What works without a model
Smart Second Brain is built in three layers, and each one is useful on its own. Nothing below requires an account, an API key, or an internet connection until you reach the agent.
What works without a provider
Section titled “What works without a provider”Search. Full-text, fuzzy, and tag/frontmatter-aware, with no setup.
Smart Graph. An interactive graph view with clustering, layout, and filtering. Out of the box it renders your wikilinks.
Open either from the command palette:
Smart Second Brain: Search notesSmart Second Brain: Open smart graph
Add an embedding model for the semantic layer
Section titled “Add an embedding model for the semantic layer”An embedding model unlocks concept-level retrieval: search matches meaning rather than exact words, and the graph gains semantic similarity edges between notes that aren’t explicitly linked.
This can run fully locally through Ollama, and a strong
starting point is qwen3-embedding:0.6b — multilingual, and small enough to be
comfortable on a laptop. No note content leaves your machine.
Configure it under Settings → Smart Second Brain → Search → Embedding indexes: add an index, pick the embedding model, and it becomes the search index. The graph picks its own index under the Graph tab, so one index can serve both or each can use a different model.
Add an AI provider for the agent
Section titled “Add an AI provider for the agent”A chat model enables the full agent: it searches your vault, reads what’s relevant, and answers with links back to the notes it used. See Connecting a provider.
Indexing
Section titled “Indexing”The first index build runs in the background and its progress appears in the status bar. Large vaults take a few minutes. Indexing is incremental after that, so only changed notes are reprocessed.