Capture & Library
Capture links, notes, quotes, and files without setup, then browse every saved knowledge type in one filtered Library.
Capture is available from every authenticated screen. Library is where captured and derived knowledge becomes browsable without forcing everything into folders first.
Capture Without Prerequisites
Choose Capture, then select Link, Note, Quote, or File. A capture does not require a tag, folder, stage, classification, embedding, or model provider.
- Link saves a page through the normal bookmark pipeline.
- Note creates a standalone note.
- Quote preserves selected evidence with its source URL and an optional interpretation.
- File imports supported material as a searchable note.
The browser extension, PWA share target, CLI, and local automation routes use the same owned knowledge store. Link and quote captures can queue in the browser while offline.
Bookmark readers retain capture-attempt history, source evidence and available preserved copies, locally hosted article images, and your scroll progress. Direct HTTP capture remains available on every installation. On fresh supported Linux installs, the installer enables an isolated native capture runtime by default; choose the core-only option when you do not want it. See Preservation and Sharing.
Browse the Unified Library
Library shows bookmarks, ordinary notes, and annotations by default. Generated entities and concepts remain available through the separate Derived view and in Graph. Search, type, and sort stay visible. Open More filters for topic, source, date, connection state, and density. Collection controls stay out of the way until you create or select a collection. Results use cursor pagination so large collections remain practical.
From Library, select up to eight eligible sources for a conversation or quiz. Arivu opens a fixed passage preview before any content can be sent to a configured provider.
Required bookmark, collection, tag, annotation, capture-attempt, and processing-job records commit together. A failed initial save cannot leave a bookmark that looks accepted but has no durable processing work behind it.
What Arivu Derives
Arivu can locally extract text, concepts, entities, and deterministic summaries. Generated knowledge is grounded in selected text or source evidence. When evidence is incomplete, Arivu may intentionally leave a generated field empty rather than fill the gap. A complete rendered capture can improve weak direct extraction, but a challenge page or failed attempt cannot replace stronger evidence or the last good reader.
Optional providers can add richer summaries and embeddings. Safe parts of a structured summary can remain available when another field fails validation, and provider outages can fall back to publisher descriptions and evidence-derived material. Missing embeddings only remove semantic-similarity relationships; saved content, text search, explicit links, and other eligible graph relationships continue.