Can I vibecode Connected Papers?

text
Build the closest honest personal consolation tool inspired by Connected Papers, but do not claim to replace its structural moat or proprietary data graphs. You must use exactly this stack: Next.js 15 (App Router), TypeScript, SQLite (via Prisma or Drizzle), and Playwright for end-to-end testing. Do not offer alternative technology stacks.

The primary objective is to build a private paper graphing and management workspace. The app should allow a user to import user-supplied URLs (e.g., from arXiv or Semantic Scholar open APIs) or upload local PDF files. It should extract basic metadata (title, authors, abstract, citations if available), allow the user to attach personal notes, and support full-text or metadata search across the local corpus.

Start from an empty directory and create the complete, fully functioning project. The application must default to a single-user, private mode. Store all user data locally in the SQLite database and local file system. Do not integrate analytics, telemetry, advertisements, or third-party authentication services. All necessary secrets and external credentials must be securely placed in a `.env` file; provide a comprehensive `.env.example` and ensure credentials are never committed.

Generate realistic, clearly labeled sample data (e.g., public domain papers or dummy metadata) that the user can easily delete. Implement the smallest, most polished user interface possible that completes the core loop (import -> extract -> view graph/list -> search). The UI must include distinct visual states for empty collections, loading/processing indicators, validation warnings, success confirmations, and clear error boundaries. Implement standard import and export features (e.g., CSV or JSON export) so the user is not locked into the platform. Ensure the interface relies on accessible keyboard navigation, semantic HTML labels, visible focus states, and high contrast.

Validate all untrusted input rigorously, and guarantee that no secrets or private file paths are ever logged. Deliberately exclude paid-tier features of the original product, such as team libraries, institutional single sign-on, licensed proprietary scholarly metadata, and publisher-specific import reliability. Do not fake integrations, network effects, or security compliance. If an external API (like Semantic Scholar) is optional, ensure the app remains useful for local files without it, and clearly explain the degraded mode.

Write focused unit tests for the SQLite data model and the metadata extraction workflow. Add one comprehensive Playwright end-to-end smoke test that proves the core loop functions correctly in a browser environment. Supply a detailed README covering setup, required permissions, architectural decisions, data storage locations, backup instructions, and limitations. Include package.json scripts for install, dev, test, build, and start. Finally, run all tests and the build step before concluding, actively fixing any errors that arise rather than just describing how to fix them.