When should I use semantic vs keyword search?

Keyword for exact matches ("find every memory with the word PostgreSQL"). Semantic for conceptual matches ("find memories about database choices" — also finds MySQL, MongoDB). Hybrid combines both and is the recommended default.

What can I search across?

Memories, episodes, procedures and conversation history. Each has its own search UI and they share the same engine. The web search tool inside chat also searches the live web (Brave / Tavily).

Are search results scored?

Yes — same five-factor scoring as context assembly: relevance 50%, importance 15%, confidence 15%, recency 10%, frequency 10%. Hybrid search weighs semantic and keyword scores together.

Are searches isolated per environment?

Yes. Search only sees the active environment. Switch environments to search a different brain.

Search

Alma search has three modes: Keyword (exact matching, fast), Semantic (conceptual similarity via vector search, finds memories that match meaning even when words differ) and Hybrid (weighted combination — recommended default). Filter by date, category, environment and more. Search runs across memories, episodes, procedures and conversations.

Three search modes: Keyword (FTS5, exact matching), Semantic (conceptual similarity via vector search), and Hybrid (weighted combination of both). Search across all memories with date and category filters.

See plans