Skip to main content
Version: 2025-02-27

RikAI2

RikAI2 combines large language model (LLM) and extractive AI technologies to understand and contextualize documents. This API allows users to extract and analyze data efficiently with natural language questions.

To gain access to RikAI2, please reach out to the sales team at Lazarus AI.

Benefits of Using RikAI2​

  • No training required.
  • Eliminates the need for custom models or weights for most cases.
  • No token limits.
  • Very low hallucination rate: the frequency at which the model's responses include false or misleading information presented as factual.
  • Includes an image transformer, meaning that the model can physically see the inputted file. This allows the model to detect images, signatures, checkboxes and tables.

Example Questions​

  • What is this document?
  • What is (X)'s name?
    • X being something such as a patient or a prescriber on the form.
  • Who signed this form?
  • When was this form signed?
  • What is the (specific field)?
    • Specific field being a header or a section.

For images, below are some potential use cases:

  • Risk analysis.
  • Note preservation and summarization.
  • Language translation.
  • Image caption generation.

For specific prompting help, refer to our RikAI2 Prompting Guide.

How to Use​

RikAI2 is available for use through endpoints in the Lazarus API framework.

There is a bulk upload endpoint to upload 1 or more files to RikAI2 for asynchronous processing. Currently, this is the only endpoint supported for RikAI2 calls.

Input and Output​

Documents may be uploaded using an inputURL or a base64 encoded string.

The question parameter contains the natural language prompt you would like to submit to RikAI2. This parameter accepts a string.

To receive the model output, use either a webhook (recommended) or an outputURL:

  • The webhook parameter is a URL where status updates and JSON responses will be sent. Click here to learn more about using webhook and webhookSendFull parameters, including an example response.
  • The outputURL receives PUT requests made by the API. Its body is either a byte stream (for a ZIP file, default) or a JSON file (if returnJSON is set to true). ZIP will contain the file you uploaded, a CSV file with a breakdown of the itemization, and both a JSON and TXT file containing the entire JSON response.

Variations​

In addition to the base RikAI2 model for processing documents and asking questions about their content, we have variations of the RikAI2 API available.

  • RikY2 is a powerful LLM with document understanding and sequential prompting capabilities; based on a lightweight, flexible version of RikAI.
  • RikAI2-Extract is one of our custom models that is tailored to be a JSON-format extraction model.