Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/elenacarino-max/Pildora4_ext_StarWars/llms.txt

Use this file to discover all available pages before exploring further.

MLflow Jedi — La Cámara de los Experimentos Perdidos — is an interactive classroom app that teaches the full MLflow experiment-tracking cycle through a Star Wars narrative. Students progress through six concept chambers, reconstruct three real Random Forest experiments, compare tracked metrics, and present an evidence-based recommendation — all without a single line of student code ever being executed on the server.

Introduction

Understand the learning flow, roles, and key concepts behind MLflow Jedi.

Quickstart

Get the app running locally or with Docker in under five minutes.

Student Guide

Walk through chambers, missions, the comparator, and the final defense.

Teacher Guide

Create sessions, monitor teams in real time, and export results.

How the activity works

1

Teacher creates a session

Log in to the Profesorado panel and click Crear nueva sesión. The app generates a unique JEDI-XXXX code and picks six questions — one per MLflow concept — from a bank of 50.
2

Teams join as Alumnado

Each team opens the Alumnado portal, enters the session code, and picks one of ten Star Wars team names. Up to ten teams can share a single session.
3

Students solve chambers and missions

Six concept chambers unlock in sequence. Each correct answer earns 20 points. After all six seals, teams complete three missions — Tatooine (guided), Coruscant (sabotaged), and Mustafar (autonomous) — that register real MLflow runs.
4

Teams defend a decision

After comparing metrics and artefacts, each team receives a holocron card with a unique challenge and submits a structured defense. The highest possible score is 250 points.

Key features

Safe code validation

Student code is parsed with Python’s ast module — it is never executed. Only whitelisted MLflow calls are accepted.

Real MLflow tracking

Three Random Forest experiments are trained on the scikit-learn Digits dataset and tracked with MLflow’s SQLite backend.

Seven holocron challenges

Each team receives one of seven differentiated defense tasks, from audit trails to production readiness checks.

Docker deployment

A single docker compose up --build starts the app with persistent storage. No external services required.

50-question bank

Every session draws a fresh random set of six questions — one per MLflow concept — ensuring variety across sessions.

Operations guide

Configure passwords, storage paths, and team limits through environment variables. Full backup and update procedures included.
A live demo is available at pildora4extstarwars-14.streamlit.app. Create a teacher session first to get a JEDI-XXXX code before entering as a student team.

Build docs developers (and LLMs) love