temperature_test

OpenAI Temperature and MaxTokens Test

This example demonstrates the Temperature and MaxTokens options in the OpenAI ChatAgent API.

What it tests:

1. High Temperature (0.9): More creative, varied responses
2. Low Temperature (0.1): More deterministic, precise responses
3. MaxTokens Limit: Controlling response length
4. Combined Options: Using multiple options together

Run the test:

export OPENAI_API_KEY="your-key"
go run main.go

Expected behavior:

• High temperature: More creative and varied sentence about robots
• Low temperature: Consistent, precise answer to "What is 2+2?"
• Token limit: Short response that cuts off around 20 tokens
• Combined: Weather analysis with specific persona, temperature, and length

API Consistency

The OpenAI ChatAgent uses the same API as Anthropic:

response, err := agent.Chat("Hello", &agents.ChatOptions{
    Temperature: 0.8,
    MaxTokens:   100,
})

Usage Statistics

OpenAI provides different token usage fields:

• response.Raw.Usage.PromptTokens - Input tokens
• response.Raw.Usage.CompletionTokens - Output tokens

Key learnings:

• Temperature affects creativity vs determinism (same as Anthropic)
• MaxTokens controls response length (hard cutoff)
• Options work together seamlessly
• API is consistent between OpenAI and Anthropic providers

Interested in AI-powered workflow automation for your company? Get started: https://aktagon.com | contact@aktagon.com