Drip AI

The rating came from a vision model asked to fill a fixed JSON template: scores for coordination, style, originality or fit for the occasion, and goal alignment, plus a summary and suggestions. Two prompt variants covered whether the user had named an occasion.

Each result was parsed into a typed struct and stored as JSONB under a version number, so older ratings still rendered correctly after the rubric or the prompt changed.

The cheap checks ran first.

File type, a 10 MB size limit and an NSFW check all short-circuited ahead of the model call and the S3 upload, so nothing was billed or stored for a request that was going to be rejected.

Every account started with a free allowance, counted on the user row. Past that, a Redis key capped ratings at three a day, with its 24-hour window set by the first rating of the day, so it reset predictably instead of sliding. Each stored rating recorded whether it was free or paid.

Purchases ran through StoreKit and RevenueCat. The webhook handler checked whether a transaction had already been processed before applying it, and followed transfers between app user IDs, so a retried event could never credit twice.

Sign in with Apple was done on the server too: the API generated its own Apple client secret from the signing key, exchanged the identity token with Apple, and validated it before issuing its own session.

Ready to upgrade your style? DRIP AI is here to give you quick, no-nonsense advice on what works and what doesn’t. Just upload your fit and let the AI do the heavy lifting, delivering tips that help you sharpen your look for any occasion, whether you’re on vacation or just hanging with friends.

We keep things simple because looking good shouldn’t be complicated. DRIP AI gives you the tools to boost your style and feel great in whatever you wear.