using ai-shell against the grok API - "Your script" section is empty often but not always

I often get answers like this:

```
┌ AI Shell
│
◇ Your script:

•
│
◇ Explanation:

The script is a simple Python program that greets a user and performs basic arithmetic.

Steps:
1. Prompt user for their name
2. Print personalized greeting
3. Ask user for two numbers
4. Calculate and display sum of numbers
```

Sometimes the answer does contain a script.

From reading other issues this might be about model quality or model behavior. It would be awesome if ai-shell would work with grok API or any other openai-api-supporting LLM endpoint :)