👩‍💻 In case you missed it: We've included an out-of-the-box code execution agent in the Gemini APIs, that will write, execute, and recursively fix and improve Python code for you in a sandboxed environment. All you have to do is add "code_execution" as a tool: model = genai.GenerativeModel( model_name="gemini-2.0-flash-exp",... generation_config=generation_config, tools='code_execution', )show more

is there a way for it to use more libraries ?

You can't use more than the pre-installed libraries just yet... but please share if there are any libraries that you'd like to have installed, that aren't yet included!

In this Free eBook, we’ll walk you through Powershell scripting basics, show you Powershell commands and scripts for performing the most common administrative tasks, and explain how you can schedule your Powershell scripts and сommands. Get your free copy ⬇️

it’s happening. i will become a code bot babysitter and mostly do infra

Are there plans to enable this as part of a function call rather than as a native tool in the Gemini api? In the event we might want to say generate multiple code chunks to be executed independently or sequentially, would that be supported?

You should be able to enable automatic function calling, and call multiple functions in a single turn: Would that work for the use cases you're thinking of?

Cool,

You don’t explain how and where Python runs and why Python??

:) tweets are still but tiny methods of communicating, more information in the docs:

code execution can output image / plots now? what MIME type of outputs? I recall first iteration of code execution can only generate... basically utf-8, right?

Please make multiple tools to enable simultaneously