Article Keyword Graph and AI Librarian

What it is

A Next.js research portal: a Chrome extension or web form ingests articles, OpenAI generates summaries and keywords, and the keywords are written to a Neo4j graph so articles connect through shared concepts in an interactive Sigma.js view. A Librarian agent recommends only articles that already exist in the indexed corpus.

The problem

Saved articles pile up unread and unconnected; a keyword graph and a no-hallucination librarian make a personal corpus navigable.

How it works

1. Supabase Postgres stores articles; Neo4j Aura stores keyword nodes linked by HAS_KEYWORD edges
2. Chrome extension ingestion, with AI summary and keyword extraction on write
3. Sigma.js/Graphology graph page with click-to-focus on keywords and articles
4. Librarian spec: reader-only, 3-5 cited recommendations per answer, no ingestion authority, and an explicit statement when the corpus lacks coverage

Stack

Disciplines

For your business

Private knowledge bases that ingest a team's reading and make it navigable through a graph and a grounded AI librarian.