Figma Make Preview: HTTP 520 when calling OpenAI API from server-side backend

Hello,

I am developing an academic wellbeing application using Figma Make Preview with a Next.js/Payload backend.

I have configured an OpenAI API key as a sensitive environment variable and implemented a server-side integration with the OpenAI Responses API.

However, requests to https://api.openai.com/v1/responses return HTTP 520 Server Error.

Diagnostic findings:

• One isolated diagnostic request was performed using synthetic data.

• The request was sent server-side with store: false.

• The response status was HTTP 520.

• No recognizable OpenAI response headers were returned, including x-request-id, openai-processing-ms, and openai-version.

• No valid AI response was received.

• The selected model, gpt-4.1-mini, supports the Responses API and structured outputs.

• No user data was modified during testing.

This may indicate a problem with the Figma Make Preview outbound proxy or network egress layer.

My questions:

1. Are server-side HTTPS requests to api.openai.com supported in Figma Make Preview?

2. Does Preview use an outbound proxy or allowlist that could cause HTTP 520?

3. Are there any known restrictions or issues affecting OpenAI API requests from Preview?

4. Is there a recommended way to integrate an external LLM into a Figma Make application without publishing it?

I would appreciate your guidance.

Thank you!