Topic: Applications

Combining semantic vector search with traditional keyword search, because each fails where the other works.

Searching a corpus for relevant material and pasting it into the prompt so the model can answer from it.

Splitting documents into pieces before embedding them — usually the real determinant of retrieval quality.

Showing a model several worked examples in the prompt so it infers the pattern you want.

Reusing the processed form of a repeated prompt prefix so you are not charged full price for it again.

The standing instruction placed before a conversation that sets a model’s role, constraints and format.

A store that indexes embeddings so the nearest ones to a query can be found quickly.