bolt
bolt
同遇到，问题点上没人回答？各种坑。。。这开源的东西放出来何意义？
> PS:个人认为这种开源搞了个常规的框架，关于LLM在测试上的应用是完全黑盒的，私以为真正有价值的思路并没透露。。。 近一年业界关于LLM在自动化测试上的应用包括相关Agent的应用设计这一类的研究paper也非常多
> Hi, > > We would greatly appreciate it if you could provide a full reproduction scripts. Based on the provided screenshots, it is not clear which models you were...
prompt = ''' 你是接口自动化测试大师，帮肋软件工程师根据给出的接口定义+入参，来生成有序和结构化的自动化测试用例，\ 如以 /iplan/task/pack/list接口为例，有8个类型都为string的输入参数 name,id,status,owner,page,page_size,vendor_task_id,business_id,\ 生成的接口自动化用例样例格式cases:[{\"case1\":\"name=null,id=-999,status=false,owner=xiaokeai,page=1,page_size=-100,vendor_task_id="",business_id=""\",{\"case2\":\"...\"}] \ 请设计覆盖输入参数的正常、异常组合的用例,生成不少于50条,按我所提供的样例格式并用中文返回" ''' @jklj077 prompt可按这个去调试试下
example: browser = await playwright.chromium.launch(headless=False, channel="chrome") #using existed chrome browser as webdriver
thanks Takuto Yuki ***@***.***> 于2025年2月16日周日 23:12写道： > @bolt163 > Please note that the Browser class in browser-use is completely different > from Playwright’s Browser; they are entirely separate in design...
+1 the same problem
the same +1 has so many bad cases with 'content': 'Xiaohongshu (Little Red Book)' , which actual content were not
验证场景估计非常难，自己想其他办法吧，别指望LLM对抗AI