I Open-Sourced Vamoriq — The .NET MAUI Boundary That Keeps AI Optional

Vamoriq gives people one small daily mission to explore a new city, personalized with AI. Mikhail Petrusheuski open-sourced it (MIT) not to show that a mobile app calls AI, but to show where it refuses to depend on it: curate a valid mission, let AI adapt it within a bounded budget, then persist the result locally.

What you’ll learn

• The product decision precedes the AI call — picking a valid curated mission first, so a model outage never means a blank screen
• Failure as a normal branch — an 8-second linked cancellation budget, returning the curated fallback on unauthenticated, timed-out, malformed, or unusable responses
• Partial output doesn’t erase curated data — mapping the AI DTO field-by-field back onto curated values, and validating domain invariants for higher-risk domains
• Offline-first means persisting the decision — storing the chosen mission, streaks, and proof in SQLite keyed by local date, so continuity doesn’t depend on the network
• The curated catalog is infrastructure — a localized, tiered, de-duplicating mission library is exactly what makes the AI layer optional
• A boundary worth stealing — the questions to ask of any AI-assisted .NET feature about what valid object exists, how much latency is allowed, and what gets persisted

Read the full post for the code and the MIT-licensed repository.