On filters: Agreed, I see a few pumpkin seed protein powders tagged as nut flavored that should be unflavored. Will address.
On the “4 servings” issue: I wasn’t handling multi-packs properly, planning on adding that. In addition that products nutrition label did not have serving count, so looks like the llm got confused and took pack count. Im skeptical about inferring serving count from total weight / serving size, as I’ve seen many examples of incorrect total weight information on products.
On 18 vs 16 servings: that’s a LLM misread from the nutrition panel. Thinking reprocessing with larger model may address it.
You can connect to it using any MCP client (such as Claude). Then ask in natural language, e.g. “What’s the cheapest unflavored whey powder right now with 4+ stars?”, and the client will hit the MCP tool with those filters and give you back the results.
$239.95 for 9068g with 16g serving size (60 calories, 10g of protein) is not 4 servings and is not 0.2g protein/$1, 1 cal/$1?
And for another product your site says 18 servings, the listing says 16 servings
I haven’t noticed the 404s, but I’ll check.
On filters: Agreed, I see a few pumpkin seed protein powders tagged as nut flavored that should be unflavored. Will address.
On the “4 servings” issue: I wasn’t handling multi-packs properly, planning on adding that. In addition that products nutrition label did not have serving count, so looks like the llm got confused and took pack count. Im skeptical about inferring serving count from total weight / serving size, as I’ve seen many examples of incorrect total weight information on products.
On 18 vs 16 servings: that’s a LLM misread from the nutrition panel. Thinking reprocessing with larger model may address it.
You can connect to it using any MCP client (such as Claude). Then ask in natural language, e.g. “What’s the cheapest unflavored whey powder right now with 4+ stars?”, and the client will hit the MCP tool with those filters and give you back the results.