docs: rebuild documentation from scratch with clean IA

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/user/overview.md

@@ -1,5 +1,29 @@
-# User Documentation
+---
+title: Overview
+sidebar_position: 1
+---
 
-Use this section if you are deploying and using scholarr.
+# What is Scholarr?
 
-- [Getting Started](./getting-started.md)
+Scholarr is a self-hosted service that tracks academic publications from Google Scholar. It runs as a Docker container with a PostgreSQL database and serves a Vue 3 frontend.
+
+## Key Concepts
+
+- **Scholar** - A tracked Google Scholar profile. Scholars are user-scoped: each user manages their own list.
+- **Publication** - A globally deduplicated academic work. Publications are shared across all users to avoid duplicate storage.
+- **Scholar-Publication Link** - Connects a scholar to a publication for a specific user. Read/unread state, favorites, and visibility live on this link, not on the publication itself.
+- **Run** - A single ingestion cycle that fetches new publications for one or more scholars.
+- **Identifier** - A DOI, arXiv ID, PMID, or other external key attached to a publication. Multiple identifiers can exist per publication.
+
+## How It Works
+
+1. You add a Google Scholar profile (by ID, URL, or name search).
+2. The scheduler periodically scrapes the profile for new publications.
+3. Each publication is fingerprinted and deduplicated against the global store.
+4. External APIs (arXiv, Crossref, OpenAlex) are queried for identifiers.
+5. Unpaywall and arXiv resolve open-access PDF URLs when a DOI is available.
+6. The dashboard shows new, unread, and all publications with filtering and search.
+
+## Safety Model
+
+Scholarr enforces rate limits and cooldowns to prevent IP bans from upstream sources. These are not configurable to zero; they are safety floors. See [Scrape Safety Runbook](../operations/scrape-safety-runbook.md) for details.