Skip to main content

OpenAI Workflows

OpenAI Component Example

This example demonstrates how to use the OpenAI built-in component with Stepflow.

Prerequisites​

  1. An OpenAI API key
  2. The Stepflow CLI built from this repository

Setup​

Set your OpenAI API key as an environment variable:

export OPENAI_API_KEY=your_api_key_here

Running the Example​

From the repository root, run:

cargo run --manifest-path stepflow-rs/Cargo.toml -- run --flow=examples/openai/openai_chat.yaml --input=examples/openai/input.json --config=examples/openai/stepflow-config.yml

Example Flow​

The example flow (openai_chat.yaml) demonstrates:

  1. Calling the OpenAI chat completion API
  2. Passing system and user messages
  3. Extracting the response from the API

Additional Configuration​

The OpenAI component supports several configuration options:

  • messages: An array of messages to send to the API (required)
  • max_tokens: Maximum number of tokens to generate (optional)
  • temperature: Controls randomness of responses (optional, between 0 and 2)
  • api_key: Override the API key from environment (optional)

Available Components​

  • builtins:/openai/chat - Uses the gpt-3.5-turbo model
  • builtins:/openai/chat/gpt4 - Uses the gpt-4 model

Example Input​

{
  "prompt": "Explain quantum computing in simple terms.",
  "system_message": "You are a helpful expert explaining complex topics to beginners."
}

Example Output​

{
  "response": "Quantum computing is like...",
  "message": {
    "role": "assistant",
    "content": "Quantum computing is like..."
  }
}