Hacker News Reader

That looks pretty cool. Would you mind explaining a little bit how you did that into more detail? Like, you ask Meilisearch some documents (keyword search) and then ask the LLMs score of those and refine the list a bit?

View on HN

It's a two-step prompt-injection process. In the first step, I wrap the raw typed-in input inside of a dynamically-generated prompt template that takes some context into account (for example, the specifics of the product that the person is currently looking at, or the recent history thereof). The prompt generates a list of Meilisearch keyword search queries. I issue all queries and harvest the results. Scoring and ranking happens in the second (also prompt-injection powered) step.

So the answer to your question is pretty much yes.

Reply on news.ycombinator.com