RecSwitch

RecSwitch is a web app + SDK that lets product teams A/B test and swap recommendation engines (vendor APIs, open-source models, or in-house services) behind a single interface. Instead of hardwiring one recommender into your app, you route traffic through RecSwitch, define experiments, and compare business metrics (CTR, conversion, revenue per session) and quality metrics (diversity, novelty, repeat rate). It includes a lightweight event pipeline to ingest clicks/purchases, a feature store-lite for user/item attributes, and a rules fallback when models fail. This is an AI app + traditional app combination: the core value is orchestration, evaluation, and safe deployment, not inventing a new algorithm. Brutal truth: most teams don’t need a “better model”; they need faster iteration, fewer integration headaches, and reliable measurement to justify recommender changes.