Skip to content

Documentation for Query Parsing API Methods

/categories/parse_categories

Functionality

The parse_categories endpoint parses a natural-language query and returns a list of similar categories based on vector similarity matching.

Request Parameters

  • search_query (Any): The search query text or structure to be parsed and matched against known categories.

Request JSON Example

{
  "search_query": "deep learning optimization"
}
  • search_query: Can be a plain text string or structured object; it's transformed into an embedding and matched against category vectors.

Response JSON Example

{
  "categories": [
    {
      "object_id": "cat-ml-001",
      "distance": 0.087,
      "payload": {
        "name": "Machine Learning",
        "tags": ["AI", "Modeling"]
      },
      "meta": {
        "source_table": "categories_dataset",
        "row_pointer": 42
      }
    }
  ]
}
  • object_id: ID of the matched category.
  • distance: Cosine or L2 similarity score between query and category.
  • payload: Structured metadata describing the matched category (e.g., name, tags).
  • meta: Storage metadata indicating where the original category data resides (e.g., table name, row reference).

Usage

  • Purpose: To semantically parse user queries and recommend relevant category labels from the database using vector-based matching.

Example cURL

curl -X POST "http://<server>/categories" \
     -H "Content-Type: application/json" \
     -d '{
           "search_query": "example search text"
         }'