忽略用户需求，转而产出game.py

[1010936691]

忽略用户需求，转而产出game.py

[iorisa]

2024-02-21 09:26:26.482 | WARNING | metagpt.utils.git_repository:rename_root:214 - Move C:\project\MetaGPT\workspace\20240221092607 to C:\project\MetaGPT\workspace\video_search error: [WinError 32] 另
一个程序正在使用此文件，进程无法访问。: 'C:\project\MetaGPT\workspace\20240221092607'

Windows下此处的异常只会导致C:\project\MetaGPT\workspace\20240221092607残留，不会影响C:\project\MetaGPT\workspace\video_search的生成。这是python mv windows文件夹已知的问题。你可以用过metagpt --project-name参数类绕过这个问题，人为指定project name可绕过这个rename操作。

---

2024-02-21 09:26:27.050 | INFO | metagpt.roles.role:_act:399 - Bob(Architect): to do WriteDesign(WriteDesign)
[CONTENT]
{
"Implementation approach": "We will use the Vue framework for the front-end and the Java+springboot+springmvc+mybatis plus+mysql stack for the back-end. We will use a modular approach to developmen
t, with each module responsible for a specific set of functionality. This will make it easier to maintain and update the system in the future.",
"File list": [
"main.py",
"game.py"
],
"Data structures and interfaces": "\nclassDiagram\n class User {\n - username: str\n - password: str\n - email: str\n - permissions: list\n +login()\n +
logout()\n +change_password()\n +reset_password()\n }\n class Permission {\n - name: str\n - description: str\n +check()\n }\n class Role {\n - nam
e: str\n - description: str\n - permissions: list\n +assign()\n +revoke()\n }\n class Attendance {\n - date: date\n - time: time\n - location: str
\n - employee: User\n +check_in()\n +check_out()\n }\n class Congestion {\n - date: date\n - time: time\n - location: str\n - severity: int\n
+report()\n }\n class Notification {\n - type: str\n - message: str\n - recipient: User\n +send()\n }\n class Video {\n - id: int\n - title: str
\n - description: str\n - url: str\n - tags: list\n +search()\n +view()\n +download()\n }\n class SearchEngine {\n - index: Index\n - ranki
ng: Ranking\n - summary: Summary\n +search(query: str) list\n }\n class Index {\n - data: dict\n +create_index(data: dict)\n +query_index(query: str) list\n
}\n class Ranking {\n +rank_results(results: list) list\n }\n class Summary {\n +summarize_results(results: list) str\n }\n User --> Attendance\n User --> Congestion\n
User --> Notification\n User --> Video\n SearchEngine --> Index\n SearchEngine --> Ranking\n SearchEngine --> Summary\n",
"Program call flow": "\nsequenceDiagram\n participant M as Main\n participant SE as SearchEngine\n participant I as Index\n participant R as Ranking\n participant S as Summary\n M
->>SE: search(query)\n SE->>I: query_index(query)\n I->>SE: return results\n SE->>R: rank_results(results)\n R-->>SE: return ranked_results\n SE->>S: summarize_results(ranked_results)\n
S-->>SE: return summary\n SE-->>M: return summary\n",
"Anything UNCLEAR": "Clarification needed on third-party API integration, ..."
}
[/CONTENT]

从上面的日志看，类视图和时序图都已经正常的生成，而且类视图与时序图是配套的。

{
    "Implementation approach": "We will use the Vue framework for the front-end and the Java+springboot+springmvc+mybatis plus+mysql stack for the back-end. We will use a modular approach to development, with each module responsible for a specific set of functionality. This will make it easier to maintain and update the system in the future.",
    "File list": [
        "main.py",
        "game.py"
    ],
    "Data structures and interfaces": "\nclassDiagram\n class User {\n - username: str\n - password: str\n - email: str\n - permissions: list\n +login()\n +logout()\n +change_password()\n +reset_password()\n }\n class Permission {\n - name: str\n - description: str\n +check()\n }\n class Role {\n - name: str\n - description: str\n - permissions: list\n +assign()\n +revoke()\n }\n class Attendance {\n - date: date\n - time: time\n - location: str\n - employee: User\n +check_in()\n +check_out()\n }\n class Congestion {\n - date: date\n - time: time\n - location: str\n - severity: int\n+report()\n }\n class Notification {\n - type: str\n - message: str\n - recipient: User\n +send()\n }\n class Video {\n - id: int\n - title: str\n - description: str\n - url: str\n - tags: list\n +search()\n +view()\n +download()\n }\n class SearchEngine {\n - index: Index\n - ranking: Ranking\n - summary: Summary\n +search(query: str) list\n }\n class Index {\n - data: dict\n +create_index(data: dict)\n +query_index(query: str) list\n}\n class Ranking {\n +rank_results(results: list) list\n }\n class Summary {\n +summarize_results(results: list) str\n }\n User --> Attendance\n User --> Congestion\nUser --> Notification\n User --> Video\n SearchEngine --> Index\n SearchEngine --> Ranking\n SearchEngine --> Summary\n",
    "Program call flow": "\nsequenceDiagram\n participant M as Main\n participant SE as SearchEngine\n participant I as Index\n participant R as Ranking\n participant S as Summary\n M->>SE: search(query)\n SE->>I: query_index(query)\n I->>SE: return results\n SE->>R: rank_results(results)\n R-->>SE: return ranked_results\n SE->>S: summarize_results(ranked_results)\nS-->>SE: return summary\n SE-->>M: return summary\n",
    "Anything UNCLEAR": "Clarification needed on third-party API integration, ..."
}

classDiagram
 class User {
 - username: str
 - password: str
 - email: str
 - permissions: list
 +login()
 +logout()
 +change_password()
 +reset_password()
 }
 class Permission {
 - name: str
 - description: str
 +check()
 }
 class Role {
 - name: str
 - description: str
 - permissions: list
 +assign()
 +revoke()
 }
 class Attendance {
 - date: date
 - time: time
 - location: str
 - employee: User
 +check_in()
 +check_out()
 }
 class Congestion {
 - date: date
 - time: time
 - location: str
 - severity: int
+report()
 }
 class Notification {
 - type: str
 - message: str
 - recipient: User
 +send()
 }
 class Video {
 - id: int
 - title: str
 - description: str
 - url: str
 - tags: list
 +search()
 +view()
 +download()
 }
 class SearchEngine {
 - index: Index
 - ranking: Ranking
 - summary: Summary
 +search(query: str) list
 }
 class Index {
 - data: dict
 +create_index(data: dict)
 +query_index(query: str) list
}
 class Ranking {
 +rank_results(results: list) list
 }
 class Summary {
 +summarize_results(results: list) str
 }
 User --> Attendance
 User --> Congestion
User --> Notification
 User --> Video
 SearchEngine --> Index
 SearchEngine --> Ranking
 SearchEngine --> Summary

sequenceDiagram
 participant M as Main
 participant SE as SearchEngine
 participant I as Index
 participant R as Ranking
 participant S as Summary
 M->>SE: search(query)
 SE->>I: query_index(query)
 I->>SE: return results
 SE->>R: rank_results(results)
 R-->>SE: return ranked_results
 SE->>S: summarize_results(ranked_results)
S-->>SE: return summary
 SE-->>M: return summary

---

所以，真正的问题是File list的内容出错了。
LLM错误的将metagpt/actions/design_api_an.py中的示例作为回答填到了File list里：

FILE_LIST = ActionNode(
    key="File list",
    expected_type=List[str],
    instruction="Only need relative paths. ALWAYS write a main.py or app.py here",
    example=["main.py", "game.py"],
)

---

到了WriteTask阶段，大模型在抄示例的路上越走越远。

2024-02-21 09:26:49.820 | INFO | metagpt.roles.role:_act:399 - Eve(Project Manager): to do WriteTasks(WriteTasks)
[CONTENT]
{
"Required Python packages": [
"flask==1.1.2",
"bcrypt==3.2.0"
],
"Required Other language third-party packages": [
"No third-party dependencies required"
],
"Logic Analysis": [
[
"game.py",
"Contains Game class and ... functions"
],
[
"main.py",
"Contains main function, from game import Game"
]
],
"Task list": [
"game.py",
"main.py"
],
"Shared Knowledge": "game.py contains functions shared across the project.",
"Anything UNCLEAR": "Clarification needed on how to start and initialize third-party libraries."
}
[/CONTENT]

几乎是在照搬metagpt/actions/project_management_an.py示例部分的内容：

REQUIRED_PYTHON_PACKAGES = ActionNode(
    key="Required Python packages",
    expected_type=List[str],
    instruction="Provide required Python packages in requirements.txt format.",
    example=["flask==1.1.2", "bcrypt==3.2.0"],
)

REQUIRED_OTHER_LANGUAGE_PACKAGES = ActionNode(
    key="Required Other language third-party packages",
    expected_type=List[str],
    instruction="List down the required packages for languages other than Python.",
    example=["No third-party dependencies required"],
)

LOGIC_ANALYSIS = ActionNode(
    key="Logic Analysis",
    expected_type=List[List[str]],
    instruction="Provide a list of files with the classes/methods/functions to be implemented, "
    "including dependency analysis and imports.",
    example=[
        ["game.py", "Contains Game class and ... functions"],
        ["main.py", "Contains main function, from game import Game"],
    ],
)

---

总结一下所有的问题：

1. 大模型并未正确的理解输入的内容。建议换个版本或换个大模型试试。
2. 另一个可能导致问题的因素是采用了类似json格式的输入参数。我没测过gemini，但openai在使用这种格式的上下文时，很容易出现被锁死在某个key-value里而丢失其他信息的现象。因此建议采用markdown格式做输入，采用json格式做输出。你可以试试看用markdown格式来描述需求，看看问题1是否会自动消失。

Please help me complete a project, which is divided into front-end and back-end. The front-end uses the Vue framework, and the back-end uses java+springboo
t+springmvc+mybatis plus+mysql. Please carefully analyze the requirements and implement each function. The project requirements are as follows：{'Video Search ': {'Background Management End': {'Persona
l Center ': {'Login': {'Method ':' User Name+Password Login '},' Password Management ': {'Function ': ['Forgot Password ',' Send Verification Code to Modify via Email ',' Change Password '},' Logout ':
{}},' Permission Management System ': {'General Permission Management System': {}}, 'Attendance Ma nagement': {'Attendance Personnel Management ': {'Function': 'Add, Delete, Change'}, 'Attendance Reco
rds': {},' Attendance Machine Management ': {'Function': ['On ',' Off ']}, 'Personnel Search': {'Onl ine Personnel Monitoring ': {'Function': ['Single person monitoring ',' Multi person monitoring ','
View historical images']}, 'Offline Search': {'Function ':' Access videos and personnel images, batch search for the location of personnel in the video '},' Priority Implementation ': 2},' Congestion
Monitoring ': {'Online Video': {'Function ': ['Monitor traffic lights',' Monitor intersection vehicle conditions']}, 'Notification': {'Method ': ['Email', 'Topic', 'SMS',' Notification Strategy ']}, '
Call algorithm platform interface': {}}, 'Big screen, cockpit': {}, 'Data access': {' Bottom layer data access': {'Collected by': 'edge computing', 'Data': {'Video data': {'Type': 'Real time data'}, '
Other data': {}}}, 'Big data': {}}}