Code Execution

CodingFleet provides an integrated Code Execution feature, allowing AI-generated code to run securely in a sandboxed environment. This enhances your workflow by automating testing and immediate feedback.

Availability: Since it's in Beta, it is available to all users.

Supported Languages

Currently, we support the following languages: Python, JavaScript, TypeScript, Ruby, Golang, Rust, C, C++, C#, Java, PHP, Perl, R, Kotlin, and few others.

Supported databases include:

  • PostgreSQL
  • SQLite
  • MySQL
  • MongoDB

Need support for additional languages or databases? Contact us directly, and we'll be happy to add support.

How It Works

  • Once enabled, your code is executed within a secure, isolated sandbox environment.
  • Every user maintains access to their sandbox continuously as long as they keep executing code within a 24-hour window.
  • The sandbox resets if no code execution occurs within a 24-hour window.
  • On the first execution of a specific language, the necessary environment and dependencies will install automatically, causing a slight initial delay. Subsequent runs will execute faster.
  • You can upload files to the sandbox environment where they will appear in the /uploads folder. Ensure "Code Execution" feature is enabled before uploading, otherwise it'll not upload to the sandbox. This feature is only available for Elite users. Maximum file size is 50MB.
  • You can download files from the sandbox environment, instruct the AI to download certain files residing in the environment such as graphs, reports, JSON, CSVs, or anything that is in the sandbox system resulted from code execution.

Supported Models

The following AI models support Code Execution:

  • OpenAI o3-mini (and o3-mini-high)
  • GPT-4o
  • GPT-4o mini
  • Claude 3.7 Sonnet Thinking
  • Claude 3.7 Sonnet
  • Claude 3.5 Sonnet
  • Claude 3.5 Haiku

We'll continuously expand model compatibility.

Important Notes and Recommendations

Web Apps or Continuous Processes

When testing web applications or other persistent processes, make sure to instruct the AI to run these applications in the background. This avoids hanging and allows further automated testing.

Claude 3.7 Sonnet Specific Limitations

During our testing, Claude 3.7 Sonnet is highly detailed and may produce overly complex solutions even for simple tasks, significantly increasing execution time. To prevent delays:

  • Explicitly instruct Claude 3.7 Sonnet to provide concise solutions.
  • Consider using alternative models for straightforward tasks requiring rapid execution.

Current Limitations

  • When using the chat tool, the AI may sometime simulate code execution rather than actually executing it. If that happens, ensure you instruct it to "actually execute the code using execute_code tool".

 

If you have any questions or need further assistance, please contact us.

   Last Updated: apr 2025