CortexLab V9 · container-per-job execution

Every code submission gets a fresh disposable runtime.

V9 sends Python JavaScript Bash and SQL to a local job controller. The controller launches one temporary container for that single job with networking disabled and strict CPU memory PID filesystem and time limits then destroys it.

isolated executor

multi-language code

print('hello from cortex v9')
xs=[1,2,3,4]
print([x*x for x in xs])

execution result

run code to inspect stdout stderr exit code and duration.

fresh container

Each submission receives a new runtime rather than sharing a long-lived execution container with previous jobs.

network off

Job containers launch with --network none and only the submitted source file is bind-mounted read-only.

hard lifecycle

Wall time output memory CPU and process counts are bounded and the runtime is automatically removed after execution.

Production isolation note. The development controller uses the Docker socket which is privileged. Put the controller on an isolated/rootless worker host before accepting untrusted public code.