Skip to main content

OpenAIGPT

OpenAI GPT Node Documentation

The OpenAI GPT Node enables you to integrate powerful AI language models from OpenAI (including GPT-4, GPT-3.5, and specialized models) into your TheoBuilder workflows. This node can generate text, answer questions, analyze content, and perform complex reasoning tasks using your business data.

What This Node Does​

The OpenAI GPT Node connects your workflow to OpenAI's advanced language models, allowing you to:

  • Generate human-like text responses based on your prompts
  • Answer questions using your company's knowledge base (RAG - Retrieval Augmented Generation)
  • Analyze and process text, images, and audio files
  • Create intelligent chatbots and virtual assistants
  • Automate content creation and customer support responses

Configuration Parameters​

Authentication Settings​

Provider

  • Field Name: providerId
  • Type: Dropdown menu with options:
    • OpenAI: Connect directly to OpenAI's API service
    • Microsoft Azure OpenAI: Use OpenAI models through Microsoft Azure
  • Default Value: OpenAI
  • Simple Description: Choose whether to use OpenAI directly or through Microsoft Azure
  • When to Change This: Select Azure if your organization uses Microsoft Azure cloud services or requires enterprise-grade security and compliance
  • Business Impact: Azure provides additional security controls and may be required for enterprise deployments

API Token (OpenAI Provider)

  • Field Name: apikey
  • Type: Password text field
  • Default Value: Empty
  • Simple Description: Your OpenAI API key for authentication
  • When to Change This: Enter your unique API key from your OpenAI account
  • Business Impact: Required for the node to function; without it, no AI processing will occur

Resource Endpoint (Azure Provider)

  • Field Name: azureEndpoint
  • Type: Text field
  • Default Value: Empty
  • Expected Format: https://your-resource-name.openai.azure.com/
  • Simple Description: The URL endpoint for your Azure OpenAI resource
  • When to Change This: Enter the specific endpoint provided by your Azure administrator
  • Business Impact: Must match your Azure configuration exactly for successful connections

Azure OpenAI API Key (Azure Provider)

  • Field Name: azureApiKey
  • Type: Password text field
  • Default Value: Empty
  • Simple Description: Your Azure-specific API key for authentication
  • When to Change This: Use the API key provided by your Azure OpenAI service
  • Business Impact: Required for Azure deployments; different from standard OpenAI keys

Azure OpenAI API Version (Azure Provider)

  • Field Name: azureApiVersion
  • Type: Text field
  • Default Value: Empty
  • Expected Format: 2023-12-01-preview (example)
  • Simple Description: The API version for your Azure OpenAI deployment
  • When to Change This: Use the version specified by your Azure administrator
  • Business Impact: Must match your Azure deployment version for compatibility

Azure OpenAI Deployment Name (Azure Provider)

  • Field Name: azureDeploymentName
  • Type: Text field
  • Default Value: Empty
  • Simple Description: The name of your specific model deployment in Azure
  • When to Change This: Enter the deployment name created in your Azure OpenAI service
  • Business Impact: Determines which specific model instance your requests will use

Model Selection​

Model

  • Field Name: model
  • Type: Searchable dropdown menu
  • Default Value: Empty
  • Simple Description: The specific AI model to use for processing
  • Available Options: Automatically populated based on your API access (GPT-4, GPT-3.5-turbo, etc.)
  • When to Change This: Choose based on your needs - GPT-4 for complex reasoning, GPT-3.5 for faster responses
  • Business Impact: Different models have varying costs, speeds, and capabilities

Pass on Previous Messages

  • Field Name: passOnPreviousMessages
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: The AI remembers previous conversation messages for context
  • Off: Each request is processed independently without conversation history
  • When to Change This: Enable for chatbot scenarios where context matters
  • Business Impact: Improves conversation quality but increases token usage and costs

Use Connected BotX Chatbot

  • Field Name: passOnPreviousMessagesUseBotX
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: Automatically uses the built-in chatbot conversation history
  • Off: Requires manual specification of conversation history source
  • When to Change This: Enable when building chatbots with TheoBuilder's chat interface
  • Business Impact: Simplifies chatbot development and ensures proper conversation flow

History Property Source

  • Field Name: passOnPreviousMessagesPropSource
  • Type: Smart text field with variable suggestions
  • Default Value: Empty
  • Simple Description: The data property containing conversation history
  • When to Change This: Specify when using custom conversation history tracking
  • Business Impact: Enables context-aware responses in custom chat implementations

Image URL(s) (Vision Models Only)

  • Field Name: imageUrl
  • Type: Smart text field with variable suggestions
  • Default Value: Empty
  • Simple Description: URLs of images to analyze along with text prompts
  • When to Change This: Provide image URLs when using GPT-4 Vision or similar models
  • Business Impact: Enables AI to analyze and respond to visual content

Request Order (Vision Models Only)

  • Field Name: imageUrlPosition
  • Type: Dropdown menu with options:
    • Appended to the end: Images are processed after the text prompt
  • Default Value: Appended to the end
  • Simple Description: Where images appear in relation to your text prompt
  • When to Change This: Currently only one option available
  • Business Impact: Affects how the AI interprets the relationship between text and images

Reasoning Effort (GPT-5 Models Only)

  • Field Name: reasoningEffort
  • Type: Dropdown menu with options:
    • Low: Faster responses with basic reasoning
    • Medium: Balanced speed and reasoning depth
    • High: Slower responses with deep reasoning
  • Default Value: Medium
  • Simple Description: How much computational effort the AI uses for reasoning
  • When to Change This: Use High for complex problems, Low for simple tasks
  • Business Impact: Higher effort improves accuracy but increases response time and costs

Verbosity Level (GPT-5 Models Only)

  • Field Name: verbosityLevel
  • Type: Dropdown menu with options:
    • Low: Concise, brief responses
    • Medium: Balanced detail level
    • High: Detailed, comprehensive responses
  • Default Value: Medium
  • Simple Description: How detailed and lengthy the AI responses should be
  • When to Change This: Adjust based on whether you need brief summaries or detailed explanations
  • Business Impact: Affects response length, processing time, and token costs

Audio File Path (Whisper Models Only)

  • Field Name: audioFilePath
  • Type: Smart text field with variable suggestions
  • Default Value: Empty
  • Simple Description: Path to audio files for transcription or analysis
  • When to Change This: Provide file paths when using Whisper or other audio-processing models
  • Business Impact: Enables speech-to-text and audio analysis capabilities

RAG Training Settings​

Training Style

  • Field Name: selectedTS
  • Type: Dropdown menu with options:
    • Questions & Answers: Train using Q&A pairs from your data
    • Text Documents: Train using full text documents
  • Default Value: Empty
  • Simple Description: The format of your training data for knowledge base creation
  • When to Change This: Choose based on your data structure - Q&A for structured knowledge, Text Documents for unstructured content
  • Business Impact: Determines how effectively the AI can answer questions about your specific business knowledge

Embedding Model

  • Field Name: embeddingModel
  • Type: Searchable dropdown menu
  • Default Value: Empty
  • Simple Description: The model used to convert your text into searchable vectors
  • When to Change This: Select based on your language and accuracy requirements
  • Business Impact: Affects the quality of knowledge retrieval and response accuracy

Training Mode

  • Field Name: trainingMode
  • Type: Dropdown menu with options:
    • Full Training: Complete training process including data processing and indexing
    • Rebuild Embeddings: Recreate text embeddings without changing the index
    • Rebuild Index Only: Recreate search index without reprocessing embeddings
    • Fetch Data Only: Only retrieve and prepare data without training
  • Default Value: Full Training
  • Simple Description: What part of the training process to execute
  • When to Change This: Use partial modes when updating existing trained models
  • Business Impact: Partial training modes save time and resources when making incremental updates

Use Approximate Similarity Index

  • Field Name: useApproximateSimilarityIndex
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: Uses faster approximate search for large datasets
  • Off: Uses precise but slower search methods
  • When to Change This: Enable for datasets with hundreds of thousands or millions of records
  • Business Impact: Significantly improves response speed for large knowledge bases with minimal accuracy loss

Index Trees

  • Field Name: indexTrees
  • Type: Number input
  • Default Value: 10
  • Valid Range: 1 to 100
  • Simple Description: Number of search trees for approximate indexing
  • When to Change This: Increase for better accuracy, decrease for faster indexing
  • Business Impact: More trees improve search accuracy but increase memory usage

Index Search Nodes

  • Field Name: indexSearchNodes
  • Type: Number input
  • Default Value: -1 (automatic)
  • Simple Description: Number of nodes to search during query processing
  • When to Change This: Set specific values for fine-tuning search performance
  • Business Impact: Higher values improve accuracy but slow down searches

RAG Inference Settings​

Distance Function

  • Field Name: distanceFunction
  • Type: Dropdown menu with options:
    • Cosine: Measures angle between vectors (recommended for most text)
    • Euclidean: Measures straight-line distance between vectors
    • Manhattan: Measures grid-like distance between vectors
    • Chebyshev: Measures maximum distance along any dimension
  • Default Value: Cosine
  • Simple Description: Mathematical method for finding similar content in your knowledge base
  • When to Change This: Cosine works best for most text applications
  • Business Impact: Affects the relevance and accuracy of retrieved information

Minimum Confidence Threshold

  • Field Name: minConfidence
  • Type: Number input
  • Default Value: 0.0
  • Valid Range: 0.0 to 1.0
  • Simple Description: Minimum similarity score required to include content in responses
  • When to Change This: Increase to get only highly relevant results, decrease to include more diverse content
  • Business Impact: Higher thresholds reduce irrelevant responses but may miss useful information

Top N Contexts

  • Field Name: topN
  • Type: Number input
  • Default Value: 0 (automatic)
  • Simple Description: Maximum number of knowledge base excerpts to include in each response
  • When to Change This: Set specific limits to control response length and focus
  • Business Impact: More contexts provide comprehensive answers but increase processing time and costs

System Message​

System Message

  • Field Name: systemMessage
  • Type: Multi-line text area with variable support
  • Default Value: Empty
  • Simple Description: Instructions that define the AI's role, personality, and behavior
  • When to Change This: Customize to make the AI act as a specific type of assistant (customer service, technical expert, etc.)
  • Business Impact: Dramatically affects response quality, tone, and appropriateness for your use case

Prompt Editor​

Prompt

  • Field Name: prompt
  • Type: Multi-line text area with variable support
  • Default Value: Empty
  • Simple Description: The main instruction or question you want the AI to respond to
  • When to Change This: Modify for each specific task or question type
  • Business Impact: The most critical setting - determines what the AI actually does

Try to Prevent Hallucinations

  • Field Name: promptPreventHallucinations
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: Adds instructions to make the AI admit when it doesn't know something
  • Off: AI may generate plausible-sounding but incorrect information
  • When to Change This: Enable for factual applications where accuracy is critical
  • Business Impact: Reduces false information but may result in more "I don't know" responses

When Hallucinating, Answer

  • Field Name: answerWhenHallucinating
  • Type: Text field
  • Default Value: "Sorry, I don't know"
  • Simple Description: What the AI should say when it doesn't have enough information
  • When to Change This: Customize the response to match your brand voice
  • Business Impact: Maintains professional communication even when the AI lacks information

Sources​

Include Sources

  • Field Name: includeSources
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: Responses include references to source documents
  • Off: Only the AI-generated response is provided
  • When to Change This: Enable when users need to verify information or access original documents
  • Business Impact: Improves transparency and trustworthiness but increases response complexity

Sources Column Name

  • Field Name: sourcesColumnName
  • Type: Text field
  • Default Value: sources
  • Simple Description: Name of the data column containing source references
  • When to Change This: Customize to match your data structure naming conventions
  • Business Impact: Ensures proper data organization and accessibility

Top N Sources

  • Field Name: topNSources
  • Type: Number input
  • Default Value: 5
  • Simple Description: Maximum number of source references to include
  • When to Change This: Adjust based on how much source information users need
  • Business Impact: More sources provide better verification but may overwhelm users

Wrap in HTML/Custom Template

  • Field Name: wrapInHtmlCustomTemp
  • Type: Toggle switch (On/Off)
  • Default Value: Off
  • On: Formats source references using custom HTML templates
  • Off: Provides source references in simple text format
  • When to Change This: Enable when you need specific formatting for web display or branded presentation
  • Business Impact: Improves visual presentation and user experience

Rendered Output Column

  • Field Name: wrapperRenderedOutputColumn
  • Type: Text field
  • Default Value: rendered_sources
  • Simple Description: Name of the column containing formatted source references
  • When to Change This: Customize to match your data naming conventions
  • Business Impact: Ensures organized data structure for downstream processing

Output Settings​

Output Option

  • Field Name: outTransformId
  • Type: Dropdown menu with options:
    • Original with appended result column: Keeps all input data and adds AI response
    • Return result column only: Provides only the AI response data
  • Default Value: Original with appended result column
  • Simple Description: How to structure the output data from this node
  • When to Change This: Choose "result only" when you only need the AI response for downstream processing
  • Business Impact: Affects data flow structure and processing efficiency

Result Property Name

  • Field Name: outColumnName
  • Type: Text field
  • Default Value: gpt_response
  • Simple Description: Name for the column containing the AI's main response
  • When to Change This: Use descriptive names that match your data organization system
  • Business Impact: Ensures clear data identification in complex workflows

Full Response Property Name

  • Field Name: fullResponseColumnName
  • Type: Text field
  • Default Value: gpt_full_response
  • Simple Description: Name for the column containing complete API response details
  • When to Change This: Customize for your data naming conventions
  • Business Impact: Provides access to detailed API information for debugging and analysis

Omit Full Response

  • Field Name: fullResponseOmit
  • Type: Toggle switch