Storage (SQLite)

The Context Graph needs a home. All data is stored in a local, encrypted SQLite database with three key extensions: FTS5 for full-text search, sqlite-vec for vector similarity queries, and WAL mode for crash-safe writes.

Data model

The database contains five core entities:

• ContextNode EPages captured in the graph (URL, title, AI summary, entities, embedding, dwell time, tags)
• ContextEdge EConnections between nodes (type, weight, timestamps)
• Space EContext workspaces (name, color, icon, position)
• Tab EIndividual browser tabs (URL, title, suspension state, position within Space)
• ResearchSession EDetected research clusters (node IDs, status, AI summary)

-- Core table: Context Graph nodes
CREATE TABLE context_node (
    id              TEXT PRIMARY KEY,   -- UUID v7
    url             TEXT NOT NULL,
    title           TEXT NOT NULL,
    domain          TEXT NOT NULL,
    ai_summary      TEXT,
    ai_entities     TEXT,               -- JSON array
    ai_embedding    BLOB,               -- 384-dim float32 vector
    visit_count     INTEGER DEFAULT 1,
    total_dwell_time INTEGER DEFAULT 0,
    first_visited   TEXT NOT NULL,
    last_visited    TEXT NOT NULL,
    user_tags       TEXT,               -- JSON array
    space_id        TEXT REFERENCES space(id),
    is_excluded     INTEGER DEFAULT 0
);

-- Full-text search index
CREATE VIRTUAL TABLE context_node_fts USING fts5(
    title, url, ai_summary, ai_entities, user_notes,
    content=context_node
);