The OpenAI node connects to OpenAI’s hosted large language models. It accepts input prompts, processes them using a selected model, and returns generated answers. This node is typically used for tasks such as reasoning, summarization, content generation, and conversational response.
Inputs
- Prompt – Text prompt for the model
- Questions – This port accepts plain text prompts or user-generated questions. These are forwarded to the OpenAI model for processing
- Documents – Document objects for context
- System – System instructions for the model
Outputs
- Text – Generated text output
- Answers – This port outputs the model-generated response based on the input received. The output is a string of generated text that can be passed to other nodes
Configuration
Model Settings
- Model – OpenAI model to use
- Default – “gpt-4o”
- Notes – Available models include gpt-4o, gpt-4, gpt-3.5-turbo
- Use GPT-4o for highest quality and complex reasoning tasks
- Use GPT-3.5-Turbo for faster responses and lower cost
- Consider fine-tuned models for specialized applications
- API Key – OpenAI API key
- Notes – Required for authentication
- Project (Organization)
- Optionally enter your OpenAI project or organization name if applicable. This is used for project-level identification in the API
- Temperature – Creativity/randomness level
- Default – 0.7
- Note – Range: 0.0-1.0
- Max Tokens – Maximum response length
- Default – 1024
- Notes – Limits output size
Advanced Settings
- Top P – Nucleus sampling parameter
- Default – 1.0
- Notes – Controls diversity
- Frequency Penalty – Penalty for token frequency
- Default – 0.0
- Notes – Range: -2.0 to 2.0
- Presence Penalty – Penalty for token presence
- Default – 0.0
- Notes – Range: -2.0 to 2.0
- System Prompt – Default system instructions
- Note – Sets model behavior
- Stop Sequences – Sequences to stop generation
- Default – []
- Notes – Custom stop tokens
- Timeout – API request timeout
- Default – 60
- Note – In seconds
Example Usage
Basic Text Generation
{
"model": "gpt-3.5-turbo",
"apiKey": "your-api-key",
"temperature": 0.7,
"maxTokens": 1024,
"topP": 1.0
}
RAG Implementation with GPT-4
For a Retrieval-Augmented Generation (RAG) implementation using GPT-4:
{
"model": "gpt-4o",
"apiKey": "your-api-key",
"temperature": 0.3,
"maxTokens": 2048,
"systemPrompt": "You are a helpful assistant that answers questions based on the provided documents. Always cite your sources and maintain a professional tone.",
"topP": 0.9,
"frequencyPenalty": 0.2,
"presencePenalty": 0.2,
"timeout": 120
}
Best Practices
Prompt Engineering
- Provide clear, specific instructions in your prompts
- Use system prompts to establish consistent behavior
- Include relevant context for knowledge-intensive tasks
- Structure prompts with clear sections for complex tasks
Performance Optimization
- Adjust temperature based on task requirements (lower for factual responses, higher for creative content)
- Set appropriate max tokens to avoid unnecessary processing
- Use streaming for responsive user interfaces
Troubleshooting
API Problems
- Authentication errors – Verify API key validity
- Rate limit exceeded – Implement request throttling or upgrade API tier
- Timeout errors – Increase timeout setting or reduce prompt/context size
Response Quality Issues
- Irrelevant responses – Refine prompts or adjust system instructions
- Inconsistent outputs – Lower temperature for more deterministic responses
- Truncated responses – Increase max tokens setting
For detailed technical information, refer to:
- OpenAI API Documentation
- OpenAI Models Overview
- OpenAI Connector Source Code /../../aparavi-connectors/connectors/llm_openai/openai.py