Create a simple webui with gradio #316
Conversation
|
Created a simple web ui application using Gradio. Modified some of the code for environment and company, adding two private variables for environment. One is used to temporarily store history for displaying to users and accepting user modifications, and the other indicates the environment type. After testing, it will not affect normal operation through the startup way. |
|
你好。提交的程序请遵守PEP8 |
代码已进行修改，请看看是否还有需要调整的地方 |
metagpt/app.py
Outdated
|
|
||
| def clear_logs(): | ||
| with open(PROJECT_ROOT / 'logs/log.txt', 'w') as f: | ||
| f.write("") |
There was a problem hiding this comment.
place the function in proper file
| with open(PROJECT_ROOT / 'logs/log.txt', 'w') as f: | ||
| f.write("") | ||
|
|
||
| async def startup(company_name : str, |
There was a problem hiding this comment.
PEP8. 可以尝试使用PyCharm自带的format或者black / flake8
metagpt/app.py
Outdated
| investment : float = 6.0, | ||
| n_round : int = 5, | ||
| code_review : bool = True, | ||
| staffs : list = ["ProjectManager", |
metagpt/app.py
Outdated
| company.invest(investment) | ||
| company.start_project(idea) | ||
| # report all output to webui | ||
| global SoftwareCompany_Company |
There was a problem hiding this comment.
我尝试了一下不使用全局变量，发现会比较难实现，可能需要将SoftwareCompany改为单例模式。另外，我参考了一些别的项目的gradio webui设计，发现有的也使用了全局变量。所以我希望能被允许在这里使用全局变量
metagpt/app.py
Outdated
| await company.run(n_round) | ||
| elif company_name == "SoftwareCompany_With_Human": | ||
| await company.continue_run() | ||
| return "Role: "+company.environment.short_term_history.role,\ |
metagpt/app.py
Outdated
| with gr.Tabs(): | ||
| with gr.TabItem("MetaGPT") as generate_tab: | ||
| company_choise = gr.Dropdown(label = "Choose the company type",\ | ||
| choices = ["SoftwareCompany", "SoftwareCompany_With_Human"],\ |
There was a problem hiding this comment.
请问这里应该如何修改为PEP8风格？是这样吗？
with gr.TabItem("MetaGPT") as generate_tab:
company_choise = gr.Dropdown(
label = "Choose the company type",
choices = ["SoftwareCompany", "SoftwareCompany_With_Human"],
value = "SoftwareCompany_With_Human"
)There was a problem hiding this comment.
@Ori-Replication 指的是字符串：class驼峰，函数下划线，但是两种风格不能混用。并且字符串与class name需要对应
metagpt/app_chs.py
Outdated
| """ | ||
| @Time : 2023/9/19 17:12 | ||
| @Author : ORI_Replication | ||
| @File : app_chs.py |
There was a problem hiding this comment.
一般而言，i18n的最佳方法是独立k个语言文件，或者一个大的EXCEL / JSON，用来做槽位与指定语言的映射。不推荐copy & paste代码来解决这个问题，因为它的维护成本太高了
metagpt/environment.py
Outdated
| """ | ||
|
|
||
| environment_type: str = Field(default='SoftwareCompany') | ||
| short_term_history: str = Field(default_factory=Memory) |
There was a problem hiding this comment.
这里short_term_history声明Memory，但是在使用Message的方式对待它。或许不声明这个变量，而是在Memory中解决这个会更好
metagpt/environment.py
Outdated
| """ | ||
|
|
||
| environment_type: str = Field(default='SoftwareCompany') |
There was a problem hiding this comment.
现在environment没有__init__ 函数，我是否可以进行定义？
There was a problem hiding this comment.
@Ori-Replication 我的意思是：不一定需要使用字符串来表达类型。表达类型的最佳实践是使用类型本身，或者使用enum
metagpt/software_company.py
Outdated
|
|
||
| class SoftwareCompanyWithHuman(BaseModel): |
|
需要解决conflicts |
metagpt/app_chs.py
Outdated
There was a problem hiding this comment.
删除SoftwareCompanyWithHuman类
There was a problem hiding this comment.
- 修改了short_term_memory的类型，成为了message类型
- 将环境类型改为human_interaction变量，如果为True，则不直接将message储存到记忆中与历史中，而是在用户修改后再储存。在app.py中有具体实现方式
|
我不太会解决这个conflict。在这一版中我将新增的一些需求的包加入进来了，但还是显示有冲突，显示我添加的gradio包有冲突。我不太会解决，请问该如何解决？
|
使用gradio创建了一个简单的webui应用。
修改了部分environment和company的代码，为environment新增了两个私有变量，一个用来临时储存history，来展示给用户和接受用户修改，另一个标明environment类型。经测试不会影响正常通过startup途径运行。