AITriggerNodeEditor Documentation

Overview

The AITriggerNodeEditor is a feature designed for the Vantage analytics and data platform. This component allows users to configure an AI trigger within their data workflows. The primary purpose of the AITriggerNodeEditor is to evaluate conditions based on data inputs and determine whether an action should be triggered, making it a crucial element for automating responses based on data changes or events.

Purpose

• Evaluate incoming data against set criteria to trigger specific actions.
• Facilitate AI-driven decision-making within data workflows.
• Allow users to customize the evaluation criteria and response thresholds.

Functionality

Key Features

• Evaluation Modes: Users can choose between aggregate evaluations (considering all data at once) or per-row evaluations (assessing each row individually).
• Custom Prompts: Define criteria for AI decision-making with customizable trigger prompts.
• Context Templates: Enable users to format contextual data that is sent to the AI for better reasoning.
• Dynamic Confidence Threshold: Allows users to set how strict or lenient the evaluation should be.
• Automatic Scheduling: Users can set up periodic checks for data to decide if triggers should activate.

Settings

Evaluation Mode

1. Name: Evaluation Mode
2. Input Type: Dropdown selector (Button Group)
3. Explanation: Determines how the AI will evaluate the input data. There are two modes:
   • Aggregate: All data is processed together, resulting in a single trigger decision.
   • Per Row: Each row is evaluated separately, allowing only the rows that meet criteria to pass through.
4. Default Value: "aggregate"

Trigger Prompt

1. Name: Trigger Prompt
2. Input Type: Textarea
3. Explanation: A customizable message that guides the AI on what elements to assess to decide if an action should be triggered. It can include specific criteria related to data changes or anomalies.
4. Default Value: Empty string

Context Template

1. Name: Context Template
2. Input Type: Textarea
3. Explanation: An optional template used to structure data sent to AI, making it easier for the AI to understand the context. Users can utilize placeholders for columns.
4. Default Value: Empty string

Confidence Threshold

1. Name: Confidence Threshold
2. Input Type: Numeric (Range Slider)
3. Explanation: Specifies the minimum confidence level for the AI trigger to activate. Adjusting this affects the sensitivity of the trigger; a lower threshold may trigger more often, while a higher one results in stricter evaluation.
4. Default Value: 70%

Max Rows to Evaluate

1. Name: Max Rows to Evaluate
2. Input Type: Numeric (Input Field)
3. Explanation: When in "per_row" evaluation mode, this limits the number of rows analyzed by the AI to manage costs associated with AI calls.
4. Default Value: 50

Check Schedule

1. Name: Schedule Enabled
2. Input Type: Boolean (Toggle Switch)
3. Explanation: Determines if the AI should automatically check the data at scheduled intervals. Enabling this allows users to set specific check types and intervals.
4. Default Value: False (not enabled)

Check Schedule Type

1. Name: Check Schedule Type
2. Input Type: Dropdown selector (Button Group)
3. Explanation: The frequency at which data checks should occur. Options include:
   • Interval
   • Daily
   • Weekly
   • Monthly
4. Default Value: "interval"

Check Interval Minutes

1. Name: Check Interval Minutes
2. Input Type: Numeric
3. Explanation: Specifies the time duration between checks when the "Interval" schedule type is selected. This allows for automated data assessments at user-defined intervals.
4. Default Value: 60 minutes

Time of Day

1. Name: Time of Day
2. Input Type: Time Input
3. Explanation: If using daily, weekly, or monthly schedules, this defines the specific time when the checks should occur.
4. Default Value: "09:00" (9 AM)

Day of Week

1. Name: Day of Week
2. Input Type: Options for weekly scheduling
3. Explanation: Specifies which day of the week the checks should take place when the 'Weekly' schedule type is selected.
4. Default Value: N/A (must be selected during configuration)

Use Cases & Examples

Use Cases

1. Anomaly Detection in Sales Data: A sales team can set up triggers to identify unusual spikes or drops in sales figures, allowing them to react quickly to market changes.

2. Quality Control in Manufacturing: In a manufacturing process, triggers can be defined to evaluate sensor data for quality control, ensuring that defective items are identified and removed quickly.

3. Customer Sentiment Analysis: A service department can utilize the AITriggerNodeEditor to monitor customer feedback and trigger alerts or actions based on negative sentiments detected in survey data.

Example Configuration

Use Case: Automatic Alert for Sales Anomalies

To address potential anomalies in sales, configure the AITriggerNodeEditor as follows:

• Evaluation Mode: Per Row
• Trigger Prompt: "Should this sale be flagged as anomalous based on previous sales data?"
• Context Template: "The sale {{sale_amount}} at {{sale_date}} deviates significantly from the average sale of {{average_sale}}."
• Confidence Threshold: 85%
• Max Rows to Evaluate: 100
• Check Schedule: Enabled
  • Check Schedule Type: Interval
  • Check Interval Minutes: 30

This configuration will ensure that every 30 minutes, the AI evaluates up to 100 recent sales records to determine if any individual sale should trigger an alert for further investigation based on strict criteria established by the user.