promptfoo/examples/lambdalabs/promptfooconfig.yaml

# yaml-language-server: $schema=https://promptfoo.dev/config-schema.json
description: Lambda Labs

# To use this example:
# 1. Set your LAMBDA_API_KEY environment variable
# 2. Run: promptfoo eval

prompts:
  - |
    You are a creative AI assistant who loves making absurd but somewhat plausible estimations.

    Please answer the following ridiculous question with step-by-step reasoning:
    {{question}}

    Make your answer entertaining but try to use reasonable assumptions and calculations.

    Your response should be at least 100 words and state the final answer at the end.

providers:
  - id: lambdalabs:chat:llama-4-maverick-17b-128e-instruct-fp8
    config:
      temperature: 0.7
      max_tokens: 2048
  - id: lambdalabs:chat:llama3.3-70b-instruct-fp8
    config:
      temperature: 0.7
      max_tokens: 2048

defaultTest:
  options:
    # Use Lambda Labs' Llama 4 model for grading too!
    provider:
      id: lambdalabs:chat:llama-4-maverick-17b-128e-instruct-fp8
      config:
        temperature: 0.2 # Lower temperature for more consistent grading
        max_tokens: 2048
  assert:
    - type: javascript
      value: output.length > 100
    - type: llm-rubric
      value: Does this response attempt to answer the question with step-by-step reasoning and plausible calculations?

tests:
  - vars:
      question: How many people would it take typing simultaneously to recreate Wikipedia from scratch within one hour?
  - vars:
      question: What's the total length of toothpaste squeezed out worldwide each morning? Could it circle Earth?
  - vars:
      question: If everyone jumped at exactly the same time, would global internet latency noticeably spike due to social media posting?
  - vars:
      question: How long would it take to fill the Grand Canyon if everyone in the U.S. poured a cup of coffee into it every day?
  - vars:
      question: If all humans suddenly started walking west at once, how long would it take Earth's rotation to noticeably slow down?
  - vars:
      question: How tall would a pile of all receipts printed globally each year be? Would it reach the moon?
  - vars:
      question: How many miles of spaghetti does humanity eat per second, and how quickly could that spaghetti stretch from New York to Los Angeles?
  - vars:
      question: Could every microwave on Earth simultaneously popping popcorn affect weather patterns?
  - vars:
      question: If you could somehow assemble all lost keys in history into a single pile, how high would the mountain of keys rise?
  - vars:
      question: If all the world's smartphones were stacked on top of each other, how many times could they reach the International Space Station?
  - vars:
      question: How many balloons would it take to lift an average-sized house like in the movie "Up", and how much helium would that require?
  - vars:
      question: If all the data stored in cloud servers was printed on paper, would the resulting paper stack reach Mars?