Ask a person what they want for lunch and they will not say “FOOD_BEVERAGE, rating above four, price band two”. They will say something light, cheap, open now, and close enough to walk to. Closing that gap is what semantic discovery means for Kadaikodi.
The designed experience does two things at once. It interprets intent — mapping a phrase like “something light and cheap nearby” onto categories, price bands, offering text and distance instead of demanding the shopper speak in filters. And it personalises the ordering, because a shopper who reorders the same idli cart every Tuesday should not have to hunt for it, while a shopper exploring a new neighbourhood should see breadth.
Today’s reality is deliberately unglamorous and stated as such. Discovery works well: free-text search, category chips, rating and price filters, and an open-or-closed badge on every card, with closed merchants genuinely unorderable rather than merely greyed out. The customer dashboard’s “Recommended for you” strip, however, is not a recommendation engine — it is the first few merchants from the nearby list. Kadaikodi would rather ship an honest list than a fake personalisation, and would rather say that here than let a marketing page imply a model that does not exist.
The upgrade path is specific. It needs a searchable representation of the catalog — offering names, descriptions, tags, categories — plus a signal of what this shopper has actually ordered, and a ranked feed computed on the server instead of a slice taken in the browser. Both halves are designed; neither is built. Kadaikodi’s API has no search field, no semantic index, and no personalised feed operation today.
Everything ranked would still respect the rules the marketplace already enforces: a closed merchant cannot be surfaced as orderable, and no ranking may leak signals across workspace boundaries.
Roadmap / target-experience scenario. What ships today is filtered discovery and a plainly-labelled nearby list — pre-launch, with no real shopper data behind any of it.
Ready to make this your story?
