Frequently asked questions¶

Top questions readers (and SEIS 666 reviewers) ask after their first pass through the book. Grouped by topic. Each answer links to the chapter or rule that goes deeper.

Pattern follows the faq-generator skill in dmccreary/claude-skills. Add a question here when it's been asked twice.

Product¶

What is the Intelligent Book vs. the app?¶

Wearly's Intelligent Book is the learning and reference companion for the styling-agent prototype. The app shows what Wearly does; the book explains how and why it reasons that way. The app is a Streamlit prototype you run locally; the book is the MkDocs-built site you're reading now, with chapters, a glossary, a learning graph, a reasoning graph, and interactive MicroSims. The two are meant to be read together — the prototype demonstrates the agent in action; the book is the structured explanation behind every line of its reasoning.

What is Wearly, in one sentence?¶

An evidence-informed, body-positive personal styling agent that reads your calendar, weather, and closet, then produces a single outfit recommendation with a visible reasoning trail you can argue with. See Chapter 1.

Who is Wearly for?¶

A single user managing a real wardrobe, who wants the 15-minute "what do I wear today" decision to feel grounded rather than guessed. The prototype is single-user and local-only.

Is Wearly a real product or a course project?¶

It's a working prototype built for SEIS 666 (Digital Transformation with AI) at the University of St. Thomas. The code runs, the documentation is the Intelligent Book you're reading, and the rule citations trace to verified sources. The roadmap chapter outlines what a production version would add.

What does Wearly NOT do?¶

It doesn't analyze body photos, infer your body shape, sell anything, talk to third-party servers (except the unauthenticated Open-Meteo weather call), or judge your wardrobe. See Chapter 10 · Privacy.

Agent reasoning¶

What makes Wearly an agent?¶

Wearly reads context first — your calendar, the weather, your closet, your fit profile — then recommends with a visible reasoning trail that cites a named rule for every decision. Add reject-and- regenerate plus persistent state across sessions and you have the full agent stance. See Chapter 1 and Chapter 3.

What are the "seven steps"?¶

Determine occasion → load profile → check weather → filter wardrobe → build outfit → check gaps → score color. Every recommendation runs this loop and surfaces a sentence (or two) per step in the reasoning panel. See Chapter 3.

What happens when I reject an outfit?¶

The rejection is recorded as an exclusion for the next run, the agent re-runs all seven steps with that constraint, and the new reasoning trail surfaces what changed. This is the reject-and-regenerate loop — arguably the most "agentic" moment in the product. See wardrobe-filtering-rules.md R5.

Why does each reasoning line carry a tag like `[occasion-rules#R8]`?¶

That's a citation chip. It resolves via rule_refs.py to a named heading inside a Skill rule file, so you (and a reviewer) can trace any sentence in the output back to a specific rule. The pattern is documented in evidence-and-references § 6a.

Personalization¶

Do I have to fill out a fit profile?¶

No. Every field on the fit profile is optional and has a safe default. If you leave body_shape blank, no body-shape-related rule fires; the agent still works. See Chapter 6.

How does Wearly decide what's "fresh"?¶

The freshness score combines a frequency penalty (how often you wear it) and a recency penalty (how recently). The score is floored at 0.4 so a much-loved piece is never pushed out of rotation. See wear-history-rules.md.

Does Wearly track what I look at, scroll past, or hover over?¶

No. It only uses explicit feedback: the reject button, the "wear this outfit today" confirmation, and the editable fit profile. No implicit behavioral tracking. This is a deliberate privacy choice — see evidence-and-references § 3.8.

Privacy¶

Where does my data go?¶

Nowhere. The prototype is local-only. The single network call is the unauthenticated Open-Meteo weather lookup; everything else stays in JSON files in the repo. See Chapter 10.

Can I delete my data?¶

Yes — delete the JSON files. The product brief commits to encrypted, opt-in, deletable data in any future production version. See Chapter 10.

Because the spirit of GDPR (view, delete, export, object) is the right design target even when the regulation doesn't technically apply. Wearly's local-only stance is the strongest form of compliance: the data never leaves the device. See evidence-and-references § 3.9.

Knowledge graph¶

What is the Wearly Knowledge Graph?¶

The Wearly Knowledge Graph is a three-layer structured-knowledge graph — domain rules, user-behavior aggregates, and runtime archetypes — generated deterministically by scripts/generate_wearly_kg.py into graph/wearly-knowledge-graph.json. 91 nodes, 187 typed edges in the seed snapshot. It's the structured context a future RAG or LLM shopping layer can query — every edge has a typed from / to / type, and the citation chain is traversable along graph hops.

What does a bigger circle mean?¶

Aggregated user signal. An item you wear often grows; a color family your closet keeps returning to grows; an occasion you attend often grows. Domain archetypes (rule packs, weather bands, skin-tone palettes) stay at base size so the user-behavior layer reads first.

Where does the per-recommendation reasoning live?¶

Inside the app, as an expander below every outfit result — a numbered reasoning trail that cites the specific rule (<pack>#R<N>) for each decision. That's the live, per-run reasoning surface; the Wearly Knowledge Graph is the structured-knowledge layer the runtime is reasoning against.

How this book works¶

What does "Level 2 intelligent textbook" mean?¶

A term from Dan McCreary's framework. A Level-2 book is built around a concept-dependency graph and ships MicroSims, glossary, FAQ, references, and Bloom-tagged concepts. Wearly is built to this level — see book/index.md.

Why are there MicroSims?¶

Because some concepts are easier to play with than to read about. Color harmony and the freshness curve are good examples. Each sim is a small self-contained p5.js or vis-network page under docs/sims/. See the sim index.

How do I cite Wearly?¶

The repo is at https://github.com/eatedalsf/styling-agent. A formal metadata file lives at book-metadata.yml in the repo root. The References page below lists the works the book itself draws on.

Pattern follows faq-generator from dmccreary/claude-skills. Answers stay short on purpose — each one points at a deeper chapter.