AI Nodes

AI Custom Prompt

An AI-powered node that enables natural language conversations using Azure OpenAI's language models. This node allows you to create interactive chatbots, virtual assistants, or automated response systems by sending customized prompts to Azure OpenAI services.

Configuration

Azure OpenAI Config

A configuration node containing your Azure OpenAI API credentials. This is required and must be set up before using this node.

Select Source

Choose the input source for the message:

• Payload: Uses msg.payload directly
• Dialogue Cloud Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
• Dialogue Cloud Text Transcript: Uses text messages from msg.payload.message

System Prompt

Defines the behavior and context for the AI. This sets the tone, purpose, and constraints for the AI assistant. If left empty, a default prompt will be used: "You are an AI assistant that helps people find information."

Include Diagnostics

When enabled, appends usage (token counts) and latencyMs (response time) keys to the output payload.

Debug Mode

When enabled, displays the complete request payload in the debug sidebar for troubleshooting purposes.

Inputs

payload string | object

The input message to be processed. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• content - The AI-generated response text
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the response was generated
• usage (optional) - Token usage information from the API; included only when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds; included only when Include diagnostics is enabled

Details

This node leverages Azure OpenAI's advanced language models to engage in natural conversations. It processes incoming messages and generates human-like responses that are contextually appropriate and meaningful.

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Error Handling

The node provides different status indicators based on operation:

• Blue dot: Processing request
• Green dot: Successful completion
• Red ring: Error occurred (details in debug panel)
• Yellow ring: Rate limited by Azure OpenAI

Integration with Dialogue Cloud

This node is designed to work seamlessly with Dialogue Cloud communication systems, particularly with transcript sources. It can process both text and transcribed audio inputs, making it ideal for creating intelligent voice and text-based interfaces.

Example Flow

A basic flow might include:

1. An input node (HTTP, WebSocket, etc.) that receives user queries
2. The AI Custom Prompt node configured with appropriate system prompt
3. An output node to deliver the AI response back to the user

AI Action Detector

An AI-powered node that identifies action items, tasks, and commitments from conversations or text content using Azure OpenAI's language models. This node analyzes text input to extract actionable information, making it ideal for project management, task tracking, and follow-up automation.

Overview

The AI Action Detector uses natural language processing to identify and extract structured information about tasks and commitments from unstructured text. It can detect:

• Task descriptions and action items
• Assigned individuals responsible for tasks
• Due dates and deadlines
• Priority levels
• Current status of tasks

Configuration

Basic Settings

• Name: Optional name for the node instance
• Azure OpenAI Config: Connection to your Azure OpenAI service (required)
• Select Source: Choose the input source format:
  • Payload: Direct text input in msg.payload
  • Dialogue Cloud Audio Transcript: Transcribed audio from (msg.payload.transcriptor.transcript)
  • Dialogue Cloud Text Transcript: Text messages from (msg.payload.message)

Detection Options

• Detect Assignees: Identifies who is responsible for each action item
• Detect Deadlines: Extracts due dates and timeframes in ISO 8601 format when possible
• Detect Priority: Determines the urgency/importance level of tasks (high, medium, low)
• Detect State: Determines the current status of tasks (pending, in progress, completed)

Advanced Settings

• Include Diagnostics: Adds usage (token counts) and latencyMs (response time) to the output payload when enabled
• Debug Mode: When enabled, shows detailed API request information in the Dialogue Studio debug panel

Inputs

payload string

Direct text input to analyze (when Payload source is selected)

payload.transcriptor.transcript string

Transcribed audio content (when Dialogue Cloud Audio Transcript source is selected)

payload.message string

Message text content (when Dialogue Cloud Text Transcript source is selected)

Outputs

payload object

A JSON object containing:

• actions: Array of detected action items with their details
• originalMessage: The analyzed message
• timestamp: ISO timestamp of when the analysis was performed
• usage (optional): Token usage information from the Azure OpenAI API, included only when Include diagnostics is enabled
• latencyMs (optional): API round-trip time in milliseconds, included only when Include diagnostics is enabled

Output Action Properties

• text: Description of the actionable task
• confidence: AI confidence score (0-1) indicating certainty of the detection
• assignee: Person assigned to the task (if detected)
• deadline: Due date in ISO 8601 format (if detected)
• priority: Task priority level (high, medium, low)
• state: Current status of the task (pending, in progress, completed)

Example Output

Example below shows output when Include diagnostics is enabled.

{
  "actions": [
    {
      "text": "Review the quarterly report",
      "confidence": 0.95,
      "assignee": "Alice",
      "deadline": "2024-03-15T00:00:00Z",
      "priority": "high",
      "state": "pending"
    },
    {
      "text": "Schedule team meeting",
      "confidence": 0.88,
      "assignee": "John",
      "deadline": null,
      "priority": "medium",
      "state": "in progress"
    }
  ],
  "usage": {
    "prompt_tokens": 145,
    "completion_tokens": 87,
    "total_tokens": 232
  },
  "latencyMs": 845,
  "timestamp": "2024-03-10T14:30:00Z"
}

Usage Examples

Example 1: Basic Task Detection

Input: "John needs to prepare the presentation slides by Friday. Sarah should review the quarterly report before next week's meeting."

Output:

{
  "actions": [
    {
      "text": "Prepare the presentation slides",
      "confidence": 0.92,
      "assignee": "John",
      "deadline": "2024-03-15T00:00:00Z",
      "priority": null,
      "state": "pending"
    },
    {
      "text": "Review the quarterly report",
      "confidence": 0.88,
      "assignee": "Sarah",
      "deadline": "2024-03-18T00:00:00Z",
      "priority": null,
      "state": "pending"
    }
  ]
}

Example 2: Meeting Minutes Analysis

Input: "During the meeting, we agreed that marketing team will finalize the campaign strategy by end of month. Alex volunteered to coordinate with the design team on the new logo - this is high priority. CEO mentioned the budget review is already in progress."

Output:

{
  "actions": [
    {
      "text": "Finalize the campaign strategy",
      "confidence": 0.94,
      "assignee": "marketing team",
      "deadline": "2024-03-31T00:00:00Z",
      "priority": "medium",
      "state": "pending"
    },
    {
      "text": "Coordinate with the design team on the new logo",
      "confidence": 0.96,
      "assignee": "Alex",
      "deadline": null,
      "priority": "high",
      "state": "pending"
    },
    {
      "text": "Budget review",
      "confidence": 0.85,
      "assignee": null,
      "deadline": null,
      "priority": null,
      "state": "in progress"
    }
  ]
}

Usage Tips

1. Confidence Scores: Consider filtering actions with low confidence scores in downstream nodes.
2. Context Awareness: The AI understands contextual information like implicit deadlines, implied priorities, and task states.
3. Integration Ideas:
   • Connect to a database to store extracted tasks
   • Send notifications when high-priority tasks are detected
   • Create calendar events for tasks with deadlines
   • Generate automatic meeting summaries with action items
4. Performance Considerations:
   • Processing longer texts may require more API tokens
   • Consider splitting very long inputs into smaller chunks

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Requirements

• Requires a valid Azure OpenAI API configuration
• Works best with clearly articulated text
• Performs optimally with conversations that have explicit action items

AI Priority Analyzer

An AI-powered node that analyzes incoming messages and assigns priority levels using Azure OpenAI's language models. Perfect for triaging customer inquiries, support tickets, or any text-based communication that requires priority-based routing.

Inputs

payload string | object

The input message to be analyzed. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

The node returns an object containing:

• priority - Numerical priority level (1 being highest priority, up to your configured maximum)
• rationale - Detailed explanation of why this priority level was assigned
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the analysis was performed
• usage (optional) - Token usage information from the Azure OpenAI API when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds when Include diagnostics is enabled
• error - Error message if processing fails

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • Dialogue Cloud Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • Dialogue Cloud Text Transcript: Uses text messages from msg.payload.message
• Priority Levels: Set the number of priority levels (default: 10)
• Low Priority Example: Provide an example of a low priority message to guide the AI
• High Priority Example: Provide an example of a high priority message to guide the AI
• Include Diagnostics: Adds usage and latencyMs fields to the output payload when enabled
• Debug Mode: Enable to get additional debugging information in the Dialogue Studio debug panel

How It Works

This node leverages Azure OpenAI's language models to analyze message content and determine appropriate priority levels. The process works as follows:

1. The node receives a message from the selected source
2. It sends the message content to Azure OpenAI with a specially crafted prompt
3. The AI model analyzes the content based on the provided examples and returns a priority level with justification
4. The node formats the result and forwards it to the next node in your flow

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Priority Scale

The priority scale ranges from 1 (highest priority) to your configured maximum (default: 10). The AI considers:

• Message urgency and criticality
• Customer needs and specific circumstances
• Business impact and time sensitivity
• Required response times

Error Handling

The node indicates its status visually:

• Blue dot: Processing the message
• Green dot: Successfully processed
• Yellow dot: Invalid JSON response
• Red ring: Error during processing

Error details are included in the payload output for troubleshooting.

Tips for Best Results

• Provide clear and distinct examples for high and low priority cases
• Consider using a consistent set of priority levels across your organization
• Use this node early in your flow to enable priority-based routing
• Combine with other nodes to create automated workflows based on priority

AI Response Generator

An AI-powered node that generates contextually appropriate responses using Azure OpenAI's language models. Perfect for customer service, automated replies, and consistent communication.

Inputs

payload string | object

The input message to respond to. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• response - The generated response text
• style - The style used for generation
• wordCount - Number of words in the response
• originalMessage: The analyzed message
• timestamp - Generation timestamp
• usage (optional) - Token usage information from Azure OpenAI when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds when Include diagnostics is enabled

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • Dialogue Cloud Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • Dialogue Cloud Text Transcript: Uses text messages from msg.payload.message
• Response Style: Select the tone and style of generated responses:
  • Professional: Formal, business-appropriate responses
  • Casual: Friendly, conversational tone
  • Technical: Detailed, technical explanations
  • Empathetic: Understanding, supportive responses
• Max Length: Maximum length of generated responses (50-1000 characters)
• Options:
  • Include Greeting: Add appropriate greeting to responses
  • Include Signature: Add signature/closing to responses
• Custom Templates: Define custom response templates in YAML format
• Include Diagnostics: Adds usage and latencyMs metrics to the output payload when enabled
• Debug Mode: Enable to get additional debugging information in the output

Custom Templates Format

Define custom response templates in YAML format:

greeting:
  formal: "Dear {name},"
  casual: "Hi {name}!"
closing:
  formal: "Best regards,"
  casual: "Thanks!"
templates:
  support:
    - "I understand your concern about {issue}..."
    - "Let me help you with {problem}..."

Troubleshooting

• Missing Configuration: Ensure Azure OpenAI Config node is properly configured with valid API key and endpoint
• Invalid Message: Check that the selected source contains valid text data
• API Error: Verify Azure OpenAI service is accessible and responding
• Invalid JSON: Enable Debug Mode to view the raw API response for troubleshooting

Notes

• Max Length parameter is specified in words, not characters
• Response generation may take a few seconds depending on the Azure OpenAI service response time
• For best results, ensure the input message contains clear context for the AI to generate an appropriate response
• The node validates input and configuration before making API calls to prevent errors

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

AI Entity Extractor

An AI-powered node that identifies and extracts various types of entities from text content using Azure OpenAI's language models. Perfect for automated data extraction, form filling, and information gathering.

Inputs

payload string | object

The input text to analyze for entities. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using Dialogue Cloud Audio Transcript source - the transcribed text from audio input.

payload.message string

When using Dialogue Cloud Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• entities - Object containing extracted entities by type with confidence scores
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the analysis was performed
• usage (optional) - Token usage information from the API, available when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds, available when Include diagnostics is enabled

Entity Output Format

The extracted entities are returned in the following format:

{
  "names": [
    {
      "value": "John Smith",
      "confidence": 0.97
    }
  ],
  "emails": [
    {
      "value": "john.smith@example.com",
      "confidence": 0.95,
      "associated_name": "John Smith"
    }
  ],
  "phones": [
    {
      "value": "+1-555-0123",
      "confidence": 0.93,
      "associated_name": "John Smith"
    }
  ],
  "accounts": [
    {
      "value": "ACC123456",
      "confidence": 0.9,
      "associated_name": "John Smith"
    }
  ],
  "dates": [
    {
      "value": "2024-03-20",
      "confidence": 0.96
    }
  ]
}

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Entity Types: Select which types of entities to extract:
  • Names: Extracts personal names
  • Emails: Extracts email addresses with associated names when possible
  • Phones: Extracts phone numbers with associated names when possible
  • Dates: Extracts dates in ISO format (YYYY-MM-DD)
  • Accounts: Extracts account numbers with associated names when possible
  • Products: Extracts product names or identifiers
• Custom Entities: Define additional entity patterns to detect
• Include Diagnostics: Adds usage and latencyMs fields to the output payload when enabled
• Debug Mode: Enable to get additional debugging information in the output

Custom Entities

Define custom entities one per line. You can optionally add a description using a colon or dash. The name (left side) becomes the entity key in the JSON output; the description is used only as a hint to the model.

order_id: internal order identifier (e.g. ABC-123-XYZ)
account_id - customer account number
plan_tier: subscription plan (basic, pro, enterprise)
product_sku: product stock keeping unit
case_id: support case identifier
addresses: physical addresses

Keys are normalized to lowercase snake_case (spaces and hyphens become underscores; other punctuation removed).

Important Notes

• Only entities with confidence scores of 0.8 or higher are included in the output
• If no built-in entity types are selected and no custom entities are defined, the node requests the default set (names, emails, phones, dates, accounts, products)
• The node attempts to associate entities like emails and phone numbers with corresponding names when possible
• Empty arrays are returned for entity types that weren't found in the input text
• The node requires a valid Azure OpenAI service configuration with an API key and endpoint

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Error Handling

The node provides status feedback and error messages for common issues:

• Missing configuration: Azure OpenAI configuration is incomplete or missing
• Invalid message: The source text is missing or invalid
• API error: An error occurred while communicating with the Azure OpenAI service
• Invalid JSON: The API response contained invalid JSON
• No entities found: No valid entities were found in the input text

AI Summarizer

An AI-powered node that generates concise summaries of incoming messages using Azure OpenAI's language models. Perfect for condensing long conversations, documents, or transcripts into brief, informative summaries while preserving key information.

Inputs

payload string | object

The input message to be summarized. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• summary - The condensed version of the input text
• keyPoints - Array of main points extracted from the text
• wordCount - Number of words in the generated summary
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the analysis was performed
• usage (optional) - Token usage information from the API when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds when Include diagnostics is enabled

payload.error string

Error message if the summarization process fails. Includes details about what went wrong.

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Number of Words: Target length for the generated summary (1-100 words)
• Include Diagnostics: Adds usage and latencyMs fields to the payload when enabled
• Debug Mode: Enable to see the full request payload sent to Azure OpenAI in the debug pane

Details

This node leverages Azure OpenAI's advanced language models to create concise summaries while maintaining the core message. The node:

• Processes text input to extract key information
• Generates summaries of specified length
• Identifies and lists main discussion points
• Maintains context and important details

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Status Indicators

• Blue dot: Processing - Request is being sent to Azure OpenAI
• Green dot: Success - Summary generated successfully
• Yellow dot: Warning - JSON parsing issues with the API response
• Red ring: Invalid message - Input validation failed
• Red dot: API error - Failed to communicate with Azure OpenAI

Requirements

• Valid Azure OpenAI service configuration (endpoint and API key)
• Proper input formatting based on the selected source
• Internet connectivity to reach the Azure OpenAI service

AI Sentiment Analyzer

An AI-powered node that analyzes the sentiment and emotional tone of incoming messages using Azure OpenAI's language models. Perfect for understanding customer satisfaction, monitoring social media sentiment, or any scenario requiring emotional analysis of text content.

Inputs

payload string | object

The input message to be analyzed. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• positive - Float value (0-1) indicating positive sentiment strength
• neutral - Float value (0-1) indicating neutral sentiment strength
• negative - Float value (0-1) indicating negative sentiment strength
• rationale - Detailed explanation of the sentiment analysis
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the analysis was performed
• usage (optional) - Token usage information from the Azure OpenAI API when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds when Include diagnostics is enabled
• error - Error message if analysis failed (only present in error cases)
• details - Additional error details (only present in error cases)

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly as text input
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Include Diagnostics: Adds usage and latencyMs metrics to the payload when enabled
• Debug Mode: Enable to get additional debugging information in the Dialogue Studio debug panel

Details

This node leverages Azure OpenAI's advanced language models to analyze the emotional content and sentiment of messages, providing normalized scores across three dimensions:

• All sentiment scores (positive, neutral, negative) are on a 0-1 scale and sum to exactly 1.0
• Each analysis includes a detailed rationale explaining the reasoning behind the assigned scores
• The node handles various error conditions and provides status feedback directly in the Dialogue Studio interface

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Status Indicators

• Blue dot (processing): Analysis request in progress
• Green dot (success): Analysis completed successfully
• Yellow dot: JSON parsing issue or unexpected response format
• Red ring: Configuration error, invalid input, or API failure

Error Handling

The node will attempt to provide meaningful error messages when issues occur:

• Missing OpenAI configuration
• Invalid or empty input messages
• API connection failures
• JSON parsing errors

AI Intent Recognizer

An AI-powered node that identifies the intent behind incoming messages using Azure OpenAI's language models. It analyzes text content and classifies it into predefined intent categories, making it ideal for automated message routing, customer service triage, and conversation flow management.

Inputs

payload string | object

The input message to be analyzed. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• intent - The identified intent category based on configured intents
• rationale - Explanation of why this intent was selected
• confidence - Confidence score (0-1) for the intent classification
• originalMessage: The analyzed message
• timestamp - ISO timestamp of when the analysis was performed
• usage (optional) - Token usage information from the API, included when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds, included when Include diagnostics is enabled

error string

Error message if JSON parsing fails or required fields are missing in the API response.

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Intents: Define your intent categories in YAML format. Each intent should include:
  • intent: The intent category name
  • description: Detailed description of what this intent represents
  • examples: Sample phrases or scenarios that match this intent
• Include Diagnostics: Adds usage and latencyMs fields to the payload when enabled
• Debug Mode: Enable to get additional debugging information in the Dialogue Studio debug panel

YAML Intent Format

- intent: Customer Support
  description: Handles payment processing, order completion, and general customer inquiries
  examples:
    - How do I complete my payment?
    - Where is my order?

- intent: Technical Support
  description: Resolves hardware issues, technical problems, and system troubleshooting
  examples:
    - My device isn't working
    - How do I fix this error?
    

Error Handling

The node handles several error conditions:

• Missing or invalid configuration - Reports error to Dialogue Studio
• Invalid message source - Sets node status to "invalid message"
• API request failures - Sets node status to "API request failed"
• JSON parsing errors - Sets node status to "invalid JSON" and includes error message in output

Details

This node leverages Azure OpenAI's advanced language models to analyze incoming messages and determine the most likely intent based on your configured categories. The node:

• Processes text input against predefined intent categories
• Provides confidence scoring for intent matching
• Includes detailed rationale for intent selection
• Handles various input formats including direct text and transcribed audio
• Returns "Other" as the intent when no match is found

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Technical Notes

• Uses Azure OpenAI with temperature = 0 for consistent results
• Implements a 10-second timeout for API requests
• Normalizes confidence scores to ensure they remain between 0-1
• Includes token usage statistics and latency metrics when Include diagnostics is enabled

Common Use Cases

• Customer service request classification
• Support ticket routing and prioritization
• Chatbot conversation flow management
• User inquiry categorization
• Automated response selection

AI Transcript Generator

An AI-powered node that generates realistic conversation transcripts based on provided input using Azure OpenAI.

Inputs

payload string | object

The input message or topic to generate a transcript from. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

payload object

A JSON object containing:

• transcript - The generated conversation transcript with timestamps and speaker names
• timestamp - ISO timestamp of when the transcript was generated
• usage (optional) - Token usage information from the Azure OpenAI API when Include diagnostics is enabled
• latencyMs (optional) - API round-trip time in milliseconds when Include diagnostics is enabled

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Configuration: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly as input
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Lines: Target number of lines for the generated transcript (1-200, default: 20)
• Include Diagnostics: Adds usage and latencyMs metrics to the payload when enabled
• Debug Mode: When enabled, shows additional debug information in the Dialogue Studio debug panel

Transcript Format

The generated transcript follows this format:

HH:MM:SS Speaker: [Dialogue]

For example:

12:01:05 Alex: Hey, did you finish the report for tomorrow?

12:01:09 Jamie: Almost there. I just need to double-check the charts.

Details

This node leverages Azure OpenAI's language models to generate realistic conversation transcripts. It provides:

• Timestamped dialogue with speaker identification
• Natural conversational flow with appropriate pacing
• Configurable transcript length
• Multiple input source options
• Structured output with API usage tracking

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.

Error Handling

The node provides status indicators and error messages for common issues:

• Missing config: Azure OpenAI configuration is incomplete
• Invalid message: Input message is missing or in incorrect format
• Failed: API request failed - check the debug panel for more information

AI Escalation Detector

Uses Azure OpenAI to analyze conversation content and determine if human agent escalation is needed based on various criteria.

Inputs

payload string | object

The input text to analyze. Can be either plain text or a structured object containing transcript data.

payload.transcriptor.transcript string

When using AnywhereNow Audio Transcript source - the transcribed text from audio input.

payload.message string

When using AnywhereNow Text Transcript source - the message text content.

Outputs

1. Escalate (top): Messages that need human attention
2. Continue (bottom): Messages that can be handled by AI

Output Format

The message payload contains:

• shouldEscalate: Boolean indicating escalation decision
• reason: Brief explanation for the decision
• originalMessage: The analyzed message
• timestamp: When the analysis was performed
• usage (optional): Token usage information from the API, included when Include diagnostics is enabled
• latencyMs (optional): API round-trip time in milliseconds, included when Include diagnostics is enabled

Configuration

• Name: Optional name for the node instance
• Azure OpenAI Config: Required connection configuration for Azure OpenAI service
• Select Source: Choose the input source:
  • Payload: Uses msg.payload directly
  • AnywhereNow Audio Transcript: Uses transcribed audio from msg.payload.transcriptor.transcript
  • AnywhereNow Text Transcript: Uses text messages from msg.payload.message
• Include Diagnostics: Adds usage and latencyMs to the payload when enabled
• Debug Mode: Enable to get additional debugging information in the node output

Escalation Criteria

The node evaluates conversations for the following escalation indicators:

• Explicit requests for human assistance
• Strong negative emotions (frustration, anger, disappointment)
• Sensitive topics (financial, medical, legal matters)
• Complex issues requiring human judgment or investigation
• Multiple failed resolution attempts
• Technical issues that can't be resolved through troubleshooting
• Compliance or policy-related questions
• Account-specific issues requiring verification

Status Indicators

• Blue dot: Currently analyzing message
• Green dot: No escalation needed
• Yellow dot: Escalation recommended
• Red ring: Error occurred

Error Handling

The node may produce errors for:

• Missing or invalid Azure OpenAI configuration
• Invalid message format or empty message
• Failed API requests or timeouts (10 second limit)
• Unexpected response format from Azure OpenAI

Usage Tips

• Use with azure-openai-config node configured with appropriate model
• For best results, provide complete conversation context rather than individual messages
• Consider using a Join node to collect conversation history before analysis
• The node is optimized for customer service scenarios but can be adapted for other use cases
• Enable Debug Mode temporarily to see the complete API payload for troubleshooting

Large inputs are truncated from the oldest content to fit the configured token cap and the model's context window. Enable Include diagnostics to see whether truncation occurred.