.png&w=256&q=75)
🤓 Latest Submissions
Rhythm
Small retail organizations may keep track of transactions, but they lack a straightforward method to detect deviations from the regular behavior of returning customers. Rhythm transforms the transaction CSV into an easily understood retention workflow. It calculates the estimated time between orders by individual, warns whenever the personal period passes, and presents the timeline and the product associated with the warning. Optionally, the retailer can draft a considerate reorder message using Gemma run locally via LM Studio on AMD Radeon RX 9070 XT. The rule-based validator discards all unsupported statements and promotional claims, while Rhythm does not send any messages. The pilot planner provides for random selection of the follow-up and the control group for evaluating actual 30-day repeats, without making claims about a forecasted lift in the campaign. The demo contains a small subset of the actual historical transactions in Online Retail, which is part of the UCI Machine Learning Repository, with anonymized customer IDs. These transactions are historic and do not involve any actual customer's information; the app is just reusing historic data and does not generate transactions at all. Alternatively, retailers can provide their own CSV; analysis takes place in the browser.
18 Oct 2026
