Context Graph
Now that you've installed Quira, let's explore the core feature that sets it apart. The Context Graph is Quira's core differentiator. It automatically captures and structures your browsing activity into an interactive knowledge graph, turning ephemeral tab sessions into persistent, queryable context. You can visualize it, query it in plain English, and export it to other tools.
Traditional: Tab -> Close -> Forget
Quira: Tab -> Auto-context node -> Thought-flow mapping -> Resume anytimeHow nodes are created
Any page you visit for 30 seconds or more automatically generates a context node. This threshold filters out accidental clicks and quick bounces, capturing only pages you actually engaged with. Pages on your domain exclusion list (e.g., email, banking) are never captured.
Each node stores rich metadata generated by the local AI:
| Field | Type | Description |
|---|---|---|
title | string | Page title at time of capture |
ai_summary | string | Local AI-generated summary (2-3 sentences) |
ai_entities | string[] | Extracted entities (people, concepts, products, technologies) |
ai_embedding | float[384] | Semantic embedding vector for similarity search |
total_dwell_time | integer | Cumulative time spent on the page (seconds) |
user_tags | string[] | User-assigned tags for manual organization |
Edge types
Edges connect nodes and carry a weight field (0.0-1.0) indicating connection strength. There are three types:
| Edge Type | Source | Description |
|---|---|---|
navigation | Browser event | User navigated from page A to page B (the actual research path) |
ai_inferred | Local AI | Topic or entity similarity detected between nodes |
user_created | Manual action | User explicitly linked two nodes |
Navigation edges are unique to Quira
Unlike traditional knowledge graphs that only capture "relatedness," navigation edges represent your actual thought flow Ethe path of your investigation. This is what makes Research Replay possible.
Three-layer access model
The Context Graph is progressively disclosed through three layers, from ambient awareness to full exploration:
| Layer | Surface | Width | Description |
|---|---|---|---|
| Layer 1 | Ambient Sidebar (KG Pulse) | 48px | Related nodes shown as glowing dots alongside the current page |
| Layer 2 | Contextual Panel | 360px | 2-hop local graph with filtering and navigation. Opened via Cmd+Shift+K |
| Layer 3 | Full Graph Explorer | Full screen | WebGPU-rendered explorer at quira://graph with timeline playback |
Storage and free tier
All Context Graph data is stored locally in an encrypted SQLite database with FTS5 for full-text search and sqlite-vec for vector similarity queries. Data never leaves your device.
The Free plan includes 500 nodes per month. You'll receive progressive notifications at 400, 480, and 500 nodes. After the limit, browsing and manual operations are unrestricted Eonly automatic node generation pauses. The Pro plan offers unlimited nodes. For details on the underlying database, see Storage (SQLite).