I’m using the
_redirects fie for proxying API requests which return
'content-type': 'text/event-stream; charset=utf-8' responses. Site Deploy is at https://deploy-preview-6--spiritwave-ai.netlify.app/
I’m noticing two things:
- The netlify response is not streaming. It arrives in a single chunk.
- If the response gets too big (takes too long), the netlify response is a 502 bad gateway
Given the nature of proxying to services like OpenAI API, longer streamed responses are becoming more common.
I already moved away from netlify functions because of 10s limit. Can you confim what the limit is on proxying for config based redirects? Given your technical architecture and cost structure, is this something hard or cost prohibitive to support?
My ask is either:
- Remove or change limit so proxying longer running LLM streams doesn’t fail.
- Include documentation that proxy redirects (and edge functions) have limited capability for LLM response streaming