Skip to main content

openai

openai​

Send messages to OpenAI's chat completion API and receive responses.

Input​

input:
  messages:
    - role: "system"
      content: "You are a helpful assistant"
    - role: "user"
      content: "Hello!"
  max_tokens: 150      # optional
  temperature: 0.7     # optional
  api_key: "sk-..."    # optional (uses OPENAI_API_KEY env var)
  • messages (required): Array of chat messages
  • max_tokens (optional): Maximum tokens in response
  • temperature (optional): Response randomness (0.0 to 1.0)
  • api_key (optional): OpenAI API key (uses environment variable if not provided)

Output​

output:
  response: "Hello! How can I help you today?"
  • response: The AI's text response

Message Format​

Each message must have:

  • role: "system", "user", or "assistant"
  • content: Text content of the message

Note: The role field uses camelCase serialization internally, but accepts both formats.

Example​

steps:
  - id: ask_ai
    component: /builtin/openai
    input:
      messages: { $from: { step: prepare_chat }, path: "messages" }
      max_tokens: 200
      temperature: 0.3

Complete AI Workflow Example​

steps:
  - id: create_prompt
    component: /builtin/create_messages
    input:
      system_instructions: "You are a code review assistant. Analyze code for potential issues."
      user_prompt: { $from: { workflow: input }, path: "code_snippet" }

  - id: analyze_code
    component: /builtin/openai
    input:
      messages: { $from: { step: create_prompt }, path: "messages" }
      max_tokens: 500
      temperature: 0.2