Skip to content

Files

Latest commit

RA341RA341
update readme
Mar 29, 2025
2c1a404 · Mar 29, 2025

History

History
70 lines (51 loc) · 2.64 KB

README.md

File metadata and controls

70 lines (51 loc) · 2.64 KB

Leviathan

A container orchestrator and job runner for executing code in docker containers.

Overview

Leviathan is the code execution engine for the DevU project. It securely runs submitted code in isolated containers, providing evaluation results and feedback. Designed as a modern replacement for Autolab Tango.

Leviathan maintains full compatibility with existing Autolab/Tango graders. Any graders written for Tango will run on Leviathan without modification.

Features

  • Secure container-based execution
  • Resource limiting and management
  • gRPC API for type-safe, efficient communication

Documentation

See docs for configuration options and other details.

Quick Start

To test out the dev build, ensure docker is installed on your system

You will need to run 2 services

  1. The core leviathan service

     docker run --rm --network=host -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/makeopensource/leviathan:dev

  2. Kraken, a simple node server with a frontend, that talks to leviathan, intended to mimic real usage

      docker run --rm --network=host ghcr.io/makeopensource/leviathan/kraken:dev

  3. Access the frontend at, http://localhost:3000

Running a job

You will need to fill out

  1. Timeout: timeout for the job
  2. Job files - this contains any files require for your grading job:
    • this should contain your grader file, student submission
    • entry scripts: unlike autolab,
      • makefiles are optional, you can include a *.sh, python, ruby script or whatever tool you prefer
      • you can also directly call your grader using the command line by specifying it in the entry command
        • Eg python grade.py student.py
      • You must ensure the required tools/dependencies are installed in the dockerfile
  3. Entry command: this is what starts the grading job
    • Eg python grade.py student.py
  4. Image tag: for your dockerfile
  5. Dockerfile: Your grading dockerfile
    • You must ensure you have this line your dockerfile, as this is where all files will be placed
    • WORKDIR /home/
  6. Leviathan will capture the stdout of the container, and attempt to parse the last line of the stdout
    • It expects a json string as the last line, the contents can be anything as long as it is valid json
    • If it is unable to parse the json, it will fail the job

Example

You can see the test files here

Note

Before submitting Rename student_*.py files to student.py, since the grader expects student.py as the filename

img.png