I live in Fairbanks, AK and grocery prices up here are high! I built this website, called cache after an Alaskan food cache, to match recipes with weekly grocery store sales.
I'm using the kroger API and Flipp flyer data. I parse out the biggest discounts and cheapest per unit items. I then filter my recipe database based on what's in my pantry and the best deals for the week and send that to anthropic to create a meal plan for the week. Recipe list is hand curated, AI was just making horrible recipes up with items not on my list.
Two things I'd like input on:
Kroger API caching. I'm querying live per zip code and will hit limits if this grows. Caching weekly per store in Firestore is the obvious move, but I've seen in-week price changes, so cached users would occasionally see stale prices. How have people handled this for retail data?
Ingredient normalization was the hard part — "garlic cloves" vs "Garlic Bulb 3ct," "yellow onion" vs "red onion" matching an onion sale but staying distinct in the recipe. I landed on a hint + variant schema with subset matching for families. Curious if anyone's done this better.
Site: cache.fit
Any thoughts and suggestions would be appreciated.
I live in Fairbanks, AK and grocery prices up here are high! I built this website, called cache after an Alaskan food cache, to match recipes with weekly grocery store sales.
I'm using the kroger API and Flipp flyer data. I parse out the biggest discounts and cheapest per unit items. I then filter my recipe database based on what's in my pantry and the best deals for the week and send that to anthropic to create a meal plan for the week. Recipe list is hand curated, AI was just making horrible recipes up with items not on my list.
Two things I'd like input on:
Kroger API caching. I'm querying live per zip code and will hit limits if this grows. Caching weekly per store in Firestore is the obvious move, but I've seen in-week price changes, so cached users would occasionally see stale prices. How have people handled this for retail data?
Ingredient normalization was the hard part — "garlic cloves" vs "Garlic Bulb 3ct," "yellow onion" vs "red onion" matching an onion sale but staying distinct in the recipe. I landed on a hint + variant schema with subset matching for families. Curious if anyone's done this better.
Site: cache.fit
Any thoughts and suggestions would be appreciated.