Tour · 3 walkthroughs · 2:59 in total
A guided tour
Short recordings of the main workflows, captioned step by step, and screenshots of each key feature. A Playwright script in the repository records them and checks each journey end to end on the way. The models and intervals come from fixed seeds, so the numbers on screen are the ones you get when you try it yourself.
Walkthrough 01 · 0:54 · captioned, no audio
Paste a recipe
Paste a recipe into the card and the browser model sorts it into quick, medium or slow as you type, with its probabilities, the words that tipped it and the method tinted by the class each word favours. One added line flips the verdict, and the fine print gives the model's hold-out accuracy with its 95% Wilson interval.
Steps and transcript
Walkthrough 02 · 1:03 · captioned, no audio
Model bake-off
The five 2021 models, Gaussian Naive Bayes, Logistic Regression, the RBF SVM, the Random Forest and the stacked ensemble, first as the report printed them and then with their uncertainty: corrected intervals over 15 repeated folds, paired fold-by-fold comparisons, Wilson intervals on 10,000 unseen recipes, McNemar's test and calibration.
Steps and transcript
Walkthrough 03 · 1:02 · captioned, no audio
Classical vs LLM
Open the bring-your-own-key settings, run the LLM evaluation on 30 held-out recipes and read it against the 2021 Random Forest on the same recipes, with Wilson intervals, a paired bootstrap and McNemar's test. Then reject the run as its human reviewer and inspect every call in the audit log.
Mocked AI response for illustration. No real model was called in this recording: the tour chose the OpenAI adapter with the model id “showcase-mock-not-a-real-model”, typed a placeholder instead of a key, and every request to the provider was intercepted and answered by a scripted stand-in whose every cue starts with “mock reply”. The numbers in its results table describe the stand-in, not any LLM.
Steps and transcript
Key features
Screenshots
Desktop views at 1440 × 900 and phone views at 390 × 844. Select one to enlarge it; the arrow keys move between them. The two LLM screenshots show mocked AI responses for illustration.
On a phone (390 × 844)
The recorder is web/e2e/showcase.spec.ts in the repository (Private GitHub repository (available on request)). Run pnpm showcase to replay every journey and rebuild these files.