> ## Documentation Index
> Fetch the complete documentation index at: https://docs.instaforge.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Playground

> Chat with the assistant like a customer, and see why it answered the way it did.

The playground is a private chat with the assistant. Nothing is sent to Instagram, and it always tests the **draft** version, so it's safe to try changes before customers see them.

<Frame caption="The playground: start over (1) and the explanation panel (2)">
  <img src="https://mintcdn.com/instaforge/HVDnlmyBMfIyfa_y/images/playground.png?fit=max&auto=format&n=HVDnlmyBMfIyfa_y&q=85&s=910692402499f649653f6eaa277a0b63" alt="The playground: start over (1) and the explanation panel (2)" width="1950" height="1353" data-path="images/playground.png" />
</Frame>

## How to test

<Steps>
  <Step title="Choose what to simulate">
    A direct message, a comment, a story reply or a mention. Comments show the public reply and the DM separately.
  </Step>

  <Step title="Write like a customer">
    Short, informal, with typos. Ask in each language the business's customers use.
  </Step>

  <Step title="Read the explanation (2)">
    **Why this answer** shows which rule matched, which knowledge was used, what the assistant remembers about "you" and the exact prompt sent to the AI.
  </Step>

  <Step title="Start over (1)">
    Clears the test chat and its memory, so you can test the first-contact experience again.
  </Step>
</Steps>

## A 10-minute test script

1. "Hi" (welcome message only)
2. "How much is a haircut?" (should be the FAQ answer, word for word)
3. "Do you do beard trims?" (should say no, not invent)
4. "Can I come Saturday afternoon?" (should offer real free times)
5. "¿Cuánto cuesta el color?" (should answer in Spanish)
6. "Can I talk to a person?" (hand-off message, then silence)

<Info>
  When an answer is wrong in a real conversation, use **Correct** next to it in the inbox. The right answer becomes an FAQ straight away.
</Info>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.