https://promptfoo.dev/ logo
how do I see the entire messages (including the in...
# ❓|questions
a
Hi, I'm using a custom LLM grader prompt and passed the test variables in the prompt. However, the LLM response says it couldn't find the variables. (It says the lack of certain input variables makes it hard to judge the output.) This means there's something wrong with the rubric prompt. I have ran promptfoo with the
--verbose
flag but it truncates the messages in the terminal. Is there a way I can see the final rubric messages so I can debug the issue? This is my dummy config
Copy code
yaml
description: "Evaluating text tone"
prompts:
  - ../../dist/prompts/tones.js
providers:
  - id: azureopenai:chat:gpt-4o
    label: "gpt4o"
    config:
      apiHost: xxxx.openai.azure.com
      temperature: 1
      apiVersion: 2024-02-15-preview
  - id: azureopenai:chat:gpt3-16k
    label: "gpt3-16k"
    config:
      apiHost: xxxx.openai.azure.com
      temperature: 1
      apiVersion: 2024-02-15-preview

defaultTest:
  assert:
    - type: javascript
      value: file://prompt.test.js
    - type: llm-rubric

  options:
    provider:
      text:
        id: azureopenai:chat:gpt-4o
        apiHost: "xxxx.openai.azure.com"
        apiVersion: 2024-02-15-preview

    rubricPrompt:
      - role: system
        content: >-
          You're an output evaluator.
          <rest of the prompt goes here>

          Calculate the score but always pass the test. Output your response in the following JSON format:
          {pass: true, score: number, reason: string}

      - role: user
        content: "Text: {{ vars.text }}\nTone of voice: {{ vars.toneOfVoice }}\n\nAI Response: {{ output }}"

tests:
  - vars:
      text: "Dummy text"
      toneOfVoice: "Informative"
i
try just
{{ text }}
and
{{ toneOfVoice }}
🙂