OpenAI

​

Configuration Options

​

Credentials Setup

1. OpenAI Account: Select or create OpenAI API credentials
   • Requires a valid OpenAI API key
   • Supports organization-specific API keys
   • Credentials are securely encrypted and stored
2. Custom Provider Settings (Optional)
   • Base URL: Override the default OpenAI API endpoint (https://api.openai.com/v1)
   • API Version: Specify API version for Azure OpenAI or other compatible services
   • Supports OpenAI-compatible providers like Azure OpenAI, LocalAI, or custom implementations

​

Task Selection

• Create chat completion: Generate conversational responses using GPT models
• Ask Assistant: Interact with OpenAI Assistants with function calling capabilities
• Create speech: Convert text to speech using OpenAI’s TTS models
• Create transcription: Convert audio files to text using Whisper
• Generate variables: Extract structured data from text using AI

​

Features

​

Create Chat Completion

​

Model Configuration

• Model Selection: Choose from available GPT models (gpt-4o, gpt-4o-mini, gpt-4-turbo, etc.)
• Temperature: Control response randomness (0.0 to 2.0, default: 1.0)
• Settings: Fine-tune model behavior for specific use cases

​

Message Management

• System: Define AI behavior and instructions
• User: Represent user input and queries
• Assistant: Include AI responses for context
• Dialogue: Reference conversation history variables

​

Dialogue Integration

• Dialogue Variable: Select variable containing conversation history
• Starts By: Choose whether conversation begins with user or assistant message

​

Response Mapping

• Message content: Save the AI’s response text
• Total tokens: Save token usage for cost tracking
• Tool results: Save results from any executed functions (if applicable)

​

Ask Assistant

​

Assistant Configuration

• Assistant ID: Select from your OpenAI Assistants
• Thread Management: Automatic conversation threading
  • Thread ID Variable: Maintain conversation context
  • Auto-creation: New threads created automatically if none exists
• Message Input: Define the user’s query or instruction

​

Function Integration

• Function Detection: Automatically fetch available functions from your assistant
• Code Execution: Write JavaScript code that runs server-side
• Variable Access: Functions can read and modify bot variables
• Parameter Handling: Receive structured parameters from the AI
• Return Values: Send function results back to the assistant

// Function: getWeather
// Parameters: { city: string }
if (city === 'Paris') {
  return 'Sunny and warm'
} else {
  // Make API call to weather service
  const response = await fetch(`https://api.weather.com/v1/current?q=${city}`)
  const data = await response.json()
  return `${data.condition}, ${data.temperature}°C`
}

​

Create Speech

​

Configuration Options

• Model Selection: Choose TTS model (tts-1, tts-1-hd for higher quality)
• Voice Selection: Pick from 6 available voices:
  • alloy: Neutral, balanced tone
  • echo: Clear, professional sound
  • fable: Warm, storytelling voice
  • onyx: Deep, authoritative tone
  • nova: Bright, energetic voice
  • shimmer: Soft, gentle sound
• Input Text: The content to convert to speech (supports variables)
• URL Storage: Automatically save generated audio URL to a variable

​

Output Management

• Generated audio files are temporary (7-day expiration)
• Files stored in MP3 format for broad compatibility
• URLs can be used directly in audio bubble blocks
• Download files before expiration for permanent storage

​

Create Transcription

​

Configuration

• Audio URL: Provide URL to audio file (MP3, WAV, M4A, etc.)
• Model: Uses Whisper-1 for accurate speech recognition
• Result Storage: Save transcribed text to a specified variable

​

Supported Formats

• Multiple audio formats supported
• Automatic language detection
• High accuracy for various accents and languages

​

Generate Variables

​

Configuration

• Model Selection: Choose appropriate GPT model for the task
• Extraction Prompt: Define what information to extract
• Variable Mapping: Specify which variables to populate
• Context Input: Reference user messages or other text sources

​

Example Usage

Extract user's information from user's last message: "{{Last message}}"

• Name: User’s name
• Email: Email address
• Phone: Phone number (optional)
• Company: Company name (optional)

• Input: “My name is John Smith and my email is john@company.com”
• Result: Name = “John Smith”, Email = “john@company.com”

​

Advanced Features

​

Function Calling and Tools

​

Tool Definition

// Example: Database query function
async function getUserData(userId) {
  const user = await database.users.findById(userId)
  return {
    name: user.name,
    subscription: user.subscription,
    lastLogin: user.lastLogin,
  }
}

​

Integration Benefits

• Dynamic Data Access: Query databases, APIs, and external services
• Real-time Processing: Execute functions during conversation flow
• Context Enhancement: Enrich AI responses with live data
• Variable Management: Update bot state based on function results

​

Vision Support

​

Supported Models

• gpt-4o and variants
• gpt-4-turbo series
• gpt-4-vision-preview

​

Image Processing

• URL Detection: Automatically identifies image URLs in messages
• Format Requirements: Images must be accessible via direct URLs
• Context Integration: Combines visual and textual information

​

Usage Example

What's in this picture?

https://example.com/image.png

​

Custom Provider Configuration

​

Azure OpenAI Integration

• Base URL: https://your-resource.openai.azure.com/
• API Version: 2023-12-01-preview (or latest)
• Authentication: Use Azure API keys
• Model Names: Use Azure deployment names

​

Compatible Providers

• LocalAI: Self-hosted OpenAI-compatible API
• Ollama: Local model serving with OpenAI API compatibility
• Custom Endpoints: Any OpenAI API-compatible service

​

Message History Management

​

Dialogue Variables

• Structure: JSON array of message objects
• Format: [{"role": "user", "content": "Hello"}, {"role": "assistant", "content": "Hi there!"}]
• Persistence: Automatically maintained between bot sessions
• Flexibility: Support both user-assistant and system message sequences

​

Threading (Assistants)

• Automatic Management: Thread IDs created and stored automatically
• Persistence: Conversations continue across sessions
• Cleanup: Implement thread lifecycle management as needed

​

Best Practices

​

Security Guidelines

​

API Key Management

• Store API keys securely in workspace credentials
• Use organization-specific keys for team access control
• Rotate keys regularly following OpenAI’s security recommendations
• Never expose API keys in client-side code or logs

​

Function Security

• Validate all function inputs before processing
• Implement proper error handling and logging
• Use environment variables for sensitive configuration
• Apply rate limiting to prevent abuse

​

Data Privacy

• Be mindful of sensitive data sent to OpenAI
• Consider data retention policies and compliance requirements
• Use Azure OpenAI for enhanced privacy and compliance needs
• Implement data anonymization where appropriate

​

Performance Optimization

​

Model Selection

• Use gpt-4o-mini for simple tasks to reduce costs
• Reserve gpt-4o for complex reasoning requirements
• Consider model-specific capabilities (vision, function calling)
• Monitor token usage and implement cost controls

​

Message Optimization

• Keep system messages concise but comprehensive
• Use dialogue variables to maintain context efficiently
• Implement message truncation for long conversations
• Cache frequently used responses when appropriate

​

Streaming Configuration

• Enable streaming for better user experience
• Handle streaming gracefully in error scenarios
• Consider network conditions and user device capabilities

​

Testing Strategies

​

Development Testing

• Test with various input types and edge cases
• Validate function calling with different parameters
• Verify error handling and graceful degradation
• Test streaming behavior and interruption handling

​

Production Monitoring

• Monitor API response times and error rates
• Track token usage and associated costs
• Log function execution results for debugging
• Implement alerting for service availability issues

​

A/B Testing

• Test different model configurations
• Compare response quality across model versions
• Evaluate user satisfaction with different approaches
• Measure conversation completion rates

​

Cost Management

​

Token Optimization

• Monitor input and output token consumption
• Implement conversation length limits
• Use cheaper models for appropriate tasks
• Consider caching for repeated queries

​

Usage Patterns

• Set up billing alerts in OpenAI dashboard
• Implement rate limiting per user/session
• Monitor peak usage times and plan accordingly
• Consider implementing usage quotas for users

​

Multiple OpenAI Blocks: Tips and Tricks

​

Streaming Limitations

• Text Concatenation: Streaming disabled when AI responses are prefixed/suffixed with text
• Block Sequencing: All blocks must complete before displaying combined results
• Format Preservation: Text formatting may be affected by surrounding content

​

Optimization Strategies

• Sequential Processing: Plan logical flow between AI blocks
• Variable Management: Use intermediate variables for complex data flows
• User Feedback: Provide loading indicators for multi-step AI processes
• Error Handling: Implement fallbacks when sequential blocks fail

​

Vision Integration Details

• Images must be accessible via direct HTTP/HTTPS URLs
• URLs should be isolated from surrounding text
• Supported formats: PNG, JPEG, GIF, WebP

• Vision processing only works with vision-capable models
• Non-vision models treat image URLs as plain text

Analyze this screenshot for UI/UX issues:

https://example.com/screenshot.png

​

Troubleshooting

​

Configuration Errors

​

”OpenAI block returned error”

• Missing Credentials: Ensure an OpenAI account is selected
• Invalid API Key: Verify API key is valid and active
• Missing Messages: Include at least one user message or dialogue reference
• Model Access: Confirm your account has access to the selected model
• Rate Limits: Check if you’ve exceeded API rate limits

​

”Authentication Failed”

1. Verify API key format and validity
2. Check organization ID if using organizational keys
3. Ensure API key has sufficient permissions
4. Test API key in OpenAI’s API playground

​

Response Issues

​

Empty or No Response

• Quota Exceeded: Add payment method to OpenAI account
• Model Overload: Try different model or retry request
• Content Filtering: Response may have been filtered
• Token Limits: Request may exceed model’s token limit

​

Inconsistent Responses

• Temperature Settings: Lower temperature for more consistent outputs
• System Messages: Refine system prompts for better guidance
• Context Length: Ensure sufficient context for complex tasks
• Model Selection: Consider using more capable models for complex reasoning

​

Function and Assistant Issues

​

Function Calls Not Working

1. Verify function names match exactly between assistant and block
2. Check function code for syntax errors
3. Ensure proper parameter handling in function code
4. Test function logic independently
5. Review assistant configuration in OpenAI dashboard

​

Assistant Thread Problems

• Thread Variable: Ensure thread ID variable is properly configured
• Thread Persistence: Verify thread ID is being saved correctly
• Assistant Access: Confirm assistant ID is valid and accessible
• Function Definitions: Check that assistant has required functions defined

​

Audio and Transcription Issues

​

Speech Generation Problems

• Input Length: Ensure text is within model limits
• Voice Selection: Verify voice parameter is valid
• Model Access: Confirm access to TTS models
• File Storage: Check file upload and storage permissions

​

Transcription Failures

• Audio Format: Ensure audio file is in supported format
• File Size: Check if audio file exceeds size limits
• URL Accessibility: Verify audio URL is publicly accessible
• Audio Quality: Poor quality audio may affect transcription accuracy

​

Performance and Cost Issues

​

Slow Response Times

• Model Selection: Use faster models for simple tasks
• Message Length: Reduce input length when possible
• Streaming: Enable streaming for better user experience
• Regional Endpoints: Use geographically closer endpoints

​

High Token Usage

• Message Management: Implement conversation length limits
• Model Efficiency: Use appropriate models for task complexity
• Prompt Optimization: Refine prompts to be more concise
• Response Mapping: Only save necessary response components

​

Custom Provider Issues

​

Azure OpenAI Configuration

• Endpoint Format: Ensure base URL follows Azure format
• API Version: Use correct API version for your deployment
• Authentication: Use Azure API key, not OpenAI key
• Model Names: Use deployment names, not OpenAI model names

​

Compatible Provider Setup

• API Compatibility: Verify provider implements OpenAI API specification
• Authentication: Check authentication method requirements
• Model Availability: Confirm model names and availability
• Feature Support: Verify support for required features (vision, functions, etc.)

​

Monitoring and Debugging

​

Enable Detailed Logging

1. Check bot execution logs for detailed error messages
2. Monitor API response codes and error details
3. Track token usage patterns and costs
4. Set up alerting for service availability

​

Testing Strategies

1. Isolation Testing: Test individual components separately
2. Edge Case Testing: Test with various input types and sizes
3. Load Testing: Verify performance under expected usage
4. Error Simulation: Test error handling and recovery

​

Getting Help

• OpenAI Status: Check OpenAI Status Page for service issues
• API Documentation: Reference OpenAI API docs for latest information
• Community Support: Engage with QuickBot community for implementation help
• Professional Support: Contact support for enterprise-level assistance