When a student enters the MLflow Jedi activity, they step into a fully themed Star Wars classroom experience that teaches MLflow experiment tracking from the ground up. Starting from a session join screen, they progress through three distinct phases: unlocking six sequential concept chambers that build the vocabulary of MLflow tracking, completing three escalating missions that apply the full 5–4–2–1 tracking pattern to a real scikit-learn Random Forest experiment, and finally comparing their tracked runs side by side before drawing on a unique holocron challenge to defend a model recommendation in front of the class. Every interaction — writing code, requesting hints, submitting a defense — is reflected in the live team score, making the journey both educational and competitive.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.
Joining a session
The instructor generates a session with a unique code in the format JEDI-XXXX (e.g.JEDI-4821). Students visit the app, enter that code, and choose one of the ten pre-defined team names from the dropdown:
- Guardianes de Naboo
- Centinelas de Coruscant
- Exploradores de Tatooine
- Cronistas de Alderaan
- Custodios de Dagobah
- Vigías de Endor
- Navegantes de Hoth
- Forjadores de Mustafar
- Archiveros de Jedha
- Protectores de Lothal
Activity phases
Phase 1 — The Six Chambers
Teams solve six sequential code challenges, one per core MLflow concept: Experiment, Run, Parameters, Metrics, Artifacts, and Model. Each chamber presents a short question, a code editor pre-filled with starter code, and up to three progressive hints. A correct answer earns a seal and unlocks the next chamber. After all six seals are collected, the team advances automatically to Phase 2.
Phase 2 — Three Missions
Teams complete three escalating missions — Tatooine (guided), Coruscant (sabotaged), and Mustafar (autonomous) — each requiring them to write a complete MLflow tracking block following the 5–4–2–1 pattern: five parameters, four metrics, two artifacts, and one model. Submitted code is validated structurally using Python’s
ast module; it is never executed. On success, the app runs an internal safe template and stores the resulting metrics for comparison.Phase 3 — Comparator & Defense
With all three mission runs tracked, teams open the comparator to view metrics and artifacts for each run side by side. They then press a button to receive their unique holocron challenge card — one of seven differentiated scenarios that frame a specific angle for model selection. Finally, they complete and seal a seven-field structured defense worth 25 points.
Scoring summary
| Event | Points |
|---|---|
| Chamber solved correctly (×6) | +20 per chamber |
| Each hint used | −5 per hint |
| Forced advance after 5 failures | −10 per chamber |
| Each mission completed (×3) | +30 per mission |
| All 3 missions completed bonus | +15 |
| Defense submitted | +25 |
| Maximum possible | 250 points |
Where to go next
The Six Chambers
Detailed guide to each chamber, attempt rules, hint structure, and the allowed MLflow calls validated by the AST engine.
Three Missions
How the Tatooine, Coruscant, and Mustafar missions work, what the 5–4–2–1 pattern means, and what the validator checks.
Comparator & Defense
How to read the run comparator, receive a holocron challenge, and submit a structured evidence-based model defense.
Quickstart
Instructor setup: generating a session code, managing teams, and exporting results.