POST /embed

Generate dense vector embeddings for up to 512 menu items in a single request. Vectors capture semantic meaning, so "Murgh Makhani" and "Butter Chicken" will have high cosine similarity.

Request

Parameter	Type	Required	Description
items	string[]	Yes	Menu item texts to embed. 1-512 items, max 500 chars each.
dimension	int	No	Output vector dimension. Valid: 128, 256, 384. Default: 128.

{
  "items": ["Chicken Biryani", "Paneer Tikka Masala", "Masala Dosa"],
  "dimension": 128
}

Response

Field	Type	Description
embeddings	float[][]	One vector per input item, in submission order
dimension	int	Dimension of each vector
count	int	Number of items embedded

{
  "embeddings": [[0.023, -0.041, ...], [0.018, -0.037, ...], [0.031, -0.028, ...]],
  "dimension": 128,
  "count": 3
}

Example

import requests

response = requests.post("https://embed.statode.com/embed",
    headers={"X-API-Key": "YOUR_KEY", "Content-Type": "application/json"},
    json={
        "items": ["Chicken Biryani", "Paneer Tikka Masala", "Masala Dosa"],
        "dimension": 128
    }
)
data = response.json()
# data["embeddings"][0] is a 128-dimensional vector for "Chicken Biryani"

curl -X POST https://embed.statode.com/embed \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"items": ["Chicken Biryani", "Paneer Tikka Masala"], "dimension": 128}'

Cost

0.05 credits per item.

Notes

• Vectors are L2-normalized. Use cosine similarity (or dot product, equivalent for unit vectors) to compare them.
• Text is automatically cleaned before encoding: promotional noise, prices, and serving sizes are stripped, common misspellings are corrected.
• Use /embed/batch for catalogs larger than 512 items.
• Higher dimensions capture more nuance but use more storage. 128 is sufficient for most search and dedup use cases.