This semantic search service is powered by a vector search enabled by large language models (LLMs). We have used an LLM embedding model to convert descriptions of each of the GLORIA sectors into vectors, or points, in a high dimensional space. When you enter a query the service converts it to vector then uses a cosine similarity metric to find the GLORIA sector nearest to the query vector.

You can visualize this process with a simple example. Imagine the words,dog, cat, and helicopter as points in a 3-D space. Because of their meaning and use, the words dog and cat will be closer together while helicopter will be further away. If you add the word puppy to the space, puppy will be located closer to dog than cat. Here we use the same idea to find the GLORIA sectors nearest in semantic similiarty to your query. This approach is called semantic search because it uses the meaning of the words, not keywords or string matching, to find the best match.

Using IMPORTHTML:

• url - The URL of the page to examine. See the documentation below on the REST endpoint for details about the URL syntax for this service.
• query - Either "list" or "table" depending on what type of structure contains the desired data.
• index - The index, starting at 1, which identifies which table should be read. See the REST endpoint description below for documentation on the tables available.

Tip: When reading tables larger than one cell, leave sufficient blank cells around the IMPORTHTML() cell.

Tip: Always use the ENCODEURL function to encode queries.

Example:

Base URL: https://little-base-bcc4.pages.dev/search

Query parameters:

• q - a URL-encoded string containing the the product or sector, with optional country, to search for. Examples: bananas or bananas%20from%20ecduador. In Google Sheets the ENCODEURL()
• output - Either json or html (default).

The endpoint will return HTML or JSON results depending on the output parameter. Syntax for both styles of results follows.

output=html Results

1. A single cell with the sector name
2. a 1×5 table with the five best matching sectors

output=json Results

            
interface GLORIA3Result {
    id: number, /* The sector ID, as a 0-based index into GLORIA */
    sector: string, /* The GLORIA sector name for this intensity coefficient. */
}