Yonro

Yori is a proactive real-world assistant built solo for the Indie Developer Track. Its vision is simple: think ahead without constantly interrupting. Instead of opening to an empty chat box, Yori presents context-aware suggestions prepared from available signals such as time, location, real nearby places, and information the user explicitly provided. Its core loop is perceive, remember, predict, prepare, present, and let the user decide. Today's implementation demonstrates this approach through real nearby-place discovery, model-assisted lunch ranking, user-stated shopping reminders, visible memory controls, and a quiet state. More advanced scenarios remain part of the product roadmap. Suggestions are prepared and presented when the user opens the app; we do not claim background sensing or push notifications yet. What it does today (public main, commit 20c616c): - Nearby lunch: Yori finds real places (Google Places) and Meta Muse Spark ranks them with a short reason based on the data it was sent. - Shopping reminder: tell Yori once what you need; when a real store is nearby, it lines up where to pick it up. It uses only stated items, not low-stock guesses. - Everyday essentials: a card built only from real place fields for a supermarket, pharmacy, convenience store or mall (the top pick is the closest by straight-line distance). - One-tap choices: Accept saves your choice and shows Undo; it does not book or buy anything yet. - A quiet state and daily prompt budget, so Yori stays silent when nothing matters. - Visible memory: what Yori learned is shown in one "Me" view where you can correct or forget it. - A compact activity trace of real steps; Ask is secondary. Our demo video, honestly labelled: - Rendered concept animations ("Concept animation") show the vision: a rainy piano lesson with an umbrella, lunch in an unfamiliar area, passing your supermarket. One scene is labelled "Vision · not yet built"; B-roll is labelled "Illustrative footage". With roadmap scenarios (fuel, travel, flight or snow disruption), they show where the general core plus adapters is heading: Yori is not just a restaurant recommender. - The Live section, recorded in Hirosaki (the developer's city): real lunch places with a live Muse Spark ranking reason, a real store (Co-op Aomori Wattoku Store, 1.3 km) and a shopping reminder card. - A montage is labelled "Real app · demo data (simulated)". Not built yet: calendar connection, fuel (simulated only), travel and flight or snow disruption, and actions that execute on Accept. Engineering: TypeScript web app with a core decision engine and pluggable context adapters (places, weather, geolocation, simulated demo provider). Muse Spark is called only from a server-side proxy. Model responses are validated against the supplied context and an expected schema; responses that fail these checks are rejected, with a labelled fallback instead. Unit, golden and end-to-end tests cover the build.