@jupel Можете скриншот дать?
"Но в моей бесплатной версии нет этого. В платной есть?"
Модули везде одинаковые что в премиум BAS что без, для Iframe нужно писать селектор, если это то про что я думаю, опять же нужен скрин
We have completely redesigned how BAS works.
Now you can edit your projects, fix bugs, and write code with Agent.
The number of changes we've made is enormous.
Right now, you just open BAS, write a prompt describing what you want to achieve, and that's it.
The only things left to manage manually are choosing the path for your project files and compiling.
https://www.youtube.com/watch?v=kTSKLk7FfHE
And it's all completely free.
The AI module has also had a huge update: it can now read email, get temporary phone numbers, receive SMS, and use an authenticator.
Besides that, we have added two new modules: Authenticator and QR Codes.
As far as I know, the model is based on DeepSeek, most likely deepseek-v4-pro
There is currently no way to configure this through the interface. It might be possible through the files, but I think this is definitely a feature worth adding
I tested the AI on one of my own projects, and overall it is genuinely useful. It works quite well with refactoring: it can split code into functions, give them reasonable names, and generally clean up the project structure. This saves a lot of time
So I think it would be very useful to add a way to provide the AI skills with BAS specific best practices. This would make it much more reliable when working with real projects
@111cbt111 said in BrowserAutomationStudio 30.7.0 has been released:
sounds amazing!!! is anything left on the roadmap besides quality mobile emulation?
In near future there will be antidetect and ai improvements. With the new features available we can do a lot.
Here are just a few ideas: automatically generate PerfectCanvas during debug, make a loop to constantly retest and fix situations which BAS script doesn't process(new message, updated selectors), automate researching site protection for antidetect(research js code, find new methods), make mcp bindings to control swarm of bots with external tools.
which model is used by Agent?
deepseek-pro for code gen and deepseek-flash for some tasks which is not involved in code gen. Using advanced models can bring some benefits, but improving our prompts and tools can help even better. After additional testing we will add this to config, just remember that western models may be very restrictive. For example, fable will always downgrade when browser fingerprints are being mentioned.
Regarding Agent, we constantly improve harness and we have tools to do that, so things which may not as precise as required today will work tomorrow.