VKG API Overview

Performance Optimization

It is important to contruct your VKG with your end-user needs in mind. This involves removing unnecessary data from VKGs to ensure speed and accuracy. Additionally, the length of the inputted document is important. When text data in a document is prepared for furthering processing in a model, it goes through a process called tokenization to break the text into smaller sub-strings (tokens). Specifically, the ideal number of tokens for the VKG is 512 or less (~800 characters), as with more tokens performance degrades.

Input and Output

The supported file formats for uploading or editing nodes are CSVs and JSONs.

The output is always a JSON unless exporting a single VKG, in which case the output is a CSV, or exporting multiple VKGs, where the output is a ZIP.

Standard Error Response

If you encounter an error in an API call, the standard body is going to be the following:

{
  "status": "FAILURE",
  "error": {
    "message": ERROR_MESSAGE
  }
}

The ERROR_MESSAGE should provide some information on how to fix the request and/or denote the underlying cause of error

Some API calls will result in additional fields in the body like failedNodes or warningNodes that is specific to the context of the call, but they will also provide additional information on the errors and/or warnings