azizul-islam-efath/ai-paper-summarizer

Purpose and Entry Points

Paper Generator is a full-stack research workspace: authenticated users organize projects, upload academic PDFs, obtain AI-derived structured summaries and cross-paper analysis, then generate formatted paper drafts. The client starts at frontend/src/main.jsx and routes through frontend/src/App.jsx; the API process starts at backend/src/server.js.

Client Application

React 19/Vite provides public authentication pages and protected research-workspace views. Layouts and route guards separate anonymous and authenticated flows; page modules cover upload, project/paper detail, summaries, and generated drafts.

• frontend/src/routes/PrivateRoute.jsx

• frontend/src/layouts/DashboardLayout.jsx

• frontend/src/pages/ProjectDetail.jsx

The browser calls backend endpoints through domain-specific service modules, while authentication/session state is held in React context.

• frontend/src/services/api.js

• frontend/src/services/paperService.js

• frontend/src/context/AuthContext.jsx

HTTP API and Control Boundary

Express routes map resource domains—auth, projects, papers, summaries, generated papers, and administration—to controllers. Middleware establishes the API boundary: JWT authentication, request validation/rate limits, multipart upload handling, and centralized failures.

• backend/src/routes/paperRoutes.js

• backend/src/routes/summaryRoutes.js

• backend/src/controllers/projectController.js

• backend/src/middleware/auth.js

• backend/src/middleware/upload.js

Persistent Research Model

Prisma is the persistence boundary over MySQL. It stores user/account state and the research workspace’s projects, paper records, summaries, and generated output, allowing asynchronous-looking processing results to be retrieved by frontend resource views.

• backend/prisma/schema.prisma

• backend/src/config/prisma.js

• backend/src/controllers/generatedPaperController.js

Cloud MySQL is documented as Aiven-hosted in deployment; local development uses DATABASE_URL.

PDF Ingestion Pipeline

A paper upload enters through multipart middleware, is retained in the backend upload storage area, parsed to text, and then controlled by the paper endpoints/controllers. This separates binary-file handling and PDF extraction from AI and persistence concerns.

• backend/src/routes/paperRoutes.js

• backend/src/services/pdfParserService.js

• backend/uploads/.gitkeep

AI Synthesis Pipeline

Gemini is isolated behind an AI client and focused services. Extracted PDF text feeds metadata extraction and structured summarization; multiple project papers can feed prompt templates and paper generation for comparison/gap analysis and publication-style drafts.

• backend/src/services/AI/geminiClient.js

• backend/src/services/AI/MetadataExtractor.js

• backend/src/services/AI/Summarizer.js

• backend/src/services/AI/PaperGenerator.js

• backend/src/services/AI/PromptTemplates.js

Identity, Notifications, and Deployment

Authentication uses JWTs with refresh-token configuration and supports password-reset delivery through Gmail SMTP. Runtime configuration validates secrets and external connection settings. The frontend is intended for Vercel SPA deployment, while the Node/Express service is deployed via Render; Gemini and Aiven MySQL remain external dependencies.

• backend/src/controllers/authController.js

• backend/src/services/emailService.js

• backend/src/config/env.js

• render.yaml

• frontend/vercel.json