I just published my second open-source project: typesafe-ai-rails, a community Ruby on Rails integration for TypeSafe AI’s System One API.

I just published my second open-source project: typesafe-ai-rails, a community Ruby on Rails integration for TypeSafe AI’s System One API. 😃 System One now has a proper Rails wrapper. The gem adds: • Rails-native configuration • helpers for Noul, Choice and Score questions • persistence-backed confidence policies • fail-closed handling for missing or low-confidence decisions • usage, latency, request-ID and cost telemetry • model-aware pricing • direct access to the underlying TypeSafe Ruby SDK when the wrapper shouldn’t get in the way One subtle part was getting the confidence semantics right: Noul returns a probability, while Choice and Score expose confidence separately. The gem deliberately keeps that distinction instead of pretending all three primitives behave the same way. It’s tested against the real TypeSafe API, including live Noul / Choice / Score responses, and published through RubyGems Trusted Publishing.