AI Insights
FeatureUnderstand the summaries, tags, sections, and collection assignment generated for saved links and PDFs.
The Extraction Pipeline
When a URL or PDF is ingested, Relink extracts the full usable text, keeps more structure for longer documents, and sends that content through the LLM pipeline to generate structured metadata. Long documents are summarized differently from short ones so they follow the source more closely and retain more notable details instead of collapsing into a generic overview.
Generated Metadata
TL;DR
A concise top-level summary displayed on the main card. For long sources, it is backed by denser structured sections in the drawer.
Structured Sections
The drawer can render AI-generated sections such as key takeaways, quotes, concepts, steps, implications, and short code snippets when the source includes notable implementation details. These replace the older flat-only insight format whenever richer structure is available.
Highlights
Important quoted or emphasized passages that help you spot memorable details quickly inside the drawer.
Tags
Broad topical labels used for filtering, tag clusters, and related-memory explanations. In the UI they are normalized to readable labels such as Software Development.
Suggested Queries
Potential questions the user might ask about this text in the future. These are generated with a lightweight prompt-focused model and form the basis for quick-start prompts in the Chat RAG experience.
Collection Assignment
If you do not choose a collection manually, AI can assign the item into a single collection folder. This classification step is intentionally separate from tags: collections are the durable folder tree, while tags remain lightweight topical metadata.