Expand description
A fielded BM25 index over a single user’s memory.
§Why this rather than a search engine
A personal memory corpus is hundreds to low thousands of short records. At that size an in-process inverted index answers in tens of microseconds, rebuilds from the OKF corpus in milliseconds, and needs no segment management, no writer lifecycle, and no crash-recovery story of its own — the canonical Markdown is the recovery story. The index is a derived, disposable artefact by design (§6.1), and treating it that way removes a whole class of “index disagrees with corpus” failures.
The MemoryIndex surface is intentionally narrow so a segment-based
backend can be substituted later without touching callers.
Structs§
- Memory
Index - An in-process fielded BM25 index.
- Query
- What a search is looking for.
- Search
Hit - A scored match.