feat: 添加一键启动脚本并更新项目配置

- 添加 start_all.sh 一键启动脚本，支持启动llama-server和FastAPI服务
- 修改启动脚本使用venv虚拟环境替代conda环境
- 更新README.md，添加一键启动脚本使用说明
- 更新py_tree_generator.py，添加final_prompt返回字段
- 禁用Qwen3模型的思考功能
- 添加RAG检索结果的终端打印
- 移除ROS2相关代码（ros2_client.py已删除）

README.md

@@ -19,6 +19,7 @@
 │   │   │   └── classifier_prompt.txt     # 指令简单/复杂分类提示词
 │   │   ├── ...
 │   ├── generated_visualizations/ # 存放最新生成的py_tree可视化图像
+│   ├── generated_reasoning_content/ # 存放最新推理链Markdown（<plan_id>.md）
 │   └── requirements.txt          # 后端服务的Python依赖
 │
 ├── tools/
@@ -26,7 +27,8 @@
 │   ├── knowledge_base/           # 【处理后】存放build_knowledge_base.py生成的.ndjson文件
 │   ├── vector_store/             # 【数据库】存放最终的ChromaDB向量数据库
 │   ├── build_knowledge_base.py   # 【步骤1】用于将原始数据转换为自然语言知识
-│   └── ingest.py                 # 【步骤2】用于将自然语言知识摄入向量数据库
+│   ├── ingest.py                 # 【步骤2】用于将自然语言知识摄入向量数据库
+│   └── test_llama_server.py      # 直接调用本地8081端口llama-server，支持 --system / --system-file
 │
 ├── /             # ROS2接口定义 (保持不变)
 └── docs/
@@ -95,6 +97,10 @@
 
 通用API Key：`OPENAI_API_KEY`
 
+推理链捕获相关：
+- `ENABLE_REASONING_CAPTURE`：是否允许模型返回含有 <think> 的原文以便捕获推理链；默认 true。
+- `REASONING_PREVIEW_LINES`：在后端日志中打印推理链预览的行数；默认 20。
+
 示例：
 ```bash
 export CLASSIFIER_MODEL="qwen2.5-1.8b-instruct"
@@ -104,6 +110,10 @@ export CLASSIFIER_BASE_URL="http://$ORIN_IP:8081/v1"
 export SIMPLE_BASE_URL="http://$ORIN_IP:8081/v1"
 export COMPLEX_BASE_URL="http://$ORIN_IP:8081/v1"
 export OPENAI_API_KEY="sk-no-key-required"
+
+# 推理链捕获（可选）
+export ENABLE_REASONING_CAPTURE=true        # 默认已为true；如需关闭，设置为 false
+export REASONING_PREVIEW_LINES=30          # 调整日志预览行数
 ```
 
 ### 测试简单模式
@@ -117,6 +127,22 @@ python test_api.py
 
 示例输入：“简单模式，起飞” 或 “起飞到10米”。返回结果为简单JSON（无 `root`）：包含 `mode`、`action`、`plan_id`、`visualization_url`。
 
+### 直接调用 llama-server（绕过后端）
+
+当仅需测试本地 8081 端口的推理服务（OpenAI 兼容接口）时，可使用内置脚本：
+
+```bash
+python tools/test_llama_server.py \
+  --system-file backend_service/src/prompts/system_prompt.txt \
+  --user "起飞到10米然后降落" \
+  --base-url "http://127.0.0.1:8081/v1" \
+  --verbose
+```
+
+说明：
+- 支持 `--system` 或 `--system-file` 自定义提示词文件；`--system-file` 优先。
+- 默认解析 OpenAI 风格返回，若包含 `<think>` 推理内容会显示在输出中（具体取决于模型和服务配置）。
+
 ---
 
 ## 工作流程
@@ -202,14 +228,96 @@ python ingest.py
 
 完成前两个阶段后，即可启动并测试后端服务。
 
-#### 1. 启动后端服务
+#### 1. 启动所有服务（推荐方式：一键启动脚本）
 
-启动服务的关键在于**按顺序激活环境**：先激活ROS 2工作空间，再激活Conda环境。
+我们提供了一个一键启动脚本 `start_all.sh`，可以自动启动所有必需的服务：
 
 ```bash
 # 1. 切换到项目根目录
 cd /path/to/your/drone
 
+# 2. 使用一键启动脚本（推荐）
+./start_all.sh start
+
+# 或者直接运行（start是默认命令）
+./start_all.sh
+```
+
+**脚本功能：**
+- 自动启动推理模型服务（llama-server，端口8081）
+- 自动启动Embedding模型服务（llama-server，端口8090）
+- 自动启动FastAPI后端服务（端口8000）
+- 自动检查端口占用、模型文件、环境配置等
+- 自动等待服务就绪
+- 统一管理日志文件（保存在 `logs/` 目录）
+
+**环境变量配置（可选）：**
+
+在运行脚本前，可以通过环境变量自定义配置：
+
+```bash
+# 设置llama-server路径（如果不在默认位置）
+export LLAMA_SERVER_DIR="/path/to/llama.cpp/build/bin"
+
+# 设置模型路径（如果不在默认位置）
+export INFERENCE_MODEL="~/models/gguf/Qwen/Qwen3-8B-GGUF/Qwen3-8B-Q4_K_M.gguf"
+export EMBEDDING_MODEL="~/models/gguf/Qwen/Qwen3-embedding-4B/Qwen3-Embedding-4B-Q4_K_M.gguf"
+
+# 设置Conda环境名称（如果使用不同的环境名）
+export CONDA_ENV="backend"
+
+# 然后运行脚本
+./start_all.sh
+```
+
+**脚本命令：**
+
+```bash
+./start_all.sh start    # 启动所有服务（默认）
+./start_all.sh stop     # 停止所有服务
+./start_all.sh restart  # 重启所有服务
+./start_all.sh status   # 查看服务状态
+```
+
+**日志查看：**
+
+所有服务的日志都保存在 `logs/` 目录下：
+```bash
+# 查看所有日志
+tail -f logs/*.log
+
+# 查看特定服务日志
+tail -f logs/inference_model.log  # 推理模型
+tail -f logs/embedding_model.log  # Embedding模型
+tail -f logs/fastapi.log          # FastAPI服务
+```
+
+#### 2. 手动启动服务（备选方式）
+
+如果您需要手动控制每个服务的启动，可以按照以下步骤操作：
+
+**启动推理模型服务：**
+
+```bash
+cd /llama.cpp/build/bin
+./llama-server -m ~/models/gguf/Qwen/Qwen3-8B-GGUF/Qwen3-8B-Q4_K_M.gguf --port 8081 --gpu-layers 36 --host 0.0.0.0 -c 8192
+```
+
+**启动Embedding模型服务：**
+
+在另一个终端中：
+```bash
+cd /llama.cpp/build/bin
+./llama-server -m ~/models/gguf/Qwen/Qwen3-embedding-4B/Qwen3-Embedding-4B-Q4_K_M.gguf --gpu-layers 36 --port 8090 --embeddings --pooling last --host 0.0.0.0
+```
+
+**启动FastAPI后端服务：**
+
+在第三个终端中：
+```bash
+# 1. 切换到项目根目录
+cd /path/to/your/drone
+
 # 2. 激活ROS 2编译环境
 #    作用：将我们编译好的`drone_interfaces`包的路径告知系统，否则Python会报`ModuleNotFoundError`。
 #    注意：此命令必须在每次打开新终端时执行一次。
@@ -222,6 +330,7 @@ conda activate backend
 cd backend_service/
 uvicorn src.main:app --host 0.0.0.0 --port 8000
 ```
+
 当您看到日志中出现 `Uvicorn running on http://0.0.0.0:8000` 时，表示服务已成功启动。
 
 #### 2. 运行API接口测试

backend_service/generated_reasoning_content/reasoning_content.md

@@ -0,0 +1,43 @@
+好的，我现在需要处理用户的任务指令，生成一个符合要求的行为树JSON。首先，我得仔细阅读用户的指令和参考知识，确保理解所有要求。
+
+用户的目标是搜索并锁定危险性最高的气球（红色>蓝色>绿色），然后跟踪30秒后进行打击。参考知识中提到了三个地点，但用户可能希望搜索的区域是这些地点附近，或者需要根据这些坐标来规划路径。
+
+首先，我需要确定无人机的起降点。参考知识中的地点坐标可能作为搜索区域的中心。比如，用户提到的“学生宿舍”坐标是(5,3,2)，但可能需要选择一个合适的中心点进行搜索。不过用户没有明确指定搜索区域，可能需要使用search_pattern来覆盖这些区域，或者直接在某个中心点周围搜索。
+
+接下来，任务流程应该是：起飞→飞往搜索区域→搜索目标→检测到目标后跟踪→打击。同时必须包含安全监控。
+
+根据用户提供的参考知识，可能需要将搜索区域设置为某个中心点。比如，用户提到的“跷跷板”在(x:15, y:-8.5, z:1.2)，但z坐标需要调整到至少1米，这里已经是1.2，没问题。或者可能选择其他地点作为搜索中心。但用户没有明确说明，可能需要假设搜索区域是这些地点附近，或者使用一个综合的中心点。
+
+不过用户可能希望无人机先飞往某个特定的搜索区。比如，参考知识中的“学生宿舍”可能是一个可能的区域，但需要确认。或者用户可能希望无人机在某个中心点周围进行搜索。例如，使用search_pattern的中心点可能选在某个已知地点，比如“跷跷板”附近，或者综合多个点。
+
+但用户没有明确指定搜索区域，所以可能需要使用search_pattern的中心点为某个已知地点，比如“跷跷板”的坐标，或者选择一个合理的中心点。例如，假设搜索区域是“跷跷板”所在的位置，那么中心坐标为(15, -8.5, 1.2)。或者可能需要将搜索区域设置为多个地点的组合，但用户没有说明，所以可能需要选择一个中心点。
+
+接下来，考虑使用search_pattern来搜索，因为目标位置未知。参数中需要指定pattern_type，比如spiral或grid。假设选择spiral模式，覆盖更大的区域。中心点可能选在某个已知地点，比如“跷跷板”的坐标，或者综合多个点。但用户没有明确，所以可能需要选择一个合理的中心点，比如“跷跷板”的坐标。
+
+然后，检测到目标后，需要跟踪30秒，再打击。所以流程是：起飞→飞往搜索区→搜索→检测→跟踪→打击→降落。
+
+安全监控必须包含在Parallel节点中，监控电池和GPS，并在紧急情况下返航。
+
+现在，构建行为树的结构：
+
+根节点是Parallel，包含主任务和安全监控。主任务是Sequence，包含起飞、飞往搜索区、搜索、检测、跟踪、打击、降落。安全监控是Selector，包含电池和GPS条件，以及紧急处理Sequence。
+
+在搜索部分，使用search_pattern，参数包括pattern_type为spiral，中心点选在某个已知地点，比如“跷跷板”的坐标(15, -8.5, 1.2)，半径可能设为50米，覆盖较大区域。目标类别是balloon，描述为“红色气球危险性高于蓝色气球高于绿色气球”，但参数中description需要描述目标属性，比如“红>蓝>绿”。
+
+检测到目标后，使用track_object，参数target_class为balloon，description为“红>蓝>绿”，track_time为30秒。然后strike_target，最后评估战损。
+
+需要确保所有参数符合约束，比如z坐标在1-5000米之间，半径在5-1000米之间，速度等参数在合理范围。
+
+现在，将各个步骤的参数填入JSON中。例如，起飞到2米，飞往搜索区的中心点可能选在“跷跷板”的坐标(15, -8.5, 1.2)，但z需要调整到至少1米，这里已经是1.2。所以fly_to_waypoint的参数x=15，y=-8.5，z=1.2，接受半径2米。
+
+搜索使用search_pattern，pattern_type为spiral，中心点为(15, -8.5, 1.2)，半径设为50米，覆盖区域。目标类别是balloon，description为“红>蓝>绿”。
+
+检测到目标后，使用track_object，参数target_class为balloon，description为“红>蓝>绿”，track_time为30秒，min_confidence设为0.7，safe_distance设为15米。
+
+然后strike_target，参数target_class为balloon，description为“红>蓝>绿”，count为1。
+
+最后，降落。
+
+安全监控部分，电池阈值设为0.35，GPS的min_satellites设为8，紧急处理Sequence包括emergency_return和land。
+
+现在，将所有步骤的参数填入JSON中，确保符合所有规则。

backend_service/requirements.txt

@@ -15,8 +15,8 @@ chromadb>=0.4.0
 # Visualization
 graphviz>=0.20.0
 
-# ROS 2 Python Client
-rclpy>=0.0.1
+# ROS 2 Python Client - 已注释，项目已与ROS2解耦
+# rclpy>=0.0.1
 
 # Document Processing
 unstructured[all]>=0.11.0
@@ -30,10 +30,10 @@ rich>=13.7.0
 # Type Hints Support
 typing-extensions>=4.8.0
 
-# ROS 2 Build Dependencies
-empy==3.3.4
-catkin-pkg>=0.4.0
-lark>=1.1.0
-colcon-common-extensions>=0.3.0
-vcstool>=0.2.0
-rosdep>=0.22.0
+# ROS 2 Build Dependencies - 已注释，项目已与ROS2解耦
+# empy==3.3.4
+# catkin-pkg>=0.4.0
+# lark>=1.1.0
+# colcon-common-extensions>=0.3.0
+# vcstool>=0.2.0
+# rosdep>=0.22.0

backend_service/src/main.py

@@ -3,13 +3,13 @@ import os
 from fastapi import FastAPI, WebSocket, WebSocketDisconnect
 from fastapi.staticfiles import StaticFiles
 import logging
-import threading
-import rclpy
+# import threading  # ROS2相关，已注释
+# import rclpy  # ROS2相关，已注释
 
 from .models import GeneratePlanRequest, ExecuteMissionRequest
 from .websocket_manager import websocket_manager
 from .py_tree_generator import py_tree_generator
-from .ros2_client import MissionActionClient
+# from .ros2_client import MissionActionClient  # ROS2相关，已注释
 
 # --- Application Setup ---
 app = FastAPI(
@@ -23,14 +23,15 @@ static_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), '..', 'gene
 app.mount("/static", StaticFiles(directory=static_dir), name="static")
 
 # --- ROS2 Node and Client Initialization ---
-rclpy.init()
-ros2_client = MissionActionClient()
+# ROS2相关代码已注释，项目已与ROS2解耦
+# rclpy.init()
+# ros2_client = MissionActionClient()
 
-def run_ros2_node():
-    """Spins the ROS2 node in a dedicated thread."""
-    logging.info("Starting to spin ROS2 node...")
-    rclpy.spin(ros2_client)
-    logging.info("ROS2 node has stopped spinning.")
+# def run_ros2_node():
+#     """Spins the ROS2 node in a dedicated thread."""
+#     logging.info("Starting to spin ROS2 node...")
+#     rclpy.spin(ros2_client)
+#     logging.info("ROS2 node has stopped spinning.")
 
 # --- API Endpoints ---
 
@@ -49,9 +50,12 @@ async def generate_plan_endpoint(request: GeneratePlanRequest):
 async def execute_mission_endpoint(request: ExecuteMissionRequest):
     """
     Receives a `py_tree.json` and sends it to the drone for execution.
+    ROS2相关功能已注释，项目已与ROS2解耦。
     """
-    ros2_client.send_goal(request.py_tree)
-    return {"status": "execution_started"}
+    # ROS2相关代码已注释
+    # ros2_client.send_goal(request.py_tree)
+    logging.warning("execute_mission endpoint called but ROS2 is disabled. Mission execution is not available.")
+    return {"status": "execution_disabled", "message": "ROS2 integration is disabled. Mission execution is not available."}
 
 @app.websocket("/ws/status")
 async def websocket_endpoint(websocket: WebSocket):
@@ -73,21 +77,23 @@ async def websocket_endpoint(websocket: WebSocket):
 async def startup_event():
     """
     On startup, get the current asyncio event loop and pass it to the websocket manager.
-    Also, start the ROS2 node in a background thread.
+    ROS2相关功能已注释，项目已与ROS2解耦。
     """
     # Configure WebSocket Manager
     loop = asyncio.get_running_loop()
     websocket_manager.set_loop(loop)
     logging.info("WebSocket event loop configured.")
 
+    # ROS2相关代码已注释
     # Start ROS2 node in a background thread
-    ros2_thread = threading.Thread(target=run_ros2_node, daemon=True)
-    ros2_thread.start()
-    logging.info("ROS2 node thread started.")
+    # ros2_thread = threading.Thread(target=run_ros2_node, daemon=True)
+    # ros2_thread.start()
+    # logging.info("ROS2 node thread started.")
 
 @app.on_event("shutdown")
 async def shutdown_event():
     logging.info("Backend service shutting down.")
-    ros2_client.destroy_node()
-    rclpy.shutdown()
-    logging.info("ROS2 node shut down successfully.")
+    # ROS2相关代码已注释
+    # ros2_client.destroy_node()
+    # rclpy.shutdown()
+    # logging.info("ROS2 node shut down successfully.")

backend_service/src/prompts/simple_mode_prompt.txt

@@ -3,14 +3,14 @@
 输出要求（必须遵守）：
 - 只输出一个JSON对象，不要任何解释或多余文本。
 - JSON结构：
-{"mode":"simple","action":{"name":"<action_name>","params":{...}}}
+{"root":{"type":"action","name":"<action_name>","params":{...}}}
 - <action_name> 与参数定义、取值范围，必须与“复杂模式”提示词（system_prompt.txt）中的定义完全一致。
-- 简单模式下不包含任何行为树结构与安全监控并行，仅输出单一原子动作。
+- 简单模式下root节点必须是action类型节点，不能是控制流节点。
 
 示例：
-- “起飞到10米” → {"mode":"simple","action":{"name":"takeoff","params":{"altitude":10.0}}}
-- “移动到(120,80,20)” → {"mode":"simple","action":{"name":"fly_to_waypoint","params":{"x":120.0,"y":80.0,"z":20.0,"acceptance_radius":2.0}}}
-- “飞机自检” → {"mode":"simple","action":{"name":"preflight_checks","params":{"check_level":"comprehensive"}}}
+- “起飞到10米” → {"root":{"type":"action","name":"takeoff","params":{"altitude":10.0}}}
+- “移动到(120,80,20)” → {"root":{"type":"action","name":"fly_to_waypoint","params":{"x":120.0,"y":80.0,"z":20.0,"acceptance_radius":2.0}}}
+- “飞机自检” → {"root":{"type":"action","name":"preflight_checks","params":{"check_level":"comprehensive"}}}
 
 —— 可用节点定义——
 ```json
@@ -21,13 +21,13 @@
     {"name": "fly_to_waypoint", "description": "导航至一个指定坐标点。使用相对坐标系（x,y,z），单位为米。", "params": {"x": "float", "y": "float", "z": "float", "acceptance_radius": "float, 可选，默认2.0"}},
     {"name": "move_direction", "description": "按指定方向直线移动。方向可为绝对方位或相对机体朝向。", "params": {"direction": "string: north|south|east|west|forward|backward|left|right", "distance": "float[1,10000], 可选, 不指定则持续移动"}},
     {"name": "orbit_around_point", "description": "以给定中心点为中心，等速圆周飞行指定圈数。", "params": {"center_x": "float", "center_y": "float", "center_z": "float", "radius": "float[5,1000]", "laps": "int[1,20]", "clockwise": "boolean, 可选, 默认true", "speed_mps": "float[0.5,15], 可选", "gimbal_lock": "boolean, 可选, 默认true"}},
-    {"name": "orbit_around_target", "description": "以目标为中心，等速圆周飞行指定圈数（需已有目标）。", "params": {"target_class": "string, 见复杂模式定义列表", "description": "string, 可选", "radius": "float[5,1000]", "laps": "int[1,20]", "clockwise": "boolean, 可选, 默认true", "speed_mps": "float[0.5,15], 可选", "gimbal_lock": "boolean, 可选, 默认true"}},
+    {"name": "orbit_around_target", "description": "以目标为中心，等速圆周飞行指定圈数（需已有目标）。", "params": {"target_class": "string, 取值同object_detect列表", "description": "string, 可选", "radius": "float[5,1000]", "laps": "int[1,20]", "clockwise": "boolean, 可选, 默认true", "speed_mps": "float[0.5,15], 可选", "gimbal_lock": "boolean, 可选, 默认true"}},
     {"name": "loiter", "description": "在当前位置上空悬停一段时间或直到条件触发。", "params": {"duration": "float, 可选[1,600]", "until_condition": "string, 可选"}},
-    {"name": "object_detect", "description": "识别特定目标对象。", "params": {"target_class": "string, 见复杂模式定义列表", "description": "string, 可选", "count": "int, 可选, 默认1"}},
+    {"name": "object_detect", "description": "识别特定目标对象。一般是用户提到的需要检测的目标；如果用户给出了需要探索的目标的优先级，比如蓝色球危险性大于红色球大于绿色球，需要检测最危险的球，此处应给出检测优先级，描述应当为 '蓝>红>绿'", "params": {"target_class": "string, 要识别的目标类别，必须为以下值之一: balloon,person, bicycle, car, motorcycle, airplane, bus, train, truck, boat, traffic_light, fire_hydrant, stop_sign, parking_meter, bench, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe, backpack, umbrella, handbag, tie, suitcase, frisbee, skis, snowboard, sports_ball, kite, baseball_bat, baseball_glove, skateboard, surfboard, tennis_racket, bottle, wine_glass, cup, fork, knife, spoon, bowl, banana, apple, sandwich, orange, broccoli, carrot, hot_dog, pizza, donut, cake, chair, couch, potted_plant, bed, dining_table, toilet, tv, laptop, mouse, remote, keyboard, cell_phone, microwave, oven, toaster, sink, refrigerator, book, clock, vase, scissors, teddy_bear, hair_drier, toothbrush", "description": "string, 可选", "count": "int, 可选, 默认1"}},
     {"name": "strike_target", "description": "对已识别目标进行打击。", "params": {"target_class": "string", "description": "string, 可选", "count": "int, 可选, 默认1"}},
     {"name": "battle_damage_assessment", "description": "战损评估。", "params": {"target_class": "string", "assessment_time": "float[5-60], 默认15.0"}},
     {"name": "search_pattern", "description": "按模式搜索。", "params": {"pattern_type": "string: spiral|grid", "center_x": "float", "center_y": "float", "center_z": "float", "radius": "float[5,1000]", "target_class": "string", "description": "string, 可选", "count": "int, 可选, 默认1"}},
-    {"name": "track_object", "description": "持续跟踪目标。", "params": {"target_class": "string, 见复杂模式定义列表", "description": "string, 可选", "track_time": "float[1,600], 默认30.0", "min_confidence": "float[0.5-1.0], 默认0.7", "safe_distance": "float[2-50], 默认10.0"}},
+    {"name": "track_object", "description": "持续跟踪目标。", "params": {"target_class": "string, 取值同object_detect列表", "description": "string, 可选", "track_time": "float[1,600], 默认30.0", "min_confidence": "float[0.5-1.0], 默认0.7", "safe_distance": "float[2-50], 默认10.0"}},
     {"name": "deliver_payload", "description": "投放物资。", "params": {"payload_type": "string", "release_altitude": "float[2,100], 默认5.0"}},
     {"name": "preflight_checks", "description": "飞行前系统自检。", "params": {"check_level": "string: basic|comprehensive"}},
     {"name": "emergency_return", "description": "执行紧急返航程序。", "params": {"reason": "string"}}
@@ -53,7 +53,6 @@
 - orbit_around_target.radius: [5, 1000]
 - orbit_around_point/target.laps: [1, 20]
 - orbit_around_point/target.speed_mps: [0.5, 15]
-- 电池阈值等同复杂模式（如需涉及）
 - 若参考知识提供坐标，必须使用并裁剪到约束范围内
 
 —— 口令转化规则（环绕类）——

backend_service/src/prompts/system_prompt.txt

@@ -1,258 +1,55 @@
-你是一个无人机任务规划专家。你的唯一任务是根据用户提供的任务指令和参考知识，生成一个结构化、可执行的行为树（Pytree）JSON描述。
-你的输出必须是一个严格的、单一的JSON对象，不包含任何形式的解释、总结或自然语言描述。
+任务：根据用户任意任务指令，生成结构化可执行的无人机行为树（Pytree）JSON。**仅输出单一JSON对象，无任何自然语言、注释或额外内容**。
 
-#### 1. 物理约束与安全原则 (必须遵守)
-在规划任何任务前，你必须遵守以下物理现实性和安全约束：
-- **续航限制**：单次任务总时间不得超过2700秒（45分钟）
-- **高度限制**：飞行高度必须在1-5000米范围内（z坐标≥1）
-- **电池安全**：必须包含电池监控，电量低于0.3触发返航，低于0.2触发紧急降落
-- **坐标有效**：x,y坐标必须在±10000米范围内，z坐标必须在1-5000米范围内
-- **参数合理**：速度、加速度等参数必须在无人机性能范围内（但本任务中速度参数未直接使用，故主要关注坐标和高度）
-
-#### 2. 可用节点定义 (必须遵守)
-你必须严格从以下JSON定义的列表中选择节点来构建行为树。不允许使用任何未定义的节点。
+## 一、核心节点定义（格式不可修改，确保后端解析）
+#### 1. 可用节点定义 (必须遵守)
+你必须严格从以下JSON定义的列表中选择节点构建行为树，不允许使用未定义节点：
 ```json
 {
   "actions": [
-    {
-      "name": "takeoff",
-      "description": "无人机从当前位置垂直起飞到指定的海拔高度。",
-      "params": {
-        "altitude": "float, 目标海拔高度(米)，范围[1, 100]，默认为2"
-      }
-    },
-    {
-      "name": "land",
-      "description": "降落无人机。可选择当前位置或返航点降落。",
-      "params": {
-        "mode": "string, 可选值: 'current'（当前位置）, 'home'（返航点）"
-      }
-    },
-    {
-      "name": "fly_to_waypoint",
-      "description": "导航至一个指定坐标点。无人机到达航点后该动作才算完成。使用相对坐标系（x,y,z），单位为米。",
-      "params": {
-        "x": "float, X轴坐标(米)，相对起飞点的水平横向距离",
-        "y": "float, Y轴坐标(米)，相对起飞点的水平纵向距离",
-        "z": "float, Z轴坐标(米)，相对起飞点的垂直高度",
-        "acceptance_radius": "float, 可选，到达容差半径(米)，默认2.0"
-      }
-    },
-    {
-      "name": "move_direction",
-      "description": "按指定方向直线移动。方向可为绝对方位或相对机体朝向。",
-      "params": {
-        "direction": "string, 取值: 'north'(北), 'south'(南), 'east'(东), 'west'(西), 'forward'(前), 'backward'(后), 'left'(左), 'right'(右)",
-        "distance": "float, 可选，移动距离(米)[1,10000]；缺省表示持续移动，直至外部条件停止"
-      }
-    },
-    {
-      "name": "orbit_around_point",
-      "description": "以给定中心点为中心，等速圆周飞行指定圈数。",
-      "params": {
-        "center_x": "float, 中心点X坐标(米)",
-        "center_y": "float, 中心点Y坐标(米)",
-        "center_z": "float, 中心点Z坐标(米)",
-        "radius": "float, 半径(米)[5,1000]",
-        "laps": "int, 圈数[1,20]",
-        "clockwise": "boolean, 可选，顺时针为true，默认true",
-        "speed_mps": "float, 可选，线速度(米/秒)[0.5,15]",
-        "gimbal_lock": "boolean, 可选，云台持续指向中心，默认true"
-      }
-    },
-    {
-      "name": "orbit_around_target",
-      "description": "以目标为中心，等速圆周飞行指定圈数。需要已确认目标。",
-      "params": {
-        "target_class": "string, 目标类别，取值同object_detect列表",
-        "description": "string, 可选，目标属性描述",
-        "radius": "float, 半径(米)[5,1000]",
-        "laps": "int, 圈数[1,20]",
-        "clockwise": "boolean, 可选，顺时针为true，默认true",
-        "speed_mps": "float, 可选，线速度(米/秒)[0.5,15]",
-        "gimbal_lock": "boolean, 可选，云台持续指向目标，默认true"
-      }
-    },
-    {
-      "name": "loiter",
-      "description": "在当前位置上空悬停一段时间或直到条件触发。",
-      "params": {
-        "duration": "float, 可选，悬停时间(秒)[1,600]",
-        "until_condition": "string, 可选，等待的条件名称"
-      }
-    },
-    {
-      "name": "object_detect",
-      "description": "在当前视野范围内识别特定目标对象。适用于定点检测，无人机应在目标大致位置悬停或保持稳定姿态。",
-      "params": {
-        "target_class": "string, 要识别的目标类别，必须为以下值之一: person, bicycle, car, motorcycle, airplane, bus, train, truck, boat, traffic_light, fire_hydrant, stop_sign, parking_meter, bench, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe, backpack, umbrella, handbag, tie, suitcase, frisbee, skis, snowboard, sports_ball, kite, baseball_bat, baseball_glove, skateboard, surfboard, tennis_racket, bottle, wine_glass, cup, fork, knife, spoon, bowl, banana, apple, sandwich, orange, broccoli, carrot, hot_dog, pizza, donut, cake, chair, couch, potted_plant, bed, dining_table, toilet, tv, laptop, mouse, remote, keyboard, cell_phone, microwave, oven, toaster, sink, refrigerator, book, clock, vase, scissors, teddy_bear, hair_drier, toothbrush",
-        "description": "string, 可选，目标属性描述（如颜色、状态等）",
-        "count": "int, 可选，需要检测的目标个数，默认1"
-      }
-    },
-    {
-      "name": "strike_target",
-      "description": "对已识别的目标进行打击。必须先使用object_detect成功识别目标后才能使用此动作。",
-      "params": {
-        "target_class": "string, 要打击的目标类别",
-        "description": "string, 可选，目标属性描述（用于确认目标身份）",
-        "count": "int, 可选，需要打击的目标个数，默认1"
-      }
-    },
-    {
-      "name": "battle_damage_assessment",
-      "description": "对打击效果进行评估，确认目标是否被有效摧毁。",
-      "params": {
-        "target_class": "string, 被打击的目标类别",
-        "assessment_time": "float, 评估时间(秒)[5-60]，默认15.0"
-      }
-    },
-    {
-      "name": "search_pattern",
-      "description": "通过执行一个系统性的移动搜索模式，在指定区域内寻找特定目标。无人机会持续移动并分析视频流，直到找到目标或完成整个搜索模式。适用于在未知区域发现目标。",
-      "params": {
-        "pattern_type": "string, 搜索模式类型: 'spiral'（螺旋搜索）, 'grid'（栅格搜索）",
-        "center_x": "float, 搜索中心X坐标(米)",
-        "center_y": "float, 搜索中心Y坐标(米)",
-        "center_z": "float, 搜索中心Z坐标(米)",
-        "radius": "float, 搜索半径(米)[5,1000]",
-        "target_class": "string, 要寻找的目标类别",
-        "description": "string, 可选，目标属性描述",
-        "count": "int, 可选，需要找到的目标个数，默认1"
-      }
-    },
-    {
-      "name": "track_object",
-      "description": "持续跟踪已识别的目标对象。无人机将保持对目标的视觉锁定并跟随其移动。必须先使用object_detect成功识别目标后才能使用此动作。",
-      "params": {
-        "target_class": "string, 要跟踪的目标类别，必须为以下值之一: person, bicycle, car, motorcycle, airplane, bus, train, truck, boat, traffic_light, fire_hydrant, stop_sign, parking_meter, bench, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe, backpack, umbrella, handbag, tie, suitcase, frisbee, skis, snowboard, sports_ball, kite, baseball_bat, baseball_glove, skateboard, surfboard, tennis_racket, bottle, wine_glass, cup, fork, knife, spoon, bowl, banana, apple, sandwich, orange, broccoli, carrot, hot_dog, pizza, donut, cake, chair, couch, potted_plant, bed, dining_table, toilet, tv, laptop, mouse, remote, keyboard, cell_phone, microwave, oven, toaster, sink, refrigerator, book, clock, vase, scissors, teddy_bear, hair_drier, toothbrush",
-        "description": "string, 可选，目标属性描述（如颜色、状态等）",
-        "track_time": "float, 跟踪持续时间(秒)[1,600]，默认30.0",
-        "min_confidence": "float, 可选，跟踪置信度阈值[0.5-1.0]，默认0.7",
-        "safe_distance": "float, 可选，保持的安全距离(米)[2-50]，默认10.0"
-      }
-    },
-    {
-      "name": "deliver_payload",
-      "description": "投放携带的物资。",
-      "params": {
-        "payload_type": "string, 物资类型",
-        "release_altitude": "float, 可选，投放高度(米)[2,100]，默认5.0"
-      }
-    },
-    {
-      "name": "preflight_checks",
-      "description": "执行飞行前系统自检。",
-      "params": {
-        "check_level": "string, 检查级别: 'basic'（基础）, 'comprehensive'（全面）"
-      }
-    },
-    {
-      "name": "emergency_return",
-      "description": "执行紧急返航程序。",
-      "params": {
-        "reason": "string, 紧急返航原因"
-      }
-    }
+    {"name":"takeoff","params":{"altitude":"float[1,100]，默认2"}},
+    {"name":"land","params":{"mode":"'current'/'home'"}},
+    {"name":"fly_to_waypoint","params":{"x":"±10000","y":"±10000","z":"[1,5000]","acceptance_radius":"默认2.0"}},
+    {"name":"move_direction","params":{"direction":"north/south/east/west/forward/backward/left/right","distance":"[1,10000]，缺省持续移动"}},
+    {"name":"orbit_around_point","params":{"center_x":"±10000","center_y":"±10000","center_z":"[1,5000]","radius":"[5,1000]","laps":"[1,20]","clockwise":"默认true","speed_mps":"[0.5,15]","gimbal_lock":"默认true"}},
+    {"name":"orbit_around_target","params":{"target_class":"见object_detect列表","description":"可选，目标属性","radius":"[5,1000]","laps":"[1,20]","clockwise":"默认true","speed_mps":"[0.5,15]","gimbal_lock":"默认true"}},
+    {"name":"loiter","params":{"duration":"[1,600]秒/until_condition:可选"}},
+    {"name":"object_detect","params":{"target_class":"person,bicycle,car,motorcycle,airplane,bus,train,truck,boat,traffic_light,fire_hydrant,stop_sign,parking_meter,bench,bird,cat,dog,horse,sheep,cow,elephant,bear,zebra,giraffe,backpack,umbrella,handbag,tie,suitcase,frisbee,skis,snowboard,sports_ball,kite,baseball_bat,baseball_glove,skateboard,surfboard,tennis_racket,bottle,wine_glass,cup,fork,knife,spoon,bowl,banana,apple,sandwich,orange,broccoli,carrot,hot_dog,pizza,donut,cake,chair,couch,potted_plant,bed,dining_table,toilet,tv,laptop,mouse,remote,keyboard,cell_phone,microwave,oven,toaster,sink,refrigerator,book,clock,vase,scissors,teddy_bear,hair_drier,toothbrush","description":"可选，","count":"默认1"}},
+    {"name":"strike_target","params":{"target_class":"同object_detect","description":"可选，目标属性","count":"默认1"}},
+    {"name":"battle_damage_assessment","params":{"target_class":"同object_detect","assessment_time":"[5,60]，默认15"}},
+    {"name":"search_pattern","params":{"pattern_type":"spiral/grid","center_x":"±10000","center_y":"±10000","center_z":"[1,5000]","radius":"[5,1000]","target_class":"同object_detect","description":"可选，目标属性","count":"默认1"}},
+    {"name":"track_object","params":{"target_class":"同object_detect","description":"可选，目标属性","track_time":"[1,600]秒（必传，不可用'duration'）","min_confidence":"[0.5,1.0]默认0.7","safe_distance":"[2,50]默认10"}},
+    {"name":"deliver_payload","params":{"payload_type":"string","release_altitude":"[2,100]默认5"}},
+    {"name":"preflight_checks","params":{"check_level":"basic/comprehensive"}},
+    {"name":"emergency_return","params":{"reason":"string"}}
   ],
   "conditions": [
-    {
-      "name": "battery_above",
-      "description": "检查电池电量是否高于指定阈值。",
-      "params": {
-        "threshold": "float, 电量阈值百分比[0.0,1.0]"
-      }
-    },
-    {
-      "name": "at_waypoint",
-      "description": "检查无人机是否在指定坐标点的容差范围内。使用相对坐标系（x,y,z），单位为米。",
-      "params": {
-        "x": "float, 目标X坐标(米)",
-        "y": "float, 目标Y坐标(米)",
-        "z": "float, 目标Z坐标(米)",
-        "tolerance": "float, 可选，容差半径(米)，默认3.0"
-      }
-    },
-    {
-      "name": "object_detected",
-      "description": "检查是否检测到特定目标对象。可用于验证 object_detect 或 search_pattern 的结果，也可作为打击的前提条件。",
-      "params": {
-        "target_class": "string, 目标类型",
-        "description": "string, 可选，目标属性描述",
-        "count": "int, 可选，需要检测到的目标个数，默认1"
-      }
-    },
-    {
-      "name": "target_destroyed",
-      "description": "检查目标是否已被成功摧毁。用于战损评估后的确认。",
-      "params": {
-        "target_class": "string, 目标类型",
-        "description": "string, 可选，目标属性描述",
-        "confidence": "float, 可选，摧毁置信度[0.5-1.0]，默认0.8"
-      }
-    },
-    {
-      "name": "time_elapsed",
-      "description": "检查自任务开始是否经过指定时间。",
-      "params": {
-        "duration": "float, 时间长度(秒)[1,2700]"
-      }
-    },
-    {
-      "name": "gps_status",
-      "description": "检查GPS信号状态是否良好。",
-      "params": {
-        "min_satellites": "int, 最小卫星数量[6,15]，默认10"
-      }
-    }
+    {"name":"battery_above","params":{"threshold":"[0.0,1.0]，必传"}},
+    {"name":"at_waypoint","params":{"x":"±10000","y":"±10000","z":"[1,5000]","tolerance":"默认3.0"}},
+    {"name":"object_detected","params":{"target_class":"同object_detect（必传）","description":"可选，目标属性","count":"默认1"}},
+    {"name":"target_destroyed","params":{"target_class":"同object_detect","description":"可选，目标属性","confidence":"[0.5,1.0]默认0.8"}},
+    {"name":"time_elapsed","params":{"duration":"[1,2700]秒"}},
+    {"name":"gps_status","params":{"min_satellites":"int[6,15]，必传（如8）"}}
   ],
   "control_flow": [
-    {
-      "name": "Sequence",
-      "description": "序列节点，按顺序执行其子节点。只有当所有子节点都成功时，它才成功。",
-      "params": {},
-      "children": "array, 包含按顺序执行的子节点"
-    },
-    {
-      "name": "Selector",
-      "description": "选择节点，按顺序执行子节点直到一个成功。如果所有子节点都失败，则失败。",
-      "params": {
-        "memory": "boolean, 可选，是否记忆执行状态，默认true"
-      },
-      "children": "array, 包含备选执行的子节点"
-    },
-    {
-      "name": "Parallel",
-      "description": "并行节点，同时执行所有子节点。支持不同的成功策略。",
-      "params": {
-        "policy": "string, 成功策略: 'all_success'（全部成功）, 'one_success'（一个成功）"
-      },
-      "children": "array, 包含并行执行的子节点"
-    }
+    {"name":"Sequence","params":{},"children":"子节点数组（按序执行，全成功则成功）"},
+    {"name":"Selector","params":{"memory":"默认true"},"children":"子节点数组（执行到成功为止）"},
+    {"name":"Parallel","params":{"policy":"all_success"},"children":"子节点数组（同时执行，严禁用'one_success'）"}
   ]
 }
 ```
 
-#### 3. JSON结构规范 (必须遵守)
-生成的JSON对象必须有一个名为`root`的键，其值是一个有效的行为树节点对象。每个节点都必须包含正确的字段。
-- **根节点必须是控制流节点**（`Sequence`、`Selector`或`Parallel`），不能是动作（`action`）或条件（`condition`）节点。
-- **动作节点和条件节点是叶子节点**，不能有`children`字段。
-- 控制流节点必须有`children`字段，且其值是一个子节点数组。
-- **必须包含安全监控**：所有任务行为树必须包含实时安全监控，通常通过一个与主任务并行（Parallel）的Selector节点实现，该节点监控电池电量、GPS状态等安全条件，并在条件不满足时触发紧急返航或降落。
-- 每个节点必须包含：
-  - `type`: 节点类型，必须是`'action'`、`'condition'`、`'Sequence'`、`'Selector'`或`'Parallel'`
-  - `name`: 来自可用节点列表的确切名称
-  - `params`: 对象，包含所需的参数（必须符合参数范围约束）
-  - `children`: 数组（仅控制流节点需要），包含子节点对象
 
-**安全监控要求详解**：
-1. **必须使用Parallel节点**：根节点必须是Parallel节点，其策略必须设置为`"policy": "all_success"`，确保主任务和安全监控同时执行
-2. **必须包含安全监控Selector**：Parallel节点的子节点中必须包含一个Selector节点用于安全监控，通常命名为`"SafetyMonitor"`
-3. **必须包含电池监控**：安全监控Selector必须包含`battery_above`条件节点，监控电池电量
-4. **必须包含GPS监控**：安全监控Selector应该包含`gps_status`条件节点，监控GPS信号状态
-5. **必须包含紧急处理流程**：安全监控Selector必须包含紧急处理Sequence，在安全条件不满足时执行紧急返航和降落
+## 二、节点必填字段（后端Schema强制要求，缺一验证失败）
+每个节点必须包含以下字段，字段名/类型不可自定义：
+1. **`type`**：  
+   - 动作节点→`"action"`，条件节点→`"condition"`，控制流节点→`"Sequence"`/`"Selector"`/`"Parallel"`（与`name`字段值完全一致）；
+2. **`name`**：必须是上述JSON中`actions`/`conditions`/`control_flow`下的`name`值（如“gps_status”不可错写为“gps_check”）；
+3. **`params`**：严格匹配上述节点的`params`定义，无自定义参数（如优先级排序不可加“priority”字段，仅用`description`）；
+4. **`children`**：仅控制流节点必含（子节点数组），动作/条件节点无此字段。
 
-**正确示例**：
+
+## 三、行为树固定结构（通用不变，确保安全验证）
+根节点必须是`Parallel`，`children`含`MainTask`（Sequence）和`SafetyMonitor`（Selector），结构不随任务类型（含优先级排序）修改：
 ```json
 {
   "root": {
@@ -263,9 +60,17 @@
       {
         "type": "Sequence",
         "name": "MainTask",
+        "params": {},
         "children": [
-          // 主任务步骤
-          {"type": "action", "name": "land", "params": {"mode": "home"}}
+          // 通用主任务步骤（含优先级排序任务示例，需按用户指令替换）：
+          {"type":"action","name":"preflight_checks","params":{"check_level":"comprehensive"}},
+          {"type":"action","name":"takeoff","params":{"altitude":10.0}},
+          {"type":"action","name":"fly_to_waypoint","params":{"x":200.0,"y":150.0,"z":10.0}}, // 搜索区坐标（用户未给时填合理值）
+          {"type":"action","name":"search_pattern","params":{"pattern_type":"grid","center_x":200.0,"center_y":150.0,"center_z":10.0,"radius":50.0,"target_class":"balloon","description":"红色"}},
+          {"type":"condition","name":"object_detected","params":{"target_class":"balloon","description":"红色"}}, // 确认高优先级目标
+          {"type":"action","name":"track_object","params":{"target_class":"balloon","description":"红色","track_time":30.0}},
+          {"type":"action","name":"strike_target","params":{"target_class":"balloon","description":"红色"}},
+          {"type":"action","name":"land","params":{"mode":"home"}}
         ]
       },
       {
@@ -273,22 +78,15 @@
         "name": "SafetyMonitor",
         "params": {"memory": true},
         "children": [
+          {"type":"condition","name":"battery_above","params":{"threshold":0.3}},
+          {"type":"condition","name":"gps_status","params":{"min_satellites":8}},
           {
-            "type": "condition",
-            "name": "battery_above",
-            "params": {"threshold": 0.3}
-          },
-          {
-            "type": "condition",
-            "name": "gps_status",
-            "params": {"min_satellites": 8}
-          },
-          {
-            "type": "Sequence",
-            "name": "EmergencyHandler",
+            "type":"Sequence",
+            "name":"EmergencyHandler",
+            "params": {},
             "children": [
-              {"type": "action", "name": "emergency_return", "params": {"reason": "safety_breach"}},
-              {"type": "action", "name": "land", "params": {"mode": "home"}}
+              {"type":"action","name":"emergency_return","params":{"reason":"safety_breach"}},
+              {"type":"action","name":"land","params":{"mode":"home"}}
             ]
           }
         ]
@@ -298,464 +96,22 @@
 }
 ```
 
-**错误示例**（缺少安全监控）：
-```json
-{
-  "root": {
-    "type": "Sequence",  // 错误：根节点不是Parallel，无法同时运行安全监控
-    "name": "MainTaskOnly",
-    "children": [
-      // 只有主任务，没有安全监控
-    ]
-  }
-}
-```
 
-错误示例（根节点为动作节点）：
-```json
-{
-  "root": {
-    "type": "action",
-    "name": "land",
-    "children": [ ... ],  // 错误：动作节点不能有子节点
-    "params": {"mode": "home"}
-  }
-}
-```
+## 四、优先级排序任务通用示例
+当用户指令中明确提出有多个待考察且具有优先级关系的物体时，节点描述须为优先级关系。比如当指令为已知有三个气球，危险级关系为红色气球大于蓝色气球大于绿色气球，要求优先跟踪最危险的气球时，节点的描述参考下表情形。
+| 用户指令场景                | `target_class`  | `description`          | 核心节点示例（search_pattern）                                                                 |
+|-----------------------------|-----------------|-------------------------|------------------------------------------------------------------------------------------------|
+| 红气球＞蓝气球＞绿气球      | `balloon`    | `（红＞蓝＞绿）`        | `{"type":"action","name":"search_pattern","params":{"pattern_type":"grid","center_x":200,"center_y":150,"center_z":10,"radius":50,"target_class":"balloon","description":"（红＞蓝＞绿）"}}` |
+| 军用卡车＞民用卡车＞面包车  | `truck`         | `（军用卡车＞民用卡车＞面包车）` | `{"type":"action","name":"object_detect","params":{"target_class":"truck","description":"（军用卡车＞民用卡车＞面包车）"}}` |
 
-##### 重要安全警告：Parallel节点使用禁忌
 
-**严禁**在安全监控场景中使用 `"policy": "one_success"` 的Parallel节点！
+## 五、高频错误规避（确保验证通过）
+1. 优先级排序不可修改`target_class`：如“民用卡车、面包车与军用卡车中，军用卡车优先”，`target_class`仍为`truck`，仅用`description`填排序规则；
+2. 在没有明确指出物体之间的优先级关系情况下，`description`字段只描述物体属性本身，严禁与用户指令中不存在的物体进行排序；
+3. `track_object`必传`track_time`：不可用`duration`替代（如跟踪30秒填`"track_time":30.0`）；
+4. `gps_status`的`min_satellites`必须在6-15之间（如8，不可缺省）；
+5. 无自定义节点：“锁定高优先级目标”需通过`object_detect`+`object_detected`实现，不可用“lock_high_risk_target”。
 
-错误模式（会导致任务中断）：
-```json
-{
-  "type": "Parallel",
-  "params": {"policy": "one_success"},  // 严禁这样使用！
-  "children": [
-    {"type": "Sequence", "name": "MainTask"},  // 主任务会被意外终止
-    {"type": "Selector", "name": "SafetyMonitor"}  // 监控条件成功会杀死主任务
-  ]
-}
-```
 
-#### 4. Selector节点memory参数使用规范
-- **默认使用** `"memory": true`：用于任务执行和监控检查，避免不必要的任务中断。
-- **仅在高优先级安全中断**时使用 `"memory": false`：如急停按钮，每个tick都检查。
-- **决策流程**：
-  - Selector用于选择长时任务 → `"memory": true`
-  - Selector用于持续监控安全条件 → `"memory": true`
-  - Selector用于最高优先级安全中断 → `"memory": false`（谨慎使用）
-
-#### 5. 搜索与检测节点使用区分
-- **object_detect**：用于已知位置的定点检测（无人机悬停或稳定时识别）。
-- **search_pattern**：用于未知区域的移动搜索（无人机按模式飞行覆盖区域）。
-- 严禁混淆使用：例如，在search_pattern后不应立即使用object_detect，除非需要进一步验证。
-
-#### 6. 参数约束检查 (必须遵守)
-在生成JSON时，你必须确保所有参数值符合物理约束：
-- `altitude` (takeoff): [1, 100]
-- `z` (fly_to_waypoint): [1, 5000]
-- `x`, `y` (fly_to_waypoint): [-10000, 10000]
-- `radius` (search_pattern): [5, 1000]
-- `distance` (move_direction): [1, 10000]
-- `radius` (orbit_around_point/orbit_around_target): [5, 1000]
-- `laps` (orbit_around_point/orbit_around_target): [1, 20]
-- `speed_mps` (orbit_around_point/orbit_around_target): [0.5, 15]
-- 电池阈值: [0.0, 1.0]
-- 等等其他参数范围。
-
-如果用户指令或参考知识提供坐标，必须使用这些坐标，但确保调整到约束范围内（例如，如果z<5，则设置为5.0）。
-
-#### 7. 标准任务范式
-所有任务必须包含安全监控。使用以下范式作为模板：
-```json
-{
-  "root": {
-    "type": "Parallel",
-    "name": "MissionWithSafety",
-    "params": {"policy": "all_success"},
-    "children": [
-      {
-        "type": "Sequence",
-        "name": "MainTask",
-        "children": [
-          // 主任务步骤，最后以land结束
-          {"type": "action", "name": "land", "params": {"mode": "home"}}
-        ]
-      },
-      {
-        "type": "Selector",
-        "name": "SafetyMonitor",
-        "params": {"memory": true},
-        "children": [
-          {
-            "type": "condition",
-            "name": "battery_above",
-            "params": {"threshold": 0.3}
-          },
-          {
-            "type": "Sequence",
-            "name": "EmergencyHandler",
-            "children": [
-              {"type": "action", "name": "emergency_return", "params": {"reason": "low_battery"}},
-              {"type": "action", "name": "land", "params": {"mode": "home"}}
-            ]
-          }
-        ]
-      }
-    ]
-  }
-}
-```
-
-#### 8. 打击任务范式
-所有任务必须包含安全监控。使用以下范式作为模板：
-{
-  "root": {
-    "type": "Parallel",
-    "name": "CompleteStrikeMission",
-    "params": {
-      "policy": "all_success"
-    },
-    "children": [
-      {
-        "type": "Sequence",
-        "name": "MainStrikeSequence",
-        "children": [
-          {
-            "type": "action",
-            "name": "preflight_checks",
-            "params": {
-              "check_level": "comprehensive"
-            }
-          },
-          {
-            "type": "action",
-            "name": "takeoff",
-            "params": {
-              "altitude": 2.0
-            }
-          },
-          {
-            "type": "action",
-            "name": "fly_to_waypoint",
-            "params": {
-              "x": 200.0,
-              "y": 150.0,
-              "z": 2.0,
-              "acceptance_radius": 2.0
-            }
-          },
-          {
-            "type": "Selector",
-            "name": "TargetAcquisitionSelector",
-            "params": {
-              "memory": true
-            },
-            "children": [
-              {
-                "type": "Sequence",
-                "name": "DirectDetectionSequence",
-                "children": [
-                  {
-                    "type": "action",
-                    "name": "loiter",
-                    "params": {
-                      "duration": 10.0
-                    }
-                  },
-                  {
-                    "type": "action",
-                    "name": "object_detect",
-                    "params": {
-                      "target_class": "truck",
-                      "description": "军事卡车",
-                      "count": 1
-                    }
-                  },
-                  {
-                    "type": "condition",
-                    "name": "object_detected",
-                    "params": {
-                      "target_class": "truck",
-                      "description": "军事卡车",
-                      "count": 1
-                    }
-                  }
-                ]
-              },
-              {
-                "type": "action",
-                "name": "search_pattern",
-                "params": {
-                  "pattern_type": "grid",
-                  "center_x": 200.0,
-                  "center_y": 150.0,
-                  "center_z": 2.0,
-                  "radius": 80.0,
-                  "target_class": "truck",
-                  "description": "军事卡车",
-                  "count": 1
-                }
-              }
-            ]
-          },
-          {
-            "type": "action",
-            "name": "strike_target",
-            "params": {
-              "target_class": "truck",
-              "description": "军事卡车",
-              "count": 1
-            }
-          },
-          {
-            "type": "action",
-            "name": "battle_damage_assessment",
-            "params": {
-              "target_class": "truck",
-              "assessment_time": 20.0
-            }
-          },
-          {
-            "type": "Selector",
-            "name": "DamageConfirmationSelector",
-            "params": {
-              "memory": true
-            },
-            "children": [
-              {
-                "type": "condition",
-                "name": "target_destroyed",
-                "params": {
-                  "target_class": "truck",
-                  "description": "军事卡车",
-                  "confidence": 0.8
-                }
-              },
-              {
-                "type": "Sequence",
-                "name": "ReStrikeSequence",
-                "children": [
-                  {
-                    "type": "action",
-                    "name": "strike_target",
-                    "params": {
-                      "target_class": "truck",
-                      "description": "军事卡车",
-                      "count": 1
-                    }
-                  },
-                  {
-                    "type": "action",
-                    "name": "battle_damage_assessment",
-                    "params": {
-                      "target_class": "truck",
-                      "assessment_time": 15.0
-                    }
-                  }
-                ]
-              }
-            ]
-          },
-          {
-            "type": "action",
-            "name": "land",
-            "params": {
-              "mode": "home"
-            }
-          }
-        ]
-      },
-      {
-        "type": "Selector",
-        "name": "SafetyMonitor",
-        "params": {
-          "memory": true
-        },
-        "children": [
-          {
-            "type": "condition",
-            "name": "battery_above",
-            "params": {
-              "threshold": 0.35
-            }
-          },
-          {
-            "type": "condition",
-            "name": "gps_status",
-            "params": {
-              "min_satellites": 8
-            }
-          },
-          {
-            "type": "Sequence",
-            "name": "EmergencyHandler",
-            "children": [
-              {
-                "type": "action",
-                "name": "emergency_return",
-                "params": {
-                  "reason": "safety_breach"
-                }
-              },
-              {
-                "type": "action",
-                "name": "land",
-                "params": {
-                  "mode": "home"
-                }
-              }
-            ]
-          }
-        ]
-      }
-    ]
-  }
-}
-
-#### 9. 跟踪任务范式
-所有任务必须包含安全监控。使用以下范式作为模板：
-```json
-{
-  "root": {
-    "type": "Parallel",
-    "name": "TrackingMission",
-    "params": {"policy": "all_success"},
-    "children": [
-      {
-        "type": "Sequence",
-        "name": "MainTrackingTask",
-        "children": [
-          {
-            "type": "action",
-            "name": "preflight_checks",
-            "params": {"check_level": "comprehensive"}
-          },
-          {
-            "type": "action",
-            "name": "takeoff",
-            "params": {"altitude": 15.0}
-          },
-          {
-            "type": "action",
-            "name": "fly_to_waypoint",
-            "params": {
-              "x": 100.0,
-              "y": 80.0,
-              "z": 15.0,
-              "acceptance_radius": 3.0
-            }
-          },
-          {
-            "type": "Selector",
-            "name": "TargetAcquisitionSelector",
-            "params": {"memory": true},
-            "children": [
-              {
-                "type": "Sequence",
-                "name": "DirectDetectionSequence",
-                "children": [
-                  {
-                    "type": "action",
-                    "name": "loiter",
-                    "params": {"duration": 10.0}
-                  },
-                  {
-                    "type": "action",
-                    "name": "object_detect",
-                    "params": {
-                      "target_class": "car",
-                      "description": "红色轿车",
-                      "count": 1
-                    }
-                  }
-                ]
-              },
-              {
-                "type": "action",
-                "name": "search_pattern",
-                "params": {
-                  "pattern_type": "spiral",
-                  "center_x": 100.0,
-                  "center_y": 80.0,
-                  "center_z": 15.0,
-                  "radius": 50.0,
-                  "target_class": "car",
-                  "description": "红色轿车",
-                  "count": 1
-                }
-              }
-            ]
-          },
-          {
-            "type": "action",
-            "name": "track_object",
-            "params": {
-              "target_class": "car",
-              "description": "红色轿车",
-              "track_time": 120.0,
-              "min_confidence": 0.7,
-              "safe_distance": 15.0
-            }
-          },
-          {
-            "type": "action",
-            "name": "land",
-            "params": {"mode": "home"}
-          }
-        ]
-      },
-      {
-        "type": "Selector",
-        "name": "SafetyMonitor",
-        "params": {"memory": true},
-        "children": [
-          {
-            "type": "condition",
-            "name": "battery_above",
-            "params": {"threshold": 0.35}
-          },
-          {
-            "type": "condition",
-            "name": "gps_status",
-            "params": {"min_satellites": 8}
-          },
-          {
-            "type": "Sequence",
-            "name": "EmergencyHandler",
-            "children": [
-              {
-                "type": "action",
-                "name": "emergency_return",
-                "params": {"reason": "safety_breach"}
-              },
-              {
-                "type": "action",
-                "name": "land",
-                "params": {"mode": "home"}
-              }
-            ]
-          }
-        ]
-      }
-    ]
-  }
-}
-```
-
-#### 10. 如何使用参考知识
-当用户提供"参考知识"（如坐标信息）时，你必须使用这些信息填充参数。例如：
-- 如果参考知识说"目标坐标: (x: 120.5, y: 80.2, z: 60.0)"，则在使用`fly_to_waypoint`时设置这些值。
-- 确保坐标符合约束（如z≥1）。
-
-环绕口令到参数的映射规则（当口令涉及“环绕/绕圈”等）：
-- “环绕XY圈” → `radius=X`, `laps=Y`，默认 `clockwise=true`, `gimbal_lock=true`，比如环绕三十两圈，意思就是以目标点为圆心，30米为半径绕2圈
-- 明确“顺时针/逆时针”时 → 设置 `clockwise=true/false`
-- 出现“等速”时 → 若未给速度则 `speed_mps` 使用默认值（如3.0）；若口令给出速度，裁剪到[0.5,15]
-- “以(中心坐标)为中心/当前位置为中心” → 使用 `orbit_around_point` 并填写 `center_x/center_y/center_z`
-- “以目标为中心/围绕目标” → 使用 `orbit_around_target`；若任务未提供目标来源，则需要在主任务中先行确认目标（通过检测/跟踪或参考知识）
-
-#### 11. 输出要求
-你的输出必须是严格的、单一的JSON对象，符合上述所有规则。不包含任何自然语言描述。
+## 六、输出要求
+仅输出1个严格符合上述所有规则的JSON对象，**确保：1. 优先级排序逻辑正确填入`description`；2. `target_class`匹配预定义列表；3. 行为树结构不变；4. 后端解析与Schema验证无错误**，无任何冗余内容。

backend_service/src/py_tree_generator.py

@@ -201,7 +201,7 @@ def _generate_pytree_schema(allowed_actions: set, allowed_conditions: set) -> di
         "sandwich", "orange", "broccoli", "carrot", "hot_dog", "pizza", "donut", "cake", "chair",
         "couch", "potted_plant", "bed", "dining_table", "toilet", "tv", "laptop", "mouse", "remote",
         "keyboard", "cell_phone", "microwave", "oven", "toaster", "sink", "refrigerator", "book",
-        "clock", "vase", "scissors", "teddy_bear", "hair_drier", "toothbrush"
+        "clock", "vase", "scissors", "teddy_bear", "hair_drier", "toothbrush","balloon"
     ]
     
     # 递归节点定义
@@ -548,6 +548,16 @@ class PyTreeGenerator:
         # Updated output directory for visualizations
         self.vis_dir = os.path.abspath(os.path.join(self.base_dir, '..', 'generated_visualizations'))
         os.makedirs(self.vis_dir, exist_ok=True)
+        # Reasoning content output directory (Markdown files)
+        self.reasoning_dir = os.path.abspath(os.path.join(self.base_dir, '..', 'generated_reasoning_content'))
+        os.makedirs(self.reasoning_dir, exist_ok=True)
+        # 控制是否允许模型返回含 <think> 的原文（不强制JSON），以便提取推理链
+        self.enable_reasoning_capture = os.getenv("ENABLE_REASONING_CAPTURE", "true").lower() in ("1", "true", "yes")
+        # 终端预览的最大行数
+        try:
+            self.reasoning_preview_lines = int(os.getenv("REASONING_PREVIEW_LINES", "20"))
+        except Exception:
+            self.reasoning_preview_lines = 20
         # 加载提示词：复杂模式复用现有 system_prompt.txt；简单模式与分类器独立提示词
         self.complex_prompt = self._load_prompt("system_prompt.txt")
         self.simple_prompt = self._load_prompt("simple_mode_prompt.txt")
@@ -564,6 +574,10 @@ class PyTreeGenerator:
         self.simple_base_url = os.getenv("SIMPLE_BASE_URL", f"http://{self.orin_ip}:8081/v1")
         self.complex_base_url = os.getenv("COMPLEX_BASE_URL", f"http://{self.orin_ip}:8081/v1")
         self.api_key = os.getenv("OPENAI_API_KEY", "sk-no-key-required")
+        # 直接在代码中指定最大输出token数（不通过环境变量）
+        self.classifier_max_tokens = 512
+        self.simple_max_tokens = 8192
+        self.complex_max_tokens = 8192
         
         # 为不同用途分别创建客户端
         self.classifier_client = openai.OpenAI(api_key=self.api_key, base_url=self.classifier_base_url)
@@ -605,6 +619,8 @@ class PyTreeGenerator:
                 return None
             context_str = "\n\n".join(retrieved_docs)
             logging.info("--- 成功检索到上下文信息 ---")
+            # 打印检索到的上下文内容
+            logging.info(f"📚 检索到的上下文内容:\n{context_str}")
             return context_str
         except Exception as e:
             logging.error(f"从向量数据库检索时发生错误: {e}")
@@ -626,7 +642,11 @@ class PyTreeGenerator:
                     {"role": "user", "content": user_prompt}
                 ],
                 temperature=0.0,
-                response_format={"type": "json_object"}
+                response_format={"type": "json_object"},  # 强制JSON输出，禁用思考功能
+                max_tokens=self.classifier_max_tokens,
+                # 禁用 Qwen3 模型的思考功能（通过 extra_body 传递）
+                # 注意：如果 API 服务器不支持此参数，会忽略
+                extra_body={"chat_template_kwargs": {"enable_thinking": False}}
             )
             class_str = classifier_resp.choices[0].message.content
             class_obj = json.loads(class_str)
@@ -655,27 +675,102 @@ class PyTreeGenerator:
             final_user_prompt += augmentation
         else:
             logging.warning("未检索到上下文或检索失败，将使用原始用户提示词。")
+        
+        # 构建完整的 final_prompt（准确反映实际发送给大模型的内容结构）
+        # 注意：RAG检索结果被添加到 user prompt 中，而不是 system prompt
+        # System Prompt: use_prompt（不包含RAG结果）
+        # User Prompt: final_user_prompt（包含原始user_prompt + RAG检索结果）
+        final_prompt = f"=== System Prompt ===\n{use_prompt}\n\n=== User Prompt ===\n{final_user_prompt}"
         for attempt in range(3):
             logging.info(f"--- 第 {attempt + 1}/3 次尝试生成Pytree ---")
             try:
                 # 简单/复杂分流到不同模型与提示词
                 client = self.simple_llm_client if mode == "simple" else self.complex_llm_client
                 model_name = self.simple_model if mode == "simple" else self.complex_model
-                response = client.chat.completions.create(
-                    model=model_name,
-                    messages=[
+                # 始终强制JSON响应并禁用思考功能
+                response_kwargs = {
+                    "model": model_name,
+                    "messages": [
                         {"role": "system", "content": use_prompt},
                         {"role": "user", "content": final_user_prompt}
                     ],
-                    temperature=0.1 if mode == "complex" else 0.0,
-                    response_format={"type": "json_object"}
-                )
-                pytree_str = response.choices[0].message.content
+                    "temperature": 0.1 if mode == "complex" else 0.0,
+                    "response_format": {"type": "json_object"},  # 始终强制JSON输出，禁用思考功能
+                    # 禁用 Qwen3 模型的思考功能（通过 extra_body 传递）
+                    # 注意：如果 API 服务器不支持此参数，会忽略
+                    "extra_body": {"chat_template_kwargs": {"enable_thinking": False}}
+                }
+                # 基于模式设定最大输出token数（直接在代码中配置）
+                response_kwargs["max_tokens"] = self.simple_max_tokens if mode == "simple" else self.complex_max_tokens
+                response = client.chat.completions.create(**response_kwargs)
+                # 兼容可能存在的 reasoning_content 字段
+                try:
+                    msg = response.choices[0].message
+                    msg_content = getattr(msg, "content", None)
+                    msg_reasoning = getattr(msg, "reasoning_content", None)
+                except Exception:
+                    msg = response.choices[0]["message"] if isinstance(response.choices[0], dict) else None
+                    msg_content = (msg or {}).get("content") if isinstance(msg, dict) else None
+                    msg_reasoning = (msg or {}).get("reasoning_content") if isinstance(msg, dict) else None
+
+                combined_text = ""
+                if isinstance(msg_reasoning, str) and msg_reasoning.strip():
+                    # 将 reasoning_content 包装为 <think>，便于统一解析
+                    combined_text += f"<think>\n{msg_reasoning}\n</think>\n"
+                if isinstance(msg_content, str) and msg_content.strip():
+                    combined_text += msg_content
+                pytree_str = combined_text if combined_text else (msg_content or "")
+                raw_full_text_for_logging = pytree_str  # 保存完整原文（含 <think>）以便失败时完整打印
+
+                # 提取 <think> 推理链内容（若存在）
+                reasoning_text = None
+                try:
+                    think_match = re.search(r"<think>([\s\S]*?)</think>", pytree_str)
+                    if think_match:
+                        reasoning_text = think_match.group(1).strip()
+                        # 去除推理文本后再尝试解析JSON
+                        pytree_str = re.sub(r"<think>[\s\S]*?</think>", "", pytree_str).strip()
+                except Exception:
+                    reasoning_text = None
                 # 单独捕获JSON解析错误并打印原始响应
                 try:
                     pytree_dict = json.loads(pytree_str)
                 except json.JSONDecodeError as e:
-                    logging.error(f"❌ JSON解析失败（第 {attempt + 1}/3 次）。原始响应如下：\n{pytree_str}")
+                    logging.error(f"❌ JSON解析失败（第 {attempt + 1}/3 次）。\n—— 完整原始文本(含<think>) ——\n{raw_full_text_for_logging}")
+                    # 尝试打印响应对象的完整结构
+                    try:
+                        raw_response_dump = None
+                        if hasattr(response, 'model_dump_json'):
+                            raw_response_dump = response.model_dump_json(indent=2, exclude_none=False)
+                        elif hasattr(response, 'dict'):
+                            raw_response_dump = json.dumps(response.dict(), ensure_ascii=False, indent=2, default=str)
+                        else:
+                            # 兜底：尝试将choices与关键字段展开
+                            safe_obj = {
+                                "id": getattr(response, 'id', None),
+                                "model": getattr(response, 'model', None),
+                                "object": getattr(response, 'object', None),
+                                "usage": getattr(response, 'usage', None),
+                                "choices": [
+                                    {
+                                        "index": getattr(c, 'index', None),
+                                        "finish_reason": getattr(c, 'finish_reason', None),
+                                        "message": {
+                                            "role": getattr(getattr(c, 'message', None), 'role', None),
+                                            "content": getattr(getattr(c, 'message', None), 'content', None),
+                                            "reasoning_content": getattr(getattr(c, 'message', None), 'reasoning_content', None)
+                                        } if getattr(c, 'message', None) is not None else None
+                                    }
+                                    for c in getattr(response, 'choices', [])
+                                ] if hasattr(response, 'choices') else None
+                            }
+                            raw_response_dump = json.dumps(safe_obj, ensure_ascii=False, indent=2, default=str)
+                        logging.error(f"—— 完整响应对象 ——\n{raw_response_dump}")
+                    except Exception as dump_e:
+                        try:
+                            logging.error(f"响应对象转储失败，repr如下：\n{repr(response)}")
+                        except Exception:
+                            pass
                     continue
 
                 # 简单/复杂分别验证与返回
@@ -702,6 +797,27 @@ class PyTreeGenerator:
                         pytree_dict['visualization_url'] = f"/static/{vis_filename}"
                     except Exception as e:
                         logging.warning(f"简单模式可视化失败: {e}")
+
+                    # 保存推理链（若有）
+                    try:
+                        if reasoning_text:
+                            reasoning_path = os.path.join(self.reasoning_dir, "reasoning_content.md")
+                            with open(reasoning_path, 'w', encoding='utf-8') as rf:
+                                rf.write(reasoning_text)
+                            logging.info(f"📝 推理链已保存: {reasoning_path}")
+                            # 终端预览（最多N行）
+                            try:
+                                lines = reasoning_text.splitlines()
+                                preview = "\n".join(lines[: self.reasoning_preview_lines])
+                                logging.info("🧠 推理链预览(前%d行)：\n%s", self.reasoning_preview_lines, preview)
+                            except Exception:
+                                pass
+                        else:
+                            logging.info("未在模型输出中发现 <think> 推理链片段。若需捕获，请设置 ENABLE_REASONING_CAPTURE=true 以放宽JSON强制格式。")
+                    except Exception as e:
+                        logging.warning(f"保存推理链Markdown失败: {e}")
+                    # 添加 final_prompt 到返回结果
+                    pytree_dict['final_prompt'] = final_prompt
                     return pytree_dict
 
                 # 复杂模式回退：若模型误返回简单结构，则自动包装为含安全监控的行为树
@@ -754,6 +870,27 @@ class PyTreeGenerator:
                     vis_path = os.path.join(self.vis_dir, vis_filename)
                     _visualize_pytree(pytree_dict['root'], os.path.splitext(vis_path)[0])
                     pytree_dict['visualization_url'] = f"/static/{vis_filename}"
+
+                    # 保存推理链（若有）
+                    try:
+                        if reasoning_text:
+                            reasoning_path = os.path.join(self.reasoning_dir, "reasoning_content.md")
+                            with open(reasoning_path, 'w', encoding='utf-8') as rf:
+                                rf.write(reasoning_text)
+                            logging.info(f"📝 推理链已保存: {reasoning_path}")
+                            # 终端预览（最多N行）
+                            try:
+                                lines = reasoning_text.splitlines()
+                                preview = "\n".join(lines[: self.reasoning_preview_lines])
+                                logging.info("🧠 推理链预览(前%d行)：\n%s", self.reasoning_preview_lines, preview)
+                            except Exception:
+                                pass
+                        else:
+                            logging.info("未在模型输出中发现 <think> 推理链片段。若需捕获，请设置 ENABLE_REASONING_CAPTURE=true 以放宽JSON强制格式。")
+                    except Exception as e:
+                        logging.warning(f"保存推理链Markdown失败: {e}")
+                    # 添加 final_prompt 到返回结果
+                    pytree_dict['final_prompt'] = final_prompt
                     return pytree_dict
                 else:
                     # 打印未通过验证的Pytree以便排查

backend_service/src/ros2_client.py

@@ -1,67 +0,0 @@
-import rclpy
-from rclpy.action import ActionClient
-from rclpy.node import Node
-import json
-from typing import Dict, Any
-import logging
-
-from drone_interfaces.action import ExecuteMission
-from .websocket_manager import websocket_manager
-
-class MissionActionClient(Node):
-    """
-    Interfaces with the drone's `ExecuteMission` ROS2 Action Server.
-    """
-    def __init__(self):
-        super().__init__('mission_action_client')
-        self._action_client = ActionClient(self, ExecuteMission, 'execute_mission')
-        self.get_logger().info("MissionActionClient initialized.")
-
-    def send_goal(self, py_tree: Dict[str, Any]):
-        """
-        Sends the mission (py_tree) to the action server.
-        """
-        if not self._action_client.server_is_ready():
-            self.get_logger().error("Action server not available, goal not sent.")
-            # Optionally, you could broadcast a status update to the frontend here
-            return
-
-        self.get_logger().info("Received request to send goal to drone.")
-        goal_msg = ExecuteMission.Goal()
-        goal_msg.py_tree_json = json.dumps(py_tree)
-
-        self.get_logger().info(f"Sending goal to action server...")
-        send_goal_future = self._action_client.send_goal_async(
-            goal_msg,
-            feedback_callback=self.feedback_callback
-        )
-
-        send_goal_future.add_done_callback(self.goal_response_callback)
-
-    def goal_response_callback(self, future):
-        goal_handle = future.result()
-        if not goal_handle.accepted:
-            self.get_logger().info('Goal rejected :(')
-            return
-
-        self.get_logger().info('Goal accepted :)')
-
-        self._get_result_future = goal_handle.get_result_async()
-        self._get_result_future.add_done_callback(self.get_result_callback)
-
-    def get_result_callback(self, future):
-        result = future.result().result
-        self.get_logger().info(f'Result: {{success: {result.success}, message: {result.message}}}')
-        # Optionally, you can broadcast the final result via WebSocket here
-
-    def feedback_callback(self, feedback_msg):
-        """
-        This callback is triggered by the action server.
-        It forwards the status to the QGC plugin via the WebSocket manager in a thread-safe manner.
-        """
-        feedback = feedback_msg.feedback
-        feedback_payload = json.dumps({"node_id": feedback.node_id, "status": feedback.status})
-        self.get_logger().info(f"Received feedback: {feedback_payload}")
-        websocket_manager.broadcast(feedback_payload)
-
-# Note: The rclpy.init() and spinning of the node will be handled in main.py

backend_service/src/websocket_manager.py

@@ -22,7 +22,8 @@ class ConnectionManager:
     def broadcast(self, message: str):
         """
         Thread-safely broadcasts a message to all active WebSocket connections.
-        This method is designed to be called from a different thread (e.g., a ROS2 callback).
+        This method is designed to be called from a different thread.
+        (Note: ROS2 callback support has been removed as the project is decoupled from ROS2)
         """
         if not self.loop:
             logging.error("Event loop not set in ConnectionManager. Cannot broadcast.")

start_all.sh

@@ -0,0 +1,406 @@
+#!/bin/bash
+
+# ==============================================================================
+# 无人机自然语言控制项目 - 一键启动脚本
+# ==============================================================================
+# 功能：启动所有必需的服务（llama-server推理模型、embedding模型、FastAPI后端）
+# 用法：./start_all.sh [选项]
+# ==============================================================================
+
+set -e  # 遇到错误立即退出
+
+# 颜色定义
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+NC='\033[0m' # No Color
+
+# 默认配置（可通过环境变量覆盖）
+PROJECT_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+LLAMA_SERVER_DIR="${LLAMA_SERVER_DIR:-~/llama.cpp/build/bin}"
+INFERENCE_MODEL="${INFERENCE_MODEL:-~/models/gguf/Qwen/Qwen3-4B/Qwen3-4B-Q5_K_M.gguf}"
+EMBEDDING_MODEL="${EMBEDDING_MODEL:-~/models/gguf/Qwen/Qwen3-Embedding-4B/Qwen3-Embedding-4B-Q4_K_M.gguf}"
+VENV_PATH="${VENV_PATH:-${PROJECT_ROOT}/backend_service/venv}"
+LOG_DIR="${PROJECT_ROOT}/logs"
+PID_FILE="${LOG_DIR}/services.pid"
+
+# 端口配置
+INFERENCE_PORT=8081
+EMBEDDING_PORT=8090
+API_PORT=8000
+
+# 创建日志目录
+mkdir -p "${LOG_DIR}"
+
+# ==============================================================================
+# 辅助函数
+# ==============================================================================
+
+print_info() {
+    echo -e "${BLUE}[INFO]${NC} $1"
+}
+
+print_success() {
+    echo -e "${GREEN}[SUCCESS]${NC} $1"
+}
+
+print_warning() {
+    echo -e "${YELLOW}[WARNING]${NC} $1"
+}
+
+print_error() {
+    echo -e "${RED}[ERROR]${NC} $1"
+}
+
+# 检查命令是否存在
+check_command() {
+    if ! command -v "$1" &> /dev/null; then
+        print_error "$1 命令未找到，请先安装"
+        return 1
+    fi
+    return 0
+}
+
+# 检查端口是否被占用
+check_port() {
+    local port=$1
+    if lsof -Pi :${port} -sTCP:LISTEN -t >/dev/null 2>&1 ; then
+        return 0  # 端口被占用
+    else
+        return 1  # 端口空闲
+    fi
+}
+
+# 等待服务就绪
+wait_for_service() {
+    local url=$1
+    local service_name=$2
+    local max_attempts=30
+    local attempt=0
+    
+    print_info "等待 ${service_name} 启动..."
+    while [ $attempt -lt $max_attempts ]; do
+        if curl -s "${url}" > /dev/null 2>&1; then
+            print_success "${service_name} 已就绪"
+            return 0
+        fi
+        attempt=$((attempt + 1))
+        sleep 1
+    done
+    
+    print_error "${service_name} 启动超时"
+    return 1
+}
+
+# 停止所有服务
+stop_services() {
+    print_info "正在停止所有服务..."
+    
+    if [ -f "${PID_FILE}" ]; then
+        while read pid; do
+            if ps -p $pid > /dev/null 2>&1; then
+                print_info "停止进程 PID: $pid"
+                kill $pid 2>/dev/null || true
+            fi
+        done < "${PID_FILE}"
+        rm -f "${PID_FILE}"
+    fi
+    
+    # 尝试通过端口停止服务
+    for port in ${INFERENCE_PORT} ${EMBEDDING_PORT} ${API_PORT}; do
+        if check_port ${port}; then
+            local pid=$(lsof -ti:${port})
+            if [ ! -z "$pid" ]; then
+                print_info "停止占用端口 ${port} 的进程 (PID: $pid)"
+                kill $pid 2>/dev/null || true
+            fi
+        fi
+    done
+    
+    print_success "所有服务已停止"
+}
+
+# 清理函数（脚本退出时调用）
+cleanup() {
+    if [ "$?" -ne 0 ]; then
+        print_error "启动过程中发生错误，正在清理..."
+    fi
+    # 注意：这里不自动停止服务，让用户手动控制
+}
+
+trap cleanup EXIT
+
+# ==============================================================================
+# 主函数
+# ==============================================================================
+
+start_services() {
+    print_info "=========================================="
+    print_info "  无人机自然语言控制项目 - 服务启动"
+    print_info "=========================================="
+    echo ""
+    
+    # 检查必要的命令
+    print_info "检查必要的命令..."
+    check_command "python3" || exit 1
+    check_command "curl" || exit 1
+    check_command "lsof" || print_warning "lsof 未安装，将无法检查端口占用"
+    echo ""
+    
+    # 检查端口占用
+    print_info "检查端口占用..."
+    if check_port ${INFERENCE_PORT}; then
+        print_warning "端口 ${INFERENCE_PORT} 已被占用，推理模型可能已在运行"
+    fi
+    if check_port ${EMBEDDING_PORT}; then
+        print_warning "端口 ${EMBEDDING_PORT} 已被占用，Embedding模型可能已在运行"
+    fi
+    if check_port ${API_PORT}; then
+        print_error "端口 ${API_PORT} 已被占用，请先停止占用该端口的服务"
+        exit 1
+    fi
+    echo ""
+    
+    # 检查llama-server（展开路径中的 ~）
+    local llama_server_dir_expanded=$(eval echo "${LLAMA_SERVER_DIR}")
+    local llama_server="${llama_server_dir_expanded}/llama-server"
+    if [ ! -f "${llama_server}" ]; then
+        print_error "llama-server 未找到: ${llama_server}"
+        print_info "请设置 LLAMA_SERVER_DIR 环境变量指向正确的路径"
+        print_info "当前路径: ${LLAMA_SERVER_DIR}"
+        print_info "展开后路径: ${llama_server_dir_expanded}"
+        exit 1
+    fi
+    print_success "找到 llama-server: ${llama_server}"
+    echo ""
+    
+    # 检查模型文件
+    local inference_model_expanded=$(eval echo "${INFERENCE_MODEL}")
+    local embedding_model_expanded=$(eval echo "${EMBEDDING_MODEL}")
+    
+    if [ ! -f "${inference_model_expanded}" ]; then
+        print_error "推理模型文件未找到: ${inference_model_expanded}"
+        print_info "请设置 INFERENCE_MODEL 环境变量指向正确的模型路径"
+        exit 1
+    fi
+    print_success "找到推理模型: ${inference_model_expanded}"
+    
+    if [ ! -f "${embedding_model_expanded}" ]; then
+        print_error "Embedding模型文件未找到: ${embedding_model_expanded}"
+        print_info "请设置 EMBEDDING_MODEL 环境变量指向正确的模型路径"
+        exit 1
+    fi
+    print_success "找到Embedding模型: ${embedding_model_expanded}"
+    echo ""
+    
+    # 检查ROS2环境
+    local ros2_setup="${PROJECT_ROOT}/install/setup.bash"
+    if [ ! -f "${ros2_setup}" ]; then
+        print_warning "ROS2 setup文件未找到: ${ros2_setup}"
+        print_warning "如果项目已与ROS2解耦，可以忽略此警告"
+    else
+        print_success "找到ROS2 setup文件: ${ros2_setup}"
+    fi
+    echo ""
+    
+    # 检查venv虚拟环境
+    local venv_path_expanded=$(eval echo "${VENV_PATH}")
+    print_info "检查venv虚拟环境: ${venv_path_expanded}"
+    if [ ! -d "${venv_path_expanded}" ]; then
+        print_error "venv虚拟环境目录不存在: ${venv_path_expanded}"
+        print_info "请先创建venv环境: python3 -m venv ${venv_path_expanded}"
+        print_info "然后安装依赖: ${venv_path_expanded}/bin/pip install -r backend_service/requirements.txt"
+        exit 1
+    fi
+    if [ ! -f "${venv_path_expanded}/bin/activate" ]; then
+        print_error "venv激活脚本不存在: ${venv_path_expanded}/bin/activate"
+        print_error "这看起来不是一个有效的venv环境"
+        exit 1
+    fi
+    print_success "venv虚拟环境存在: ${venv_path_expanded}"
+    echo ""
+    
+    # 初始化PID文件
+    > "${PID_FILE}"
+    
+    # ==========================================================================
+    # 启动推理模型服务
+    # ==========================================================================
+    print_info "启动推理模型服务 (端口 ${INFERENCE_PORT})..."
+    cd "${llama_server_dir_expanded}"
+    nohup ./llama-server \
+        -m "${inference_model_expanded}" \
+        --port ${INFERENCE_PORT} \
+        --gpu-layers 36 \
+        --host 0.0.0.0 \
+        -c 8192 \
+        > "${LOG_DIR}/inference_model.log" 2>&1 &
+    local inference_pid=$!
+    echo $inference_pid >> "${PID_FILE}"
+    print_success "推理模型服务已启动 (PID: $inference_pid)"
+    print_info "日志文件: ${LOG_DIR}/inference_model.log"
+    echo ""
+    
+    # ==========================================================================
+    # 启动Embedding模型服务
+    # ==========================================================================
+    print_info "启动Embedding模型服务 (端口 ${EMBEDDING_PORT})..."
+    nohup ./llama-server \
+        -m "${embedding_model_expanded}" \
+        --gpu-layers 36 \
+        --port ${EMBEDDING_PORT} \
+        --embeddings \
+        --pooling last \
+        --host 0.0.0.0 \
+        > "${LOG_DIR}/embedding_model.log" 2>&1 &
+    local embedding_pid=$!
+    echo $embedding_pid >> "${PID_FILE}"
+    print_success "Embedding模型服务已启动 (PID: $embedding_pid)"
+    print_info "日志文件: ${LOG_DIR}/embedding_model.log"
+    echo ""
+    
+    # ==========================================================================
+    # 等待模型服务就绪
+    # ==========================================================================
+    print_info "等待模型服务就绪..."
+    sleep 3  # 给服务一些启动时间
+    
+    # 等待推理模型服务
+    if ! wait_for_service "http://localhost:${INFERENCE_PORT}/health" "推理模型服务"; then
+        # 如果health端点不存在，尝试检查根路径
+        if ! wait_for_service "http://localhost:${INFERENCE_PORT}/v1/models" "推理模型服务"; then
+            print_warning "推理模型服务可能未完全就绪，但将继续启动"
+        fi
+    fi
+    
+    # 等待Embedding模型服务
+    if ! wait_for_service "http://localhost:${EMBEDDING_PORT}/health" "Embedding模型服务"; then
+        if ! wait_for_service "http://localhost:${EMBEDDING_PORT}/v1/models" "Embedding模型服务"; then
+            print_warning "Embedding模型服务可能未完全就绪，但将继续启动"
+        fi
+    fi
+    echo ""
+    
+    # ==========================================================================
+    # 启动FastAPI后端服务
+    # ==========================================================================
+    print_info "启动FastAPI后端服务 (端口 ${API_PORT})..."
+    cd "${PROJECT_ROOT}"
+    
+    # 激活venv虚拟环境并启动FastAPI服务
+    # 使用bash -c来在新的shell中激活venv环境
+    bash -c "
+        # 激活ROS2环境（如果存在）
+        if [ -f '${ros2_setup}' ]; then
+            source '${ros2_setup}'
+        fi
+        # 激活venv虚拟环境
+        source '${venv_path_expanded}/bin/activate' && \
+        cd '${PROJECT_ROOT}/backend_service' && \
+        uvicorn src.main:app --host 0.0.0.0 --port ${API_PORT}
+    " > "${LOG_DIR}/fastapi.log" 2>&1 &
+    local api_pid=$!
+    echo $api_pid >> "${PID_FILE}"
+    print_success "FastAPI服务已启动 (PID: $api_pid)"
+    print_info "日志文件: ${LOG_DIR}/fastapi.log"
+    echo ""
+    
+    # 等待FastAPI服务就绪
+    sleep 3
+    if wait_for_service "http://localhost:${API_PORT}/docs" "FastAPI服务"; then
+        print_success "所有服务已成功启动！"
+    else
+        print_warning "FastAPI服务可能未完全就绪，请检查日志: ${LOG_DIR}/fastapi.log"
+    fi
+    echo ""
+    
+    # 显示服务访问信息
+    print_info "=========================================="
+    print_info "  服务启动完成！"
+    print_info "=========================================="
+    print_info "推理模型API: http://localhost:${INFERENCE_PORT}/v1"
+    print_info "Embedding模型API: http://localhost:${EMBEDDING_PORT}/v1"
+    print_info "FastAPI后端: http://localhost:${API_PORT}"
+    print_info "API文档: http://localhost:${API_PORT}/docs"
+    print_info ""
+    print_info "日志文件位置:"
+    print_info "  - 推理模型: ${LOG_DIR}/inference_model.log"
+    print_info "  - Embedding模型: ${LOG_DIR}/embedding_model.log"
+    print_info "  - FastAPI服务: ${LOG_DIR}/fastapi.log"
+    print_info ""
+    print_info "按 Ctrl+C 停止所有服务"
+    print_info "=========================================="
+    echo ""
+    
+    # 设置信号处理，确保Ctrl+C时能清理
+    trap 'print_info "\n正在停止服务..."; stop_services; exit 0' INT TERM
+    
+    # 等待所有后台进程（保持脚本运行）
+    print_info "所有服务正在运行中，查看日志请使用:"
+    print_info "  tail -f ${LOG_DIR}/*.log"
+    echo ""
+    
+    # 等待所有后台进程
+    wait
+}
+
+# ==============================================================================
+# 脚本入口
+# ==============================================================================
+
+case "${1:-start}" in
+    start)
+        start_services
+        ;;
+    stop)
+        stop_services
+        ;;
+    restart)
+        stop_services
+        sleep 2
+        start_services
+        ;;
+    status)
+        print_info "检查服务状态..."
+        if [ -f "${PID_FILE}" ]; then
+            print_info "已记录的服务进程:"
+            while read pid; do
+                if ps -p $pid > /dev/null 2>&1; then
+                    print_success "PID $pid: 运行中"
+                else
+                    print_warning "PID $pid: 已停止"
+                fi
+            done < "${PID_FILE}"
+        else
+            print_info "未找到PID文件，服务可能未启动"
+        fi
+        echo ""
+        print_info "端口占用情况:"
+        for port in ${INFERENCE_PORT} ${EMBEDDING_PORT} ${API_PORT}; do
+            if check_port ${port}; then
+                local pid=$(lsof -ti:${port})
+                print_success "端口 ${port}: 被占用 (PID: $pid)"
+            else
+                print_warning "端口 ${port}: 空闲"
+            fi
+        done
+        ;;
+    *)
+        echo "用法: $0 {start|stop|restart|status}"
+        echo ""
+        echo "命令说明:"
+        echo "  start   - 启动所有服务（默认）"
+        echo "  stop    - 停止所有服务"
+        echo "  restart - 重启所有服务"
+        echo "  status  - 查看服务状态"
+        echo ""
+        echo "环境变量配置:"
+        echo "  LLAMA_SERVER_DIR  - llama-server所在目录 (默认: ~/llama.cpp/build/bin)"
+        echo "  INFERENCE_MODEL   - 推理模型路径 (默认: ~/models/gguf/Qwen/Qwen3-4B/Qwen3-4B-Q5_K_M.gguf)"
+        echo "  EMBEDDING_MODEL   - Embedding模型路径 (默认: ~/models/gguf/Qwen/Qwen3-Embedding-4B/Qwen3-Embedding-4B-Q4_K_M.gguf)"
+        echo "  VENV_PATH         - venv虚拟环境路径 (默认: \${PROJECT_ROOT}/backend_service/venv)"
+        exit 1
+        ;;
+esac
+

tools/response.json

@@ -1,149 +0,0 @@
-{
-    "plan_id": "9f743b03-8ba7-4a06-9260-337463887fc2",
-    "root": {
-        "children": [
-            {
-                "children": [
-                    {
-                        "name": "preflight_checks",
-                        "params": {
-                            "check_level": "comprehensive"
-                        },
-                        "type": "action"
-                    },
-                    {
-                        "name": "takeoff",
-                        "params": {
-                            "altitude": 2
-                        },
-                        "type": "action"
-                    },
-                    {
-                        "name": "fly_to_waypoint",
-                        "params": {
-                            "acceptance_radius": 2,
-                            "x": 5,
-                            "y": 3,
-                            "z": 2
-                        },
-                        "type": "action"
-                    },
-                    {
-                        "children": [
-                            {
-                                "children": [
-                                    {
-                                        "name": "loiter",
-                                        "params": {
-                                            "duration": 10
-                                        },
-                                        "type": "action"
-                                    },
-                                    {
-                                        "name": "object_detect",
-                                        "params": {
-                                            "count": 1,
-                                            "description": "学生",
-                                            "target_class": "person"
-                                        },
-                                        "type": "action"
-                                    }
-                                ],
-                                "name": "DirectDetectionSequence",
-                                "type": "Sequence"
-                            },
-                            {
-                                "name": "search_pattern",
-                                "params": {
-                                    "center_x": 5,
-                                    "center_y": 3,
-                                    "center_z": 2,
-                                    "count": 1,
-                                    "description": "学生",
-                                    "pattern_type": "spiral",
-                                    "radius": 50,
-                                    "target_class": "person"
-                                },
-                                "type": "action"
-                            }
-                        ],
-                        "name": "TargetAcquisitionSelector",
-                        "params": {
-                            "memory": true
-                        },
-                        "type": "Selector"
-                    },
-                    {
-                        "name": "track_object",
-                        "params": {
-                            "description": "学生",
-                            "min_confidence": 0.7,
-                            "safe_distance": 10,
-                            "target_class": "person",
-                            "track_time": 20
-                        },
-                        "type": "action"
-                    },
-                    {
-                        "name": "land",
-                        "params": {
-                            "mode": "home"
-                        },
-                        "type": "action"
-                    }
-                ],
-                "name": "MainTrackingTask",
-                "type": "Sequence"
-            },
-            {
-                "children": [
-                    {
-                        "name": "battery_above",
-                        "params": {
-                            "threshold": 0.35
-                        },
-                        "type": "condition"
-                    },
-                    {
-                        "name": "gps_status",
-                        "params": {
-                            "min_satellites": 8
-                        },
-                        "type": "condition"
-                    },
-                    {
-                        "children": [
-                            {
-                                "name": "emergency_return",
-                                "params": {
-                                    "reason": "safety_breach"
-                                },
-                                "type": "action"
-                            },
-                            {
-                                "name": "land",
-                                "params": {
-                                    "mode": "home"
-                                },
-                                "type": "action"
-                            }
-                        ],
-                        "name": "EmergencyHandler",
-                        "type": "Sequence"
-                    }
-                ],
-                "name": "SafetyMonitor",
-                "params": {
-                    "memory": true
-                },
-                "type": "Selector"
-            }
-        ],
-        "name": "TrackingMission",
-        "params": {
-            "policy": "all_success"
-        },
-        "type": "Parallel"
-    },
-    "visualization_url": "/static/py_tree.png"
-}

tools/test_api.py

@@ -3,6 +3,8 @@
 
 import requests
 import json
+import os
+from datetime import datetime
 
 # --- Configuration ---
 # The base URL of your running FastAPI service
@@ -12,19 +14,51 @@ BASE_URL = "http://127.0.0.1:8000"
 ENDPOINT = "/generate_plan"
 
 # The user prompt we will send for the test
-TEST_PROMPT = "飞到学生宿舍"
+TEST_PROMPT = "已知目标检测红色气球危险性高于蓝色气球高于绿色气球，飞往搜索区搜索并锁定危险性最高的气球，对其跟踪30秒后进行打击操作"
+
+# Log file path (will be created in the same directory as this script)
+LOG_FILE = os.path.join(os.path.dirname(__file__), "api_test.log")
+
+def write_log(message, print_to_console=True):
+    """
+    Write a message to the log file in append mode.
+    Supports multi-line messages - only the first line gets timestamp.
+    
+    Args:
+        message: The message to write (can be multi-line)
+        print_to_console: Whether to also print to console (default: True)
+    """
+    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    
+    # Split message into lines and add timestamp to first line only
+    lines = message.split('\n')
+    log_entries = [f"[{timestamp}] {lines[0]}\n"]
+    for line in lines[1:]:
+        log_entries.append(f"{' ' * (len(timestamp) + 3)}{line}\n")  # Indent continuation lines
+    
+    try:
+        with open(LOG_FILE, "a", encoding="utf-8") as f:
+            f.writelines(log_entries)
+    except Exception as e:
+        print(f"⚠️  Warning: Failed to write to log file: {e}")
+    
+    if print_to_console:
+        print(message)
 
 def test_generate_plan():
     """
     Sends a request to the /generate_plan endpoint and validates the response.
+    All results are logged to the log file for continuous tracking.
     """
     url = BASE_URL + ENDPOINT
     payload = {"user_prompt": TEST_PROMPT}
     headers = {"Content-Type": "application/json"}
 
-    print("--- API Test: Generate Plan ---")
-    print(f"✅ URL: {url}")
-    print(f"✅ Sending Prompt: \"{TEST_PROMPT}\"")
+    # Write separator and test start info to log
+    write_log("=" * 80, print_to_console=False)
+    write_log("--- API Test: Generate Plan ---")
+    write_log(f"URL: {url}")
+    write_log(f"Sending Prompt: \"{TEST_PROMPT}\"")
 
     try:
         # Send the POST request
@@ -36,40 +70,85 @@ def test_generate_plan():
         # Parse the JSON response
         data = response.json()
 
-        print("✅ Received Response:")
-        print(json.dumps(data, indent=2, ensure_ascii=False))
+        # Extract and log organized prompt if available in response
+        organized_prompt = None
+        if isinstance(data, dict):
+            # Check for various possible field names for organized prompt
+            organized_prompt = data.get("organized_prompt") or \
+                             data.get("processed_prompt") or \
+                             data.get("final_prompt") or \
+                             data.get("enhanced_prompt") or \
+                             data.get("user_prompt_enhanced")
+        
+        write_log("✅ Received Response:")
+        
+        # Log organized prompt if found
+        if organized_prompt:
+            write_log("\n📝 组织后的Prompt:")
+            write_log(organized_prompt)
+        else:
+            # If not in response, log the original prompt for reference
+            write_log("\n📝 原始Prompt:")
+            write_log(f"   {TEST_PROMPT}")
+            write_log("   (注: 组织后的prompt未在API响应中返回，如需查看请检查后端日志)")
+        
+        response_json = json.dumps(data, indent=2, ensure_ascii=False)
+        write_log("\n完整响应内容:")
+        write_log(response_json)
 
         # --- Validation ---
-        print("\n--- Validation Checks ---")
+        write_log("\n--- Validation Checks ---")
+        
+        validation_results = []
         
         # 1. Check if the response is a dictionary
         if isinstance(data, dict):
-            print("PASS: Response is a valid JSON object.")
+            validation_results.append("PASS: Response is a valid JSON object.")
         else:
-            print("FAIL: Response is not a valid JSON object.")
+            validation_results.append("FAIL: Response is not a valid JSON object.")
+            # Write all validation results to log before returning
+            for result in validation_results:
+                write_log(result)
+            write_log("=" * 80, print_to_console=False)
+            write_log("", print_to_console=False)  # Empty line for readability
             return
 
         # 2. Check for the existence of the 'root' key
         if "root" in data and isinstance(data['root'], dict):
-            print("PASS: Response contains a valid 'root' key.")
+            validation_results.append("PASS: Response contains a valid 'root' key.")
         else:
-            print("FAIL: Response does not contain a valid 'root' key.")
+            validation_results.append("FAIL: Response does not contain a valid 'root' key.")
 
         # 3. Check for the existence and format of the 'visualization_url' key
         if "visualization_url" in data and data["visualization_url"].endswith(".png"):
-            print(f"PASS: Response contains a valid 'visualization_url': {data['visualization_url']}")
+            validation_results.append(f"PASS: Response contains a valid 'visualization_url': {data['visualization_url']}")
         else:
-            print("FAIL: Response does not contain a valid 'visualization_url'.")
+            validation_results.append("FAIL: Response does not contain a valid 'visualization_url'.")
+
+        # Write all validation results to log
+        for result in validation_results:
+            write_log(result)
+
+        # Write test completion marker
+        write_log("✅ Test completed successfully")
+        write_log("=" * 80, print_to_console=False)
+        write_log("", print_to_console=False)  # Empty line for readability
 
     except requests.exceptions.RequestException as e:
-        print(f"\n❌ TEST FAILED: Could not connect to the server.")
-        print("   Please make sure the backend service is running.")
-        print(f"   Error details: {e}")
+        error_msg = f"❌ TEST FAILED: Could not connect to the server.\n   Please make sure the backend service is running.\n   Error details: {e}"
+        write_log(error_msg)
+        write_log("=" * 80, print_to_console=False)
+        write_log("", print_to_console=False)  # Empty line for readability
     except json.JSONDecodeError:
-        print(f"\n❌ TEST FAILED: The server response was not valid JSON.")
-        print(f"   Response text: {response.text}")
+        error_msg = f"❌ TEST FAILED: The server response was not valid JSON.\n   Response text: {response.text}"
+        write_log(error_msg)
+        write_log("=" * 80, print_to_console=False)
+        write_log("", print_to_console=False)  # Empty line for readability
     except Exception as e:
-        print(f"\n❌ TEST FAILED: An unexpected error occurred: {e}")
+        error_msg = f"❌ TEST FAILED: An unexpected error occurred: {e}"
+        write_log(error_msg)
+        write_log("=" * 80, print_to_console=False)
+        write_log("", print_to_console=False)  # Empty line for readability
 
 if __name__ == "__main__":
     test_generate_plan()

tools/test_llama_server.py

@@ -0,0 +1,174 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+
+import os
+import sys
+import json
+import argparse
+from typing import Any, Dict
+
+import requests
+
+
+def build_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(
+        description="调用本地 llama-server (OpenAI兼容) 进行推理，支持自定义系统/用户提示词"
+    )
+    parser.add_argument(
+        "--base-url",
+        default=os.getenv("SIMPLE_BASE_URL", "http://127.0.0.1:8081/v1"),
+        help="llama-server 的基础URL（默认: http://127.0.0.1:8081/v1，或环境变量 SIMPLE_BASE_URL）",
+    )
+    parser.add_argument(
+        "--model",
+        default=os.getenv("SIMPLE_MODEL", "local-model"),
+        help="模型名称（默认: local-model，或环境变量 SIMPLE_MODEL）",
+    )
+    parser.add_argument(
+        "--system",
+        default="You are a helpful assistant.",
+        help="系统提示词（system role）",
+    )
+    parser.add_argument(
+        "--system-file",
+        default=None,
+        help="系统提示词文件路径（txt）；若提供，则覆盖 --system 的字符串",
+    )
+    parser.add_argument(
+        "--user",
+        default=None,
+        help="用户提示词（user role）；若不传则从交互式输入读取",
+    )
+    parser.add_argument(
+        "--temperature",
+        type=float,
+        default=0.2,
+        help="采样温度（默认: 0.2）",
+    )
+    parser.add_argument(
+        "--max-tokens",
+        type=int,
+        default=4096,
+        help="最大生成Token数（默认: 4096）",
+    )
+    parser.add_argument(
+        "--timeout",
+        type=float,
+        default=120.0,
+        help="HTTP超时时间秒（默认: 120）",
+    )
+    parser.add_argument(
+        "--verbose",
+        action="store_true",
+        help="打印完整返回JSON",
+    )
+    return parser.parse_args()
+
+
+def call_llama_server(
+    base_url: str,
+    model: str,
+    system_prompt: str,
+    user_prompt: str,
+    temperature: float,
+    max_tokens: int,
+    timeout: float,
+) -> Dict[str, Any]:
+    endpoint = base_url.rstrip("/") + "/chat/completions"
+    headers: Dict[str, str] = {"Content-Type": "application/json"}
+
+    # 兼容需要API Key的代理/服务（llama-server通常不强制）
+    api_key = os.getenv("OPENAI_API_KEY")
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
+
+    payload: Dict[str, Any] = {
+        "model": model,
+        "messages": [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt},
+        ],
+        "temperature": temperature,
+        "max_tokens": max_tokens,
+    }
+
+    resp = requests.post(endpoint, headers=headers, data=json.dumps(payload), timeout=timeout)
+    resp.raise_for_status()
+    return resp.json()
+
+
+def main() -> None:
+    args = build_args()
+
+    user_prompt = args.user
+    if not user_prompt:
+        try:
+            user_prompt = input("请输入用户提示词: ")
+        except KeyboardInterrupt:
+            print("\n已取消。")
+            sys.exit(1)
+
+    # 解析系统提示词：优先使用 --system-file
+    system_prompt = args.system
+    if args.system_file:
+        try:
+            with open(args.system_file, "r", encoding="utf-8") as f:
+                system_prompt = f.read()
+        except Exception as e:
+            print("\n❌ 读取系统提示词文件失败：")
+            print(str(e))
+            sys.exit(1)
+
+    try:
+        print("--- llama-server 推理 ---")
+        print(f"Base URL: {args.base_url}")
+        print(f"Model: {args.model}")
+        if args.system_file:
+            print(f"System(from file): {args.system_file}")
+        else:
+            print(f"System: {system_prompt}")
+        print(f"User: {user_prompt}")
+
+        data = call_llama_server(
+            base_url=args.base_url,
+            model=args.model,
+            system_prompt=system_prompt,
+            user_prompt=user_prompt,
+            temperature=args.temperature,
+            max_tokens=args.max_tokens,
+            timeout=args.timeout,
+        )
+
+        if args.verbose:
+            print("\n完整返回JSON:")
+            print(json.dumps(data, ensure_ascii=False, indent=2))
+
+        # 尝试按OpenAI兼容格式提取assistant内容
+        content = None
+        try:
+            content = data["choices"][0]["message"]["content"]
+        except Exception:
+            pass
+
+        if content is not None:
+            print("\n模型输出:")
+            print(content)
+        else:
+            # 兜底打印
+            print("\n无法按OpenAI兼容字段解析内容，原始返回如下：")
+            print(json.dumps(data, ensure_ascii=False))
+
+    except requests.exceptions.RequestException as e:
+        print("\n❌ 请求失败：请确认 llama-server 已在 8081 端口启动并可访问。")
+        print(f"详情: {e}")
+        sys.exit(2)
+    except Exception as e:
+        print("\n❌ 发生未预期的错误：")
+        print(str(e))
+        sys.exit(3)
+
+
+if __name__ == "__main__":
+    main()
+
+

tools/test_validate/test_results.csv

@@ -1 +1,12 @@
 instruction_index,instruction,run_number,success,attempts,response_time,plan_id,error,timestamp
+1,起飞后移动到学生宿舍上方降落,1,False,1,2.048215866088867,,,2025-12-02 20:44:56
+2,起飞后移动到学生宿舍上方查找蓝色的车,1,True,1,14.806509971618652,8a6f282e-c306-4249-962c-d47d48c31bad,,2025-12-02 20:45:12
+3,起飞后移动到学生宿舍上方寻找蓝色的车,1,True,1,15.240672826766968,f298e2f4-9295-4ffd-8fff-0d0eb9a0ee6c,,2025-12-02 20:45:28
+4,起飞后移动到学生宿舍上方检测蓝色的车,1,True,1,13.8105788230896,31733491-2030-43b1-a5e4-eb1300b8d23f,,2025-12-02 20:45:43
+5,飞到学生宿舍上方查找蓝色的车,1,True,1,12.74257755279541,4c855ef4-c251-48cd-b464-4816bc62fbb5,,2025-12-02 20:45:57
+6,飞到学生宿舍上方查找蓝色车辆并进行打击,1,True,1,16.117226600646973,63d0e7c3-dcbb-40f0-b76b-6f0191c6512f,,2025-12-02 20:46:14
+7,起飞后移动到学生宿舍上方搜索蓝色车辆，并进行打击,1,True,1,16.25989079475403,1b4a537e-c1be-4abf-897e-c21b677b83b7,,2025-12-02 20:46:31
+8,起飞到学生宿舍上方搜索被困人员，并为被困人员投递救援物资,1,True,1,16.014280796051025,f88ea46f-5e0b-48fb-b1da-326d287af3d6,,2025-12-02 20:46:48
+9,飞到学生宿舍上方搜索方圆10米范围内的蓝色车辆,1,True,1,15.530286073684692,f56c811a-8304-4c68-8260-01643928bf3e,,2025-12-02 20:47:05
+10,飞到学生宿舍上方搜索半径为10米区域范围内的蓝色车辆,1,True,1,16.660754919052124,07a13346-3026-4dce-a976-4e0faa132248,,2025-12-02 20:47:23
+11,起飞到学生宿舍搜索有没有被困人员，然后抛洒救援物资,1,True,1,14.128317832946777,16426d41-4f02-4e27-a05e-f4eb84d6c935,,2025-12-02 20:47:38

tools/test_validate/test_summary.csv

@@ -1,12 +1,12 @@
 instruction_index,instruction,total_runs,successful_runs,success_rate,avg_response_time,min_response_time,max_response_time,total_response_time
-1,起飞后移动到学生宿舍上方降落,10,10,100.00%,7.91s,7.73s,8.96s,79.07s
-2,起飞后移动到学生宿舍上方查找蓝色的车,10,9,90.00%,14.18s,9.40s,25.44s,127.59s
-3,起飞后移动到学生宿舍上方寻找蓝色的车,10,10,100.00%,11.95s,7.00s,13.08s,119.49s
-4,起飞后移动到学生宿舍上方检测蓝色的车,10,10,100.00%,12.21s,9.09s,13.09s,122.08s
-5,飞到学生宿舍上方查找蓝色的车,10,10,100.00%,12.99s,12.79s,13.07s,129.90s
-6,飞到学生宿舍上方查找蓝色车辆并进行打击,10,10,100.00%,19.17s,19.08s,19.38s,191.70s
-7,起飞后移动到学生宿舍上方搜索蓝色车辆，并进行打击,10,10,100.00%,19.18s,19.07s,19.25s,191.85s
-8,起飞到学生宿舍上方搜索被困人员，并为被困人员投递救援物资,10,10,100.00%,15.65s,14.89s,15.91s,156.47s
-9,飞到学生宿舍上方搜索方圆10米范围内的蓝色车辆,10,10,100.00%,14.04s,13.97s,14.15s,140.40s
-10,飞到学生宿舍上方搜索半径为10米区域范围内的蓝色车辆,10,10,100.00%,11.64s,11.29s,14.16s,116.39s
-11,起飞到学生宿舍搜索有没有被困人员，然后抛洒救援物资,10,10,100.00%,16.02s,14.96s,24.95s,160.19s
+1,起飞后移动到学生宿舍上方降落,1,0,0.00%,N/A,N/A,N/A,0.00s
+2,起飞后移动到学生宿舍上方查找蓝色的车,1,1,100.00%,14.81s,14.81s,14.81s,14.81s
+3,起飞后移动到学生宿舍上方寻找蓝色的车,1,1,100.00%,15.24s,15.24s,15.24s,15.24s
+4,起飞后移动到学生宿舍上方检测蓝色的车,1,1,100.00%,13.81s,13.81s,13.81s,13.81s
+5,飞到学生宿舍上方查找蓝色的车,1,1,100.00%,12.74s,12.74s,12.74s,12.74s
+6,飞到学生宿舍上方查找蓝色车辆并进行打击,1,1,100.00%,16.12s,16.12s,16.12s,16.12s
+7,起飞后移动到学生宿舍上方搜索蓝色车辆，并进行打击,1,1,100.00%,16.26s,16.26s,16.26s,16.26s
+8,起飞到学生宿舍上方搜索被困人员，并为被困人员投递救援物资,1,1,100.00%,16.01s,16.01s,16.01s,16.01s
+9,飞到学生宿舍上方搜索方圆10米范围内的蓝色车辆,1,1,100.00%,15.53s,15.53s,15.53s,15.53s
+10,飞到学生宿舍上方搜索半径为10米区域范围内的蓝色车辆,1,1,100.00%,16.66s,16.66s,16.66s,16.66s
+11,起飞到学生宿舍搜索有没有被困人员，然后抛洒救援物资,1,1,100.00%,14.13s,14.13s,14.13s,14.13s

tools/test_validate/test_validity.py

@@ -17,7 +17,7 @@ SUMMARY_CSV = "test_summary.csv"
 LOG_FILE = "api_test_log.txt"
 
 # 测试参数
-TESTS_PER_INSTRUCTION = 10
+TESTS_PER_INSTRUCTION = 1
 MAX_RETRIES = 3
 RETRY_DELAY = 2
 
@@ -207,20 +207,38 @@ def read_instructions(filename):
         return []
 
 def write_log_entry(log_file, instruction_idx, run_number, prompt, result):
-    """写入详细日志"""
+    """写入详细日志，包含完整的API响应"""
     timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
     
     with open(log_file, 'a', encoding='utf-8') as f:
         f.write(f"\n{'='*80}\n")
         f.write(f"指令 #{instruction_idx} - 运行 #{run_number} - {timestamp}\n")
         f.write(f"HTTP状态: {result.get('http_status', 'N/A')}\n")
-        f.write(f"指令: {prompt}\n")
+        f.write(f"原始指令: {prompt}\n")
         f.write(f"尝试次数: {result['attempts']}\n")
         f.write(f"响应时间: {result['response_time']:.2f}秒\n")
         f.write(f"结果: {'✅ 成功' if result['success'] else '❌ 失败'}\n")
         
-        if result['success']:
-            f.write("验证结果:\n")
+        if result['success'] and result.get('data'):
+            data = result['data']
+            
+            # 提取并记录组织后的prompt（如果存在）
+            organized_prompt = None
+            if isinstance(data, dict):
+                organized_prompt = data.get("organized_prompt") or \
+                                 data.get("processed_prompt") or \
+                                 data.get("final_prompt") or \
+                                 data.get("enhanced_prompt") or \
+                                 data.get("user_prompt_enhanced")
+            
+            if organized_prompt:
+                f.write(f"\n📝 组织后的Prompt:\n")
+                f.write(f"{organized_prompt}\n")
+            else:
+                f.write(f"\n📝 组织后的Prompt: (未在响应中返回)\n")
+            
+            # 记录验证结果
+            f.write("\n验证结果:\n")
             for check_name, check_result in result['validation_checks'].items():
                 f.write(f"  {check_name}: {'✅' if check_result else '❌'}\n")
             
@@ -229,8 +247,27 @@ def write_log_entry(log_file, instruction_idx, run_number, prompt, result):
             
             if result['invalid_conditions']:
                 f.write(f"⚠️  无效条件节点: {result['invalid_conditions']}\n")
+            
+            # 记录完整的API响应
+            f.write(f"\n完整API响应:\n")
+            try:
+                response_json = json.dumps(data, indent=2, ensure_ascii=False)
+                f.write(response_json)
+                f.write("\n")
+            except Exception as e:
+                f.write(f"⚠️  无法序列化响应数据: {e}\n")
+                f.write(f"原始数据: {str(data)}\n")
         else:
             f.write(f"错误信息: {result['error']}\n")
+            # 即使失败也尝试记录响应数据（如果有）
+            if result.get('data'):
+                f.write(f"\n部分响应数据:\n")
+                try:
+                    response_json = json.dumps(result['data'], indent=2, ensure_ascii=False)
+                    f.write(response_json)
+                    f.write("\n")
+                except Exception:
+                    f.write(f"原始数据: {str(result['data'])}\n")
 
 def generate_summary_report(instructions, results_summary):
     """
@@ -311,6 +348,10 @@ def main():
             write_log_entry(LOG_FILE, instruction_idx, run_number, prompt, result)
             
             # 记录结果
+            plan_id = ""
+            if result.get("success") and result.get("data") and isinstance(result["data"], dict):
+                plan_id = result["data"].get("plan_id", "")
+            
             detailed_result = {
                 "instruction_index": instruction_idx,
                 "instruction": prompt,
@@ -318,7 +359,7 @@ def main():
                 "success": result["success"],
                 "attempts": result["attempts"],
                 "response_time": result["response_time"],
-                "http_status": result.get("http_status"),
+                "plan_id": plan_id,
                 "error": result["error"] or "",
                 "timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S")
             }
@@ -363,6 +404,10 @@ def main():
     # 生成统计摘要
     generate_summary_report(instructions, results_summary)
     
+    # 计算总统计
+    total_tests = len(instructions) * TESTS_PER_INSTRUCTION
+    total_successful = sum(summary['success_count'] for summary in results_summary)
+    
     # 打印最终统计
     print(f"\n{'='*60}")
     print("📈 最终测试统计")
@@ -370,7 +415,10 @@ def main():
     print(f"总测试次数: {total_tests}")
     print(f"成功次数: {total_successful}")
     print(f"失败次数: {total_tests - total_successful}")
-    print(f"总成功率: {(total_successful / total_tests * 100):.2f}%")
+    if total_tests > 0:
+        print(f"总成功率: {(total_successful / total_tests * 100):.2f}%")
+    else:
+        print(f"总成功率: N/A")
     
     # 打印每个指令的统计
     print(f"\n📋 每个指令的统计:")