The problem: Most people can't read dog body language well. A wagging tail doesn't always mean happy, tucked ears can signal many things.
My approach: 1. Rule engine scans for keywords (tail, ears, pacing) → instant interpretation 2. AI layer (Pollinations) adds emotional context and actionable tips 3. Breed-aware filtering since a Chihuahua and Great Dane communicate differently 4. History logging so families stay consistent
Technical bits I'm proud of: - Zero-framework vanilla JS that still feels modern - Smart fallback: AI fails → rules work → offline mode with cached audio - CSP bypass using CF Worker for embedding restricted iframes - localStorage-based i18n without bundling MB of JSON
It's been helpful for 620k+ households. All client-side, free, no tracking beyond Clarity analytics.
Try it: https://dogtranslator.org
Happy to discuss the AI prompting strategy or architecture choices!