Resolve

Resolve answers a question about a piece of text using answers you write. You do not upload a labeled dataset, and Resolve does not use Lake gold. Labeled spreadsheet columns are ML.

Data Intelligence → Resolve (/resolve)

Compared to ML, Forecast, and Loop

	Resolve	ML	Forecast	Loop
Input	A piece of text	Feature columns	A time series	Live feature rows
Output	One choice, or a score	Category or number	Future values	A prediction that can learn
Data	Answers you write	Gold or samples	Gold or a sample	Loop wizard

Full chooser: What is Heimdall?

Two kinds of answer

The kind is fixed when you create the Resolve. Each Resolve asks one question.

Kind	What you define	What you get back
Choice	2–255 named answers, each with a short description	The answer it picked, a confidence, and a probability for every answer
Score	2–10 levels, ordered from low to high	A score on that scale, a confidence, probabilities, and a legend of the levels

Flow

1. Create a Resolve
2. Deploy (API key)
3. Send text (API guide)
4. Monitor on the Resolve API tab

Common mistakes

Mistake	Fix
Looking for a training dataset	Resolve does not train. Write the question and the answers
Expecting Resolve to read Lake gold	Use ML or Forecast for a table. Use Loop when each prediction later gets an outcome
Changing Choice to Score after create	Create a new Resolve. The type stays fixed

Next

• Create a Resolve
• Resolve API integration