Semantic Search

Search a menu by intent, not keywords. The model handles typos, abstract queries, and cross-lingual terms across 100+ languages.

88 msAPI

Try a query

Search query

Search corpus. Paste your own menu to search it, one item per line.

Corpus items20 items / 50 max

Verification is not set up here, so live runs are paused. You can still browse the examples.

3 resultscurated example

88 msAPI

read as something spicy

What just happened

Abstract intent query. No keyword overlap with the corpus.

Spicy Tom Yum Soupcosine 0.213
relevance
0.928
Spicy Tuna Rollcosine 0.154
relevance
0.921
Spicy Beef Tacoscosine 0.254
relevance
0.890

The bar is the reranker relevance score. The faint tick behind it is the raw cosine similarity.

Use this in your app

Pre-filled with your query and corpus. Drop in a key and run it.

curl -X POST https://dish-embed.latimal.com/search \
  -H "X-API-Key: $LATIMAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "query": "something spicy",
  "corpus": [
    "Margherita Pizza",
    "Spicy Beef Tacos",
    "Tonkotsu Ramen",
    "Pad Thai",
    "Falafel Plate",
    "Spicy Tuna Roll",
    "Pho Bo",
    "餃子",
    "Cheeseburger",
    "Gelato",
    "دجاج شاورما",
    "비빔밥",
    "Green Papaya Salad",
    "Spicy Tom Yum Soup",
    "Iced Latte",
    "Fresh Lemonade",
    "Garlic Bread",
    "Tiramisu",
    "Chili Chicken Wings",
    "Mango Smoothie"
  ],
  "top_k": 8
}'

Search your real catalog

This corpus is curated. The honest test is your menu. Get a free key with 5,000 credits, or send your catalog to the founder.

Get a key, 5,000 free credits Talk to the founder