# Four-paper direct-citation demonstration

These four records are synthetic teaching data from SciMaps `TestDataSeeder.seedBibliometricProject`. They are not real publications or research findings. The CSV files describe the selected corpus and its internal citations; they are not an import package for the hosted application.

Reproduce the screenshot in a SciMaps development checkout with the existing seeded data:

1. Open **Seed Bibliometrics Project** in Bibliometrics. Clear any previous selection.
2. Select the four exact titles in `sources.csv` and confirm **4 selected**.
3. Open **Direct Citation Network**. Map by **Papers**, use years **2021–2024**, **Top N nodes = 4**, **Layout = Circular**, and **Node size basis = Uniform**.
4. Keep **Include isolates** and **Show arrows** enabled, **Cluster filter = All**, **Edge scope = All**, and both minimum thresholds at **0**.
5. Choose **Fit graph to view**. Expect four nodes and the three directed links in `citations.csv`. Each arrow runs from the citing paper to the cited paper. References outside the selected corpus are excluded.

The sample identifiers correspond to seeded OpenAlex-like IDs ending `W_E2E_SOURCE_01` through `W_E2E_SOURCE_04`; they are not live provider records. Entering the synthetic sample DOIs in the hosted application's Add papers flow does not reproduce this example.

The repository capture uses the real analysis service and graph renderer, with unexpected external provider calls rejected:

```powershell
$env:SCIMAPS_E2E_REPORT_GENERATION_FIXTURE = 'true'
mvn -q test-compile
mvn -q '-Dexec.cleanupDaemonThreads=false' '-Dexec.classpathScope=test' '-Dexec.mainClass=com.scimaps.e2e.IsolatedE2eLauncher' '-Dexec.args=--config=playwright.guide-capture.config.ts' org.codehaus.mojo:exec-maven-plugin:3.6.4:java
```

Docker, Java/Maven, Node, and the Playwright browser must be available. The launcher creates disposable PostgreSQL and Mailpit containers and uses the existing seeded development account. No research-provider API key is required. The automated capture verifies the exact source titles and directed endpoints before saving the image and evidence.
