kramme:search-learnings

You can search past learnings by running a BM25 full-text search query against your local SQLite database to retrieve and rank relevant results by relevance score. This approach quickly surfaces matching entries.

Yes, you can filter full-text search results by applying optional category and project filters alongside the BM25 query. This narrows the retrieved learnings to match specific contextual criteria from the database.

Sanitizing user inputs before constructing a BM25 SQL query involves treating search terms as parameterized inputs to prevent injection. The skill automatically sanitizes inputs before building the SQLite full-text search query.

The best way to retrieve specific past guidance is performing a BM25 full-text search over the persistent learnings database. This ranks entries by relevance to quickly surface the most applicable historical guidance.

Yes, BM25 search works directly with a local SQLite database for quick retrieval of past learnings. It constructs structured SQL queries to return relevant results from the persistent local storage without external dependencies.

You need to use BM25 full-text search when your learnings database grows large and simple keyword matching becomes insufficient. BM25 ranking surfaces the most relevant entries by calculating term frequency and relevance.