Dev Container Workflow

[KCui0327]

🚀 Description

Mist will need to support two different types of user workflows:

1. The user receives a dev. container for them to develop with and have the freedom to download packages they need
2. The user submits a job to Mist to do either:
   • ML Model Training
   • Hosting a Flask Server (For GenAI Genesis)

For this issue, you will ONLY focus on the first workflow.

---

✅ Objectives

🧑‍💻 Development Container Workflow

• Provide a Docker Container where:
  • Container is provisioned per user session
  • Runs as an unprivileged user
  • No sudo access or ability to modify system directories
  • Only user-space installs allowed:
    • Python packages via pip install --user
    • Conda packages (if Conda is included)
    • Binaries in home directory or workspace
  • Preinstalled Stack:
    • Python 3.10+
    • Libraries: torch, transformers, datasets, pandas, scikit-learn, etc.
    • Tools: git, curl, vim, wget, jupyter
  • Persistent Storage
    • Per-User Volume:
      • Each user is provisioned a dedicated persistent volume
      • Data survives container shutdowns or restarts
      • Isolation:
        • Only accessible by the owning user
        • No cross-user access permitted

---

🎯 Acceptance Criteria

• Environment Behaviour
  • Persistent volume
  • Preinstalled packages and tools available
  • User can install packages to their user space
  • All system-level modification attempts (e.g., apt install) are blocked
• Testing
  • Unit tests for container provisioning logic
  • Integration test for user package install (e.g., pip install --user matplotlib)
  • Integration test for persistent file write/read
  • Security test: confirm restricted access to system directories