nokori Generate

With Generate, use real-time data to create highly accurate AI generated content.

From building Question-Answer services, to completing text generation tasks that require specific data from your internal or external data-services, Generate is a sophisticated, multi-faceted AI generation tool.

Hallucination

A common problem with many of today's most popular Large Language Models (LLMs) is a tendency to "hallucinate" when questions are asked that relate to information it wasn't trained on.

While progress is being made to teach models how to generate code on the fly to go fetch data and then generate answers using that data, that process is long, agonizing, and not very useful in a commercial context.

This is where nokori Generate steps in.

Data-backed Generation

Using nokori Generate, we can easily answer natural language questions using data we already have available in our existing databases, internal documents, or other available data sources.

Airbnb listing example

Consider an example where we have an AirBnb listing description and we want to know how many bedrooms it is:

Example generate from an AirBnb listing

import nokori from '@nokori'

const { data, error } = nokori.ai.generate({
  prompt: "How many bedrooms does this property have?",
  context: [
    "Welcome to our bright and spacious Diamond in the Bluff cabin! Where memories of your vacation will last forever. The breathtaking mountain views offer lush trees in the summer, the changing colors of fall, snowcapped peaks in the winter and the colorful blossoms of spring, definitely a sight to see. Enjoy three king bedrooms and one queen bedroom, four bathrooms, and ample living space for family gatherings. Shops and eateries within walking distance. Our cabin offers flat screen smart TV in each bedroom, washer/drier, blue-ray & DVD player, high-speed Wi-Fi and kid-friendly amenities like baby-gate, pack&play and highchair."
  ]
})

The response:

This property offers three king bedrooms, and additionally one queen bedroom.

Multiple documents example

Sometimes multiple documents or texts are useful when generating responses that are more wholistic in nature. Using nokori Generate, you can pass multiple documents at request time along with your prompt.

Consider getting answers from multiple reviews of a product you sell:

Example generate from multiple product reviews

const { data, error } = nokori.ai.generate({
  prompt: "What do customers like best about this place?",
  context: [
    "10 out of 10 would recommend. Be sure to try to famous smoked cheese dip!",
    "This is what I came for. The atmospher with the music and ecclectic art is to die for. Cheese dip = bomb.com",
    "Servers are nice. That dip is really something special though.",
    "Its not really one unique thing that makes this place so special. Its the culmination of all the little things and that smoky flavor on everything that just really makes this place unforgettable."
  ]
})

The response:

Customers like the famous smoked cheese dip, the atmosphere with music and eclectic art, and the friendly servers. The unique smoky flavor on dishes adds to the overall unforgettable experience.

TL;DR Generation

Want to generate a succint TL;DR for content that isn't meant to be a detailed summary, but just the gist?

Here's a Skift article about the vacation rental brand Vacasa being notified of NASDAQ delisting recently.

Let's ask nokori Generate to create a TL;DR for this:

Create a TL;DR for a travel article

import nokori from '@nokori/js-sdk'

const { data, error } = nokori.ai.generate({
  prompt: "Generate a TL;DR; for this news article",
  context: ["{{The entire news article}}"]
})

The response:

TL;DR generation result

TL;DR: Vacation rental property manager Vacasa received a Nasdaq delisting notice for failing to meet the minimum bid price listing rule. The company has 180 days to regain compliance. Vacasa faced workforce downsizing, increased homeowner churn, and reduced property management forecasts, although Q4 2022 revenue exceeded expectations. Both Vacasa and Sonder, which also received a delisting notice, went public via SPAC mergers.

Conclusion

Data-backed generation is the fastest way yet to add real intelligence to your applications, websites, and/or customer service experiences using data you already have.

Get started with nokori Generate today. Reach out to us at hello[at]nokori.com if you have questions or feedback.