Semantic search over your memories

The query text travels in the URL: the server embeds it, keeps it in its search log, and it appears in web server access logs. To keep the query off the server, use POST with a vector instead. Sealed memories appear only if their embedding was sent when they were stored; they come back without text and, in this text search, without their blob: use POST, or GET /v1/memory/{id}/blob, to decrypt them on the client. Score = 0.5 × semantic similarity + 0.3 × recency (one-week half-life) + 0.2 × importance/10.