fix: enforce valid JSON response in interior design assistant

[Anirudh2112]

Description:

Problem:
The interior_design_assistant app expected the model to return a valid JSON output. However, in some cases, the model's output was either invalid JSON or missing expected fields, causing runtime errors and breaking the functionality.

Solution:

Implemented response_format in the turn.create calls for list_items and suggest_alternatives to enforce structured JSON outputs.

Key Changes:
Added JSON schema in response_format for both methods:

• list_items: Ensures output includes description (string) and items (list of strings).
• suggest_alternatives: Ensures output is an array of objects with description (string).

[yanxi0830]

broken import here?

[yanxi0830]

You may have forgot to push changes related to enforce_response_format?

[ashwinb]

This is not what the issue really is referring to. You need to use the response_format argument for chat_completion() to ensure that the underlying LLM is generating the response corresponding to the grammar you provided. Simply doing json.loads() and checking against your desired format is NOT sufficient.

[Anirudh2112]

Thank you for the clarification! I’ll update the implementation accordingly and push the changes soon.

[Anirudh2112]

I’ve pushed the latest changes to implement response_format to enforce structured JSON outputs. This addresses the original issue by ensuring the LLM generates valid JSON responses directly. Let me know if there’s anything else to address!

[Anirudh2112]

Hi @yanxi0830, can you please go through the changes I've made and let me know if everything works as intended? Thank you.

[Anirudh2112]

Hi @ashwinb @yanxi0830, I've addressed all the review comments on my PRs. Just wanted to check if there are any other issues I should address or if there's anything else needed from my side to help move these forward. Thanks!