法：
|
|
|
+
|
|
|
+```python
|
|
|
+# 同步版本
|
|
|
+from langgraph.checkpoint.redis import RedisSaver
|
|
|
+
|
|
|
+checkpointer = RedisSaver.from_conn_string("redis://localhost:6379")
|
|
|
+checkpointer.setup() # 重要：创建索引
|
|
|
+
|
|
|
+# 获取最新checkpoint
|
|
|
+read_config = {"configurable": {"thread_id": "your_thread_id"}}
|
|
|
+checkpoint_tuple = checkpointer.get_tuple(read_config)
|
|
|
+
|
|
|
+# 异步版本
|
|
|
+from langgraph.checkpoint.redis.aio import AsyncRedisSaver
|
|
|
+
|
|
|
+async with AsyncRedisSaver.from_conn_string("redis://localhost:6379") as checkpointer:
|
|
|
+ await checkpointer.asetup()
|
|
|
+ checkpoint_tuple = await checkpointer.aget_tuple(read_config)
|
|
|
+```
|
|
|
+
|
|
|
+## 💡 **建议**
|
|
|
+
|
|
|
+我建议使用**方法二（LangGraph API）**，因为：
|
|
|
+
|
|
|
+1. 这是官方推荐方式，更可靠
|
|
|
+
|
|
|
+2. 自动处理版本兼容性问题
|
|
|
+
|
|
|
+3. 提供标准化的数据结构
|
|
|
+
|
|
|
+4. 您的代码会更加健壮和可维护
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+ ### 方法二参考代码
|
|
|
+
|
|
|
+```python
|
|
|
+from flask import Flask, jsonify
|
|
|
+from langgraph.checkpoint.redis import RedisSaver
|
|
|
+from datetime import datetime
|
|
|
+import asyncio
|
|
|
+
|
|
|
+app = Flask(__name__)
|
|
|
+
|
|
|
+# 初始化Redis Checkpointer
|
|
|
+checkpointer = RedisSaver.from_conn_string("redis://localhost:6379")
|
|
|
+checkpointer.setup() # 创建必要的索引
|
|
|
+
|
|
|
+# 简单的节点状态映射
|
|
|
+NODE_STATUS = {
|
|
|
+ "__start__": {"name": "开始", "icon": "🚀"},
|
|
|
+ "agent": {"name": "AI思考中", "icon": "🤖"},
|
|
|
+ "prepare_tool_input": {"name": "准备工具", "icon": "🔧"},
|
|
|
+ "tools": {"name": "执行查询", "icon": "⚙️"},
|
|
|
+ "update_state_after_tool": {"name": "处理结果", "icon": "🔄"},
|
|
|
+ "format_final_response": {"name": "生成回答", "icon": "📝"},
|
|
|
+ "__end__": {"name": "完成", "icon": "✅"}
|
|
|
+}
|
|
|
+
|
|
|
+@app.route('/api/status/<thread_id>')
|
|
|
+def get_status(thread_id):
|
|
|
+ try:
|
|
|
+ # 使用LangGraph API获取最新checkpoint
|
|
|
+ read_config = {"configurable": {"thread_id": thread_id}}
|
|
|
+
|
|
|
+ # 获取最新的checkpoint tuple
|
|
|
+ checkpoint_tuple = checkpointer.get_tuple(read_config)
|
|
|
+
|
|
|
+ if not checkpoint_tuple:
|
|
|
+ return jsonify({"error": "未找到执行线程"}), 404
|
|
|
+
|
|
|
+ # 从checkpoint中提取状态信息
|
|
|
+ checkpoint = checkpoint_tuple.checkpoint
|
|
|
+
|
|
|
+ # 获取suggested_next_step (这里需要根据实际的checkpoint结构调整)
|
|
|
+ # 通常在channel_values中或者通过分析当前状态推断
|
|
|
+ channel_values = checkpoint.get('channel_values', {})
|
|
|
+
|
|
|
+ # 根据您的实际LangGraph实现，suggested_next_step可能在不同位置
|
|
|
+ # 这里提供几种可能的获取方式：
|
|
|
+
|
|
|
+ # 方式1: 直接从channel_values获取
|
|
|
+ suggested_next_step = channel_values.get('suggested_next_step', '')
|
|
|
+
|
|
|
+ # 方式2: 从其他可能的位置获取
|
|
|
+ if not suggested_next_step:
|
|
|
+ # 可能在metadata中
|
|
|
+ metadata = checkpoint_tuple.metadata or {}
|
|
|
+ suggested_next_step = metadata.get('suggested_next_step', '')
|
|
|
+
|
|
|
+ # 方式3: 根据当前节点状态推断下一步
|
|
|
+ if not suggested_next_step:
|
|
|
+ current_node = channel_values.get('current_node', '')
|
|
|
+ # 根据业务逻辑推断下一步...
|
|
|
+
|
|
|
+ # 判断是否完成
|
|
|
+ is_completed = (not suggested_next_step or
|
|
|
+ suggested_next_step == "__end__" or
|
|
|
+ channel_values.get('status') == 'completed')
|
|
|
+
|
|
|
+ if is_completed:
|
|
|
+ return jsonify({
|
|
|
+ "status": "completed",
|
|
|
+ "name": "完成",
|
|
|
+ "icon": "✅",
|
|
|
+ "timestamp": datetime.now().isoformat()
|
|
|
+ })
|
|
|
+ else:
|
|
|
+ # 获取当前步骤信息
|
|
|
+ node_info = NODE_STATUS.get(suggested_next_step, {
|
|
|
+ "name": "执行中",
|
|
|
+ "icon": "⚙️"
|
|
|
+ })
|
|
|
+
|
|
|
+ return jsonify({
|
|
|
+ "status": "running",
|
|
|
+ "name": node_info["name"],
|
|
|
+ "icon": node_info["icon"],
|
|
|
+ "timestamp": datetime.now().isoformat()
|
|
|
+ })
|
|
|
+
|
|
|
+ except Exception as e:
|
|
|
+ return jsonify({"error": f"获取状态失败: {str(e)}"}), 500
|
|
|
+
|
|
|
+# 异步版本 (如果需要)
|
|
|
+@app.route('/api/status-async/<thread_id>')
|
|
|
+def get_status_async(thread_id):
|
|
|
+ try:
|
|
|
+ # 使用异步API
|
|
|
+ async def _get_async_status():
|
|
|
+ from langgraph.checkpoint.redis.aio import AsyncRedisSaver
|
|
|
+
|
|
|
+ async with AsyncRedisSaver.from_conn_string("redis://localhost:6379") as async_checkpointer:
|
|
|
+ await async_checkpointer.asetup()
|
|
|
+
|
|
|
+ read_config = {"configurable": {"thread_id": thread_id}}
|
|
|
+ checkpoint_tuple = await async_checkpointer.aget_tuple(read_config)
|
|
|
+
|
|
|
+ if not checkpoint_tuple:
|
|
|
+ return None
|
|
|
+
|
|
|
+ return checkpoint_tuple.checkpoint
|
|
|
+
|
|
|
+ # 在Flask中运行异步代码
|
|
|
+ loop = asyncio.new_event_loop()
|
|
|
+ asyncio.set_event_loop(loop)
|
|
|
+
|
|
|
+ try:
|
|
|
+ checkpoint = loop.run_until_complete(_get_async_status())
|
|
|
+ if not checkpoint:
|
|
|
+ return jsonify({"error": "未找到执行线程"}), 404
|
|
|
+
|
|
|
+ # 处理checkpoint数据...
|
|
|
+ # (相同的逻辑)
|
|
|
+
|
|
|
+ finally:
|
|
|
+ loop.close()
|
|
|
+
|
|
|
+ except Exception as e:
|
|
|
+ return jsonify({"error": str(e)}), 500
|
|
|
+
|
|
|
+@app.route('/api/health')
|
|
|
+def health_check():
|
|
|
+ """健康检查"""
|
|
|
+ try:
|
|
|
+ # 测试checkpointer连接
|
|
|
+ test_config = {"configurable": {"thread_id": "health_check"}}
|
|
|
+ checkpointer.get_tuple(test_config) # 这不会报错，即使没有数据
|
|
|
+
|
|
|
+ return jsonify({
|
|
|
+ "status": "healthy",
|
|
|
+ "checkpointer": "connected",
|
|
|
+ "timestamp": datetime.now().isoformat()
|
|
|
+ })
|
|
|
+ except Exception as e:
|
|
|
+ return jsonify({
|
|
|
+ "status": "unhealthy",
|
|
|
+ "error": str(e)
|
|
|
+ }), 500
|
|
|
+
|
|
|
+if __name__ == '__main__':
|
|
|
+ app.run(debug=True, host='0.0.0.0', port=5000)
|
|
|
+```
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+## Checkpoint在Redis中的存储的值
|
|
|
+
|
|
|
+### key:
|
|
|
+
|
|
|
+```
|
|
|
+checkpoint:wang1:20250729235038043:__empty__:1f06c943-fd28-6539-801f-41a94c25bd43
|
|
|
+```
|
|
|
+
|
|
|
+其中“wang1:20250729235038043”就是thread_id，它是由 {user_id}:{时间戳}组成的。
|
|
|
+
|
|
|
+### value:
|
|
|
+
|
|
|
+```json
|
|
|
+ {
|
|
|
+ "type": "ai",
|
|
|
+ "content": "系统中档口最多的服务区是**南城服务区**，共拥有**39个档口**。该结果通过统计各服务区内未删除档口的数量并按降序排列得出，已通过系统验证确保数据准确性。"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "question": "请问这个服务区有几个餐饮档口？"
|
|
|
+ },
|
|
|
+ "id": "call_498bcec43e2c4b3499e6e1"
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "name": "",
|
|
|
+ "args": {
|
|
|
+
|
|
|
+ },
|
|
|
+ "id": "call_f71ffe8603ed42da879a17",
|
|
|
+ "type": "tool_call"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "invalid_tool_calls": []
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "ToolMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "SELECT COUNT(*) AS 餐饮档口数量 \nFROM bss_branch b \nJOIN bss_service_area sa ON b.service_area_id = sa.id \nWHERE sa.service_area_name = '南城服务区' \nAND b.classify = '餐饮' \nAND b.delete_ts IS NULL \nAND sa.delete_ts IS NULL;",
|
|
|
+ "type": "tool",
|
|
|
+ "name": "generate_sql",
|
|
|
+ "id": "3dfacea1-93af-444c-bfb4-2995f336f274",
|
|
|
+ "tool_call_id": "call_498bcec43e2c4b3499e6e1",
|
|
|
+ "status": "success"
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "ToolMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "Error: is not a valid tool, try one of [generate_sql, valid_sql, run_sql].",
|
|
|
+ "type": "tool",
|
|
|
+ "name": "",
|
|
|
+ "id": "1763e2eb-7be4-4e7e-bae3-19f9e0b1b4b5",
|
|
|
+ "tool_call_id": "call_f71ffe8603ed42da879a17",
|
|
|
+ "status": "error"
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "AIMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "",
|
|
|
+ "additional_kwargs": {
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "index": 0,
|
|
|
+ "id": "call_09fe3a57143947e5b85766",
|
|
|
+ "function": {
|
|
|
+ "arguments": "{\"history_messages\": [{\"type\": \"human\", \"content\": \"请问系统中哪个服务区档口最多？\"}, {\"type\": \"ai\", \"content\": \"系统中档口最多的服务区是**南城服务区**，共拥有**39个档口**。该结果通过统计各服务区内未删除档口的数量并按降序排列得出，已通过系统验证确保数据准确性。\"}], \"question\": \"请问这个服务区有几个餐饮档口？\"}",
|
|
|
+ "name": "generate_sql"
|
|
|
+ },
|
|
|
+ "type": "function"
|
|
|
+ }
|
|
|
+ ]
|
|
|
+ },
|
|
|
+ "response_metadata": {
|
|
|
+ "finish_reason": "tool_calls",
|
|
|
+ "model_name": "qwen3-235b-a22b"
|
|
|
+ },
|
|
|
+ "type": "ai",
|
|
|
+ "id": "run--d61425aa-94cf-4e0c-837a-9ac7c48f4930-0",
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "name": "generate_sql",
|
|
|
+ "args": {
|
|
|
+ "history_messages": [
|
|
|
+ {
|
|
|
+ "type": "human",
|
|
|
+ "content": "请问系统中哪个服务区档口最多？"
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "type": "ai",
|
|
|
+ "content": "系统中档口最多的服务区是**南城服务区**，共拥有**39个档口**。该结果通过统计各服务区内未删除档口的数量并按降序排列得出，已通过系统验证确保数据准确性。"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "question": "请问这个服务区有几个餐饮档口？"
|
|
|
+ },
|
|
|
+ "id": "call_09fe3a57143947e5b85766"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "invalid_tool_calls": []
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "ToolMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "SELECT COUNT(*) AS 餐饮档口数量 FROM bss_branch b JOIN bss_service_area sa ON b.service_area_id = sa.id WHERE sa.service_area_name = '南城服务区' AND b.classify = '餐饮' AND b.delete_ts IS NULL;",
|
|
|
+ "type": "tool",
|
|
|
+ "name": "generate_sql",
|
|
|
+ "id": "8016bd5b-9ad7-43b3-90f2-1994b798b06b",
|
|
|
+ "tool_call_id": "call_09fe3a57143947e5b85766",
|
|
|
+ "status": "success"
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "AIMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "",
|
|
|
+ "additional_kwargs": {
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "index": 0,
|
|
|
+ "id": "call_0753617fb1174fab9074eb",
|
|
|
+ "function": {
|
|
|
+ "arguments": "{\"sql\": \"SELECT COUNT(*) AS \\u9910\\u996e\\u6863\\u53e3\\u6570\\u91cf \\nFROM bss_branch b \\nJOIN bss_service_area sa ON b.service_area_id = sa.id \\nWHERE sa.service_area_name = '南城服务区' \\nAND b.classify = '餐饮' \\nAND b.delete_ts IS NULL;\", \"question\": \"\\u8bf7\\u95ee\\u8fd9\\u4e2a\\u670d\\u52a1\\u533a\\u6709\\u51e0\\u4e2a\\u9910\\u996e\\u6863\\u53e3\\uff1f\"}",
|
|
|
+ "name": "valid_sql"
|
|
|
+ },
|
|
|
+ "type": "function"
|
|
|
+ }
|
|
|
+ ]
|
|
|
+ },
|
|
|
+ "response_metadata": {
|
|
|
+ "finish_reason": "tool_calls",
|
|
|
+ "model_name": "qwen3-235b-a22b"
|
|
|
+ },
|
|
|
+ "type": "ai",
|
|
|
+ "id": "run--4c0c02fa-efbb-4cd6-96c0-d80392c99aa9-0",
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "name": "valid_sql",
|
|
|
+ "args": {
|
|
|
+ "sql": "SELECT COUNT(*) AS 餐饮档口数量 \nFROM bss_branch b \nJOIN bss_service_area sa ON b.service_area_id = sa.id \nWHERE sa.service_area_name = '南城服务区' \nAND b.classify = '餐饮' \nAND b.delete_ts IS NULL;",
|
|
|
+ "question": "请问这个服务区有几个餐饮档口？"
|
|
|
+ },
|
|
|
+ "id": "call_0753617fb1174fab9074eb",
|
|
|
+ "type": "tool_call"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "invalid_tool_calls": []
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "ToolMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "SQL验证通过：语法正确且字段存在",
|
|
|
+ "type": "tool",
|
|
|
+ "name": "valid_sql",
|
|
|
+ "id": "62bde32c-02a9-4e36-aa76-5387cb03ce65",
|
|
|
+ "tool_call_id": "call_0753617fb1174fab9074eb",
|
|
|
+ "status": "success"
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "AIMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "",
|
|
|
+ "additional_kwargs": {
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "index": 0,
|
|
|
+ "id": "call_b0600de998f5467d96a3f2",
|
|
|
+ "function": {
|
|
|
+ "arguments": "{\"sql\": \"SELECT COUNT(*) AS \\u9910\\u996e\\u6863\\u53e3\\u6570\\u91cf \\nFROM bss_branch b \\nJOIN bss_service_area sa ON b.service_area_id = sa.id \\nWHERE sa.service_area_name = '\\u5357\\u57ce\\u670d\\u52a1\\u533a' \\nAND b.classify = '\\u9910\\u996e' \\nAND b.delete_ts IS NULL;\", \"question\": \"\\u8bf7\\u95ee\\u8fd9\\u4e2a\\u670d\\u52a1\\u533a\\u6709\\u51e0\\u4e2a\\u9910\\u996e\\u6863\\u53e3\\uff1f\"}",
|
|
|
+ "name": "run_sql"
|
|
|
+ },
|
|
|
+ "type": "function"
|
|
|
+ }
|
|
|
+ ]
|
|
|
+ },
|
|
|
+ "response_metadata": {
|
|
|
+ "finish_reason": "tool_calls",
|
|
|
+ "model_name": "qwen3-235b-a22b"
|
|
|
+ },
|
|
|
+ "type": "ai",
|
|
|
+ "id": "run--6144e643-450a-4ec2-932a-230f666dd90c-0",
|
|
|
+ "tool_calls": [
|
|
|
+ {
|
|
|
+ "name": "run_sql",
|
|
|
+ "args": {
|
|
|
+ "sql": "SELECT COUNT(*) AS 餐饮档口数量 \nFROM bss_branch b \nJOIN bss_service_area sa ON b.service_area_id = sa.id \nWHERE sa.service_area_name = '南城服务区' \nAND b.classify = '餐饮' \nAND b.delete_ts IS NULL;",
|
|
|
+ "question": "请问这个服务区有几个餐饮档口？"
|
|
|
+ },
|
|
|
+ "id": "call_b0600de998f5467d96a3f2",
|
|
|
+ "type": "tool_call"
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "invalid_tool_calls": []
|
|
|
+ }
|
|
|
+ },
|
|
|
+ {
|
|
|
+ "lc": 1,
|
|
|
+ "type": "constructor",
|
|
|
+ "id": [
|
|
|
+ "langchain",
|
|
|
+ "schema",
|
|
|
+ "messages",
|
|
|
+ "ToolMessage"
|
|
|
+ ],
|
|
|
+ "kwargs": {
|
|
|
+ "content": "[{\"\\u9910\\u996e\\u6863\\u53e3\\u6570\\u91cf\":6}]",
|
|
|
+ "type": "tool",
|
|
|
+ "name": "run_sql",
|
|
|
+ "id": "488da67c-ea4f-46f4-a201-d1bb40c25931",
|
|
|
+ "tool_call_id": "call_b0600de998f5467d96a3f2",
|
|
|
+ "status": "success"
|
|
|
+ }
|
|
|
+ }
|
|
|
+ ],
|
|
|
+ "user_id": "wang1",
|
|
|
+ "thread_id": "wang1:20250729235038043",
|
|
|
+ "suggested_next_step": "summarize_final_answer",
|
|
|
+ "branch:to:agent": null
|
|
|
+ },
|
|
|
+ "channel_versions": {
|
|
|
+ "__start__": "00000000000000000000000000000018.0.2372340570136553",
|
|
|
+ "messages": "00000000000000000000000000000033.0.3402020632542543",
|
|
|
+ "user_id": "00000000000000000000000000000018.0.6157289048059329",
|
|
|
+ "thread_id": "00000000000000000000000000000018.0.3438965440425772",
|
|
|
+ "suggested_next_step": "00000000000000000000000000000034.0.4358036384739292",
|
|
|
+ "branch:to:agent": "00000000000000000000000000000034.0.30470240061347376",
|
|
|
+ "branch:to:prepare_tool_input": "00000000000000000000000000000032.0.683834676177241",
|
|
|
+ "branch:to:tools": "00000000000000000000000000000033.0.4617350256378949",
|
|
|
+ "branch:to:update_state_after_tool": "00000000000000000000000000000034.0.7301276046246093",
|
|
|
+ "branch:to:format_final_response": "00000000000000000000000000000016.0.4685210381789543"
|
|
|
+ },
|
|
|
+ "
|
|
|
+```
|
|
|
+
|
wangxq