1 Comment
Jul 19, 2023Liked by Ian Johnson ✨

Re: pain point 1 – What if you had federated similarity-based cacheing of first-time queries?

If someone else has already queried something and we have the response immediately available in the cache, we return it. This saves you the API usage and improves response time significantly.

Not only that but you could create a similarity index to return prompts based on similar queries. If someone said "Give me a list of 10 colors" vs. "Generate a list of 10 colors", you could just return the same prompt response to that.

Expand full comment