I should write a blog post, too since I recently had similar issues piping an AWS Bedrock response stream from an LLM straight through to the browser using SSE. The problem with SSE and large prompts is that any data needs to be passed in through search parameters IN the URL.

Having experienced to be "hugged to death" by external services in the past, I can feel the intimidation of the creator. However, I would not completely agree with putting all the blame on OpenAI here, its not that simple. I think part of the problem is how the database site has been built.

Oh dear, did you loosen it when helping that lady? 🙈