promptfoo/examples/llama-cpp/promptfooconfig.yaml

# yaml-language-server: $schema=https://promptfoo.dev/config-schema.json
providers:
  - id: llama
    label: llama3-8b
    config:
      baseUrl: http://localhost:8080

prompts:
  - 'Write a funny tweet about {{topic}}'
  - "Human: Write a funny tweet about {{topic}}\nAI:"
  - "Q: Create a humorous tweet about {{topic}}\nA:"
  - '[INST] Write a funny tweet about {{topic}} [/INST]'
  - "<human>: Craft a witty tweet about {{topic}}\n<assistant>:"
  - "### Instruction: Compose a clever tweet on the topic of {{topic}}\n### Response:"
  - "System: You are a witty AI.\nUser: Tweet about {{topic}}\nAssistant:"
  - "<s>[INST] <<SYS>>\nYou're a comedy writer.\n<</SYS>>\nWrite a funny tweet about {{topic}} [/INST]"
  - "{'role': 'system', 'content': 'You specialize in viral tweets.'}\n{'role': 'user', 'content': 'Write a hilarious tweet about {{topic}}'}\n{'role': 'assistant', 'content':"
  - "# Context: You are a social media expert known for viral content.\n# Human: Create a funny tweet about {{topic}}\n# AI:"

tests:
  - vars:
      topic: 'San Francisco'
    assert:
      - type: icontains
        value: 'San Francisco'
      - type: answer-relevance