Skip to content
BranchesTags

Latest commit

 

History

319 Commits

Folders and files

NameName
Last commit message
Last commit date
.github/workflows
.github/workflows
 
 
academify
academify
 
 
images
images
 
 
README.md
README.md
 
 

Repository files navigation

Final Project – Web Application Development and Security

Course Code: COMP6703001
Course Name: Web Application Development and Security
Institution: BINUS University International
Project: Academify — Student Community & Collaboration Platform Youtube Video: https://www.youtube.com/watch?v=uriarSAFQII

1. Project Information

Field Value
Project Title Academify
Project Domain Option 8 — Student Community & Collaboration Platform
Class L4AC
Live URL https://e2526-wads-b4ac-02.csbihub.id
API Docs /api-docs on live URL (Swagger UI)

Group Members (max 3 — same class)

Name Student ID Role GitHub Username
Vickelsteins August Santoso 2802505941 Full Stack Vickel317
Harris Ekaputra Suryadi 2802400502 Full Stack HES2209
Kevin Makmur Kurniawan 2802547553 Full Stack kevMkr

2. Instructor & Repository Access

Share this repository with:

3. Project Overview

3.1 Problem Statement

University students need a single place to discuss coursework, join subject communities, collaborate on projects, share files, schedule study events, and message peers — without scattered WhatsApp groups and unmoderated spam.

Target users: Students and lecturers at BINUS (extensible to any academic community).

3.2 Solution Overview

Academify is a full-stack web platform where users:

  • Join forums (subject communities) and post threads with comments and likes
  • Use real-time chat (DMs + collaboration space channels via Socket.IO)
  • Share files with upload validation and optional virus scanning
  • Schedule and RSVP to events
  • Get AI-assisted moderation, recommendations, and thread summaries

AI is core: post moderation, forum/thread recommendations, and discussion summarization — all with documented automated tests (see §10).

4. Technology Stack (mandatory)

Layer Technology
Frontend Next.js 16 (App Router), React 19, Tailwind CSS
Backend Node.js via Next.js API routes + Socket.IO server
API RESTful JSON API, documented with Swagger
Database PostgreSQL + Prisma ORM
Auth Better Auth (email/password + optional Google OAuth)
Storage MinIO (S3-compatible)
AI Ollama (self-hosted LLM) with heuristic fallbacks
Containerization Docker multi-stage Dockerfile + docker-compose
CI/CD GitHub Actions → Docker Hub → VPS deploy
Version Control GitHub

Note: Firebase is not used. Auth is handled entirely by Better Auth + PostgreSQL.

5. System Architecture

5.1 Architecture diagram

Alternative Text

5.2 Architecture explanation

  • Frontend: Next.js App Router with server components (data fetching) and client components (interactivity).
  • API: REST handlers in app/api/** return JSON; business logic in lib/**.
  • Realtime: Browser connects to a separate Socket.IO server (socket-server/, lib/socket-client.ts) for DMs and collab-space typing. Set NEXT_PUBLIC_SOCKET_URL in env (port 3100 in Docker/prod).
  • Database: Prisma only — never accessed directly from the browser.
  • Security: Session cookies (HTTP-only), middleware proxy, role checks in API routes, input sanitization.
  • AI: Server-side calls to Ollama; rate-limited; fallbacks when AI is down.

6. API Design (MANDATORY)

All API's begin with /api/.

6.1 API Endpoints

Authentication Endpoints

Method Endpoint Description Auth Required
POST /auth/sign-up/email Register a new account No
POST /auth/sign-in/email Log in with email and password No
POST /auth/sign-out Log out current user Yes
GET /auth/me Retrieve current user session Yes

User Endpoints

Method Endpoint Description Auth Required
GET /users/ Retrieve all users Yes
GET /users/{userId} Retrieve a specific user's profile Yes
GET /users/{userId}/posts Retrieve posts by a specific user Yes
GET /users/{userId}/events Retrieve events by a specific user Yes
PUT /users/{userId}/avatar Update user avatar Yes
PUT /users/{userId}/banner Update user banner Yes
POST /users/{userId}/follow Follow/unfollow a user Yes
GET /users/connections Retrieve user connections Yes
POST /profile/setup Complete profile setup Yes

Post & Comment Endpoints

Method Endpoint Description Auth Required
GET /posts/ Retrieve all posts Yes
POST /posts/ Create a new post Yes
GET /posts/{postId} Retrieve a specific post Yes
PATCH /posts/{postId} Update a post Yes
DELETE /posts/{postId} Delete a post Yes
POST /posts/{postId}/like Like/unlike a post Yes
GET /posts/{postId}/comments Retrieve comments for a post Yes
POST /posts/{postId}/comments Add a comment to a post Yes
GET /comments/{commentId} Retrieve a specific comment Yes
PATCH /comments/{commentId} Update a comment Yes
DELETE /comments/{commentId} Delete a comment Yes
POST /comments/{commentId}/like Like/unlike a comment Yes

Forum & Category Endpoints

Method Endpoint Description Auth Required
GET /categories/ Retrieve all categories Yes
GET /categories/{id} Retrieve a specific category Yes
GET /forums/{forumId}/members Retrieve forum members Yes
POST /forums/{forumId}/membership Join/leave a forum Yes

Event Endpoints

Method Endpoint Description Auth Required
GET /events/ Retrieve all events Yes
POST /events/ Create a new event Yes
GET /events/{eventId} Retrieve a specific event Yes
PATCH /events/{eventId} Update an event Yes
DELETE /events/{eventId} Delete an event Yes
POST /events/{eventId}/rsvp RSVP to an event Yes
GET /events/{eventId}/attendees Retrieve event attendees (Prisma EventAttendee) Yes

Messaging Endpoints

Method Endpoint Description Auth Required
GET /messages/ Retrieve user's messages Yes
POST /messages/ Send a new message Yes
GET /messages/{userId} Retrieve conversation with user Yes
GET /messages/space/{spaceId} Retrieve space messages Yes

File Endpoints

Method Endpoint Description Auth Required
GET /files/ Retrieve user's files Yes
POST /files/ Upload a file Yes
GET /files/{fileId} Retrieve a specific file Yes
DELETE /files/{fileId} Delete a file Yes
POST /files/{fileId}/share Share a file Yes
POST /files/scan Scan file for malware Yes
POST /storage/upload Upload to object storage Yes
POST /storage/presign Get presigned upload URL Yes
POST /storage/delete Delete from object storage Yes

Moderation & Admin Endpoints

Method Endpoint Description Auth Required
GET /admin/users Retrieve all users (admin) Yes (Admin)
GET /admin/users/{userId} Retrieve user details (admin) Yes (Admin)
PATCH /admin/users/{userId}/role Update user role Yes (Admin)
GET /admin/analytics Retrieve platform analytics Yes (Admin)
GET /reports/ Retrieve reports (scoped to moderated forums for forum mods) Yes (Forum mod / Admin)
GET /reports/{reportId} Retrieve a specific report Yes (Forum mod / Admin)
POST /reports/{reportId}/review Review a report Yes (Forum mod / Admin)
POST /reports/{reportId}/action Take action on report Yes (Forum mod / Admin)
GET /moderation/queue Retrieve moderation queue (forum-scoped for forum mods) Yes (Forum mod / Admin)
POST /moderation/approve/{postId} Approve a post Yes (Forum mod / Admin)
POST /moderation/delete/{postId} Delete flagged content Yes (Forum mod / Admin)
POST /moderation/revert/{postId} Revert moderation action Yes (Forum mod / Admin)
POST /moderation/warn/{userId} Warn a user (sets accountStatus; no action block) Yes (Admin)
POST /moderation/suspend/{userId} Suspend a user (blocks post/comment/DM/upload) Yes (Admin)
POST /moderation/ban/{userId} Ban a user (same restrictions as suspend) Yes (Admin)
POST /moderation/restore/{userId} Restore user to active Yes (Admin)
GET /moderation/logs Retrieve moderation logs Yes (Forum mod / Admin)

AI Endpoints

Method Endpoint Description Auth Required
POST /ai/moderate AI content moderation Yes
GET /ai/summarize/{postId} Summarize a post Yes
GET /ai/recommend AI topic recommendations Yes
GET /ai/recommend/forums AI forum recommendations Yes
GET /ai/health Check AI service health Yes

Collaboration Endpoints

Method Endpoint Description Auth Required
GET /collaboration/ Retrieve collaboration spaces Yes
POST /collaboration/ Create a collaboration space Yes
GET /collaboration/{spaceId} Retrieve a specific space Yes
PATCH /collaboration/{spaceId} Update a space Yes
DELETE /collaboration/{spaceId} Delete a space Yes

Miscellaneous Endpoints

Method Endpoint Description Auth Required
GET /search/ Search posts, users, forums Yes
GET /swagger/ API documentation No

6.2 API Documentation

Example: POST /api/auth/sign-in/email

Request JSON:

{
  "email": "user@example.com",
  "password": "password",
  "name": "string"
}

Response JSON:

{
  "success": true,
  "user": {
    "id": "user_123",
    "email": "user@example.com",
    "name": "John Doe",
    "role": "STUDENT"
  }
}

7. Database Design

7.1 Database Choice

Explain why you chose:

  • PostgreSQL — We chose PostgreSQL because our application requires strong relational integrity for complex data relationships between users, forums, posts, comments, events, messages, and moderation logs. PostgreSQL provides ACID compliance, robust indexing, and mature support for the relational queries our platform demands. Combined with Prisma ORM, we get type-safe database access, automatic migrations, and protection against SQL injection. Hosting on Neon provides serverless scaling with connection pooling for production reliability.

7.2 Schema / Data Structure

Academify ERD

Key models in our schema:

Model Description
User Core user profile with role-based attributes (STUDENT, LECTURER, ADMIN), privacy settings, and collaboration metadata
ForumHub Discussion forums with moderators and members
ForumMember Many-to-many relationship between users and forums
ForumModerator Moderator assignments per forum
Post Forum posts with moderation status, AI scoring, and summaries
Comment Nested comments on posts with like support
CommentLike Like tracking for comments
PostLike Like tracking for posts
Event Scheduled events with RSVP and attendee tracking
EventAttendee Event participation with roles (GUEST, HOST)
File File attachments linked to posts or collaboration spaces
Message Direct messaging between users and within collaboration spaces
CollabSpace Collaboration spaces within forums with role-based access
SpaceMember Membership in collaboration spaces
ReportReview User-submitted reports for content moderation
ModerationActionLog Audit log for all moderation actions taken
Follow User follow relationships
AuthUser Better Auth user table
AuthSession Better Auth session table
AuthAccount Better Auth account/provider table

8. AI Features (mandatory)

8.1 AI feature list

AI Feature Purpose Type
Content moderation Auto-approve/flag/block new posts NLP classification
Thread summarization Key points + open questions from discussion NLP generation
Thread recommendations Suggest posts based on profile & activity Recommendation
Forum recommendations Suggest forums to join Recommendation

Content moderation is the one AI feature whose output quality is measured rather than asserted, check 10.5 AI decision quality for the labeled evaluation and its limitations.

8.2 AI Integration Flow

Explain:

Content Moderation:

  • Input → AI processing → Output
    • User submits a post or comment → Text is sent to local Ollama instance for classification → AI returns moderation decision (APPROVED, FLAGGED, or BLOCKED) → Result stored in moderationStatus field on the Post → Content visibility enforced based on status
  • How AI results are used in the system
    • Posts with PENDING status are held for review. APPROVED posts are immediately visible. FLAGGED posts enter the moderation queue for moderator action. Low-risk off_topic-only flags are auto-downgraded to APPROVED to reduce false positives on beginner questions. If Ollama is unavailable, profanity-based heuristic fallback is used.

Thread Summarization:

  • Input → AI processing → Output
    • User opens a thread → Top 20 comments by likes + post body are collected → Text sent to Ollama for summarization → AI returns key points and open questions → Summary cached as JSON on the Post model
  • How AI results are used in the system
    • Cached summaries are displayed at the top of thread views. Summaries are regenerated only when comment count changes significantly, reducing unnecessary AI calls.

Topic & Forum Recommendations:

  • Input → AI processing → Output
    • User's profile interests + forum metadata are collected → Semantic similarity computed via Ollama → Ranked list of relevant forums returned → Personalized feed filtered by engagement patterns
  • How AI results are used in the system
    • Recommendations appear in the sidebar and "Recommended for You" section. Results are personalized per user and refreshed on each session.

9. Security Implementation (mandatory)

Control Implementation
Authentication Better Auth session cookies (lib/auth.ts), HTTP-only, secure in production
Authorization Roles: student, lecturer, admin; per-forum moderators via ForumModerator (content queue, reports, approve/delete scoped to their forums); platform admin for account sanctions (warn / suspend / ban / restore)
Input validation lib/validation.ts on API bodies
XSS mitigation lib/sanitization.ts — HTML stripped from user text
SQL injection Prisma parameterized queries only
CSRF SameSite cookies + same-origin API calls
Rate limiting lib/ai/rate-limit.ts, lib/rate-limit.ts on AI endpoints
File security Extension/MIME checks, size limits, ClamAV (clamscan + freshclam in Docker)
API keys Ollama/MinIO secrets in env only — never in client bundle
Realtime chat Socket.IO over WebSocket; CORS on socket server; NEXT_PUBLIC_SOCKET_URL for client
Moderation audit ModerationActionLog table

10. Testing Documentation

Test layers

Layer Tool What it tests Location
Unit Jest Pure functions, helpers, AI mappers __tests__/*.test.ts
Component / page Jest + React Testing Library UI rendering, forms, interactions __tests__/*.test.tsx
API (mocked DB) Jest + NextRequest Route handlers, auth guards, validation __tests__/api-*.test.ts, security-critical.test.ts
Integration (real DB) Jest + PostgreSQL API ↔ Prisma ↔ PostgreSQL __tests__/integration/*.int.test.ts
E2E Playwright Full browser smoke tests on running app e2e/*.spec.ts
AI quality Custom harness + live Ollama Moderation decision accuracy against labeled ground truth eval/

E2E tests the whole stack (browser → Next.js → API → DB). Playwright automates Chromium. E2E is slower and runs against a live server; unit/integration tests run in Node.

Current coverage (~66% statements) focuses on critical paths. See the coverage matrix below.

Running tests

npm test                 # Unit + component + mocked API tests
npm run test:coverage    # Coverage report → coverage/
npm run test:integration # Real PostgreSQL integration tests
npm run test:e2e         # Playwright browser smoke tests
npm run eval:moderation  # AI moderation quality vs. labeled ground truth (needs Ollama)

For integration tests locally:

createdb academify_test   # once
TEST_DATABASE_URL=postgresql://user:pass@localhost:5432/academify_test npx prisma migrate deploy
npm run test:integration

CI runs lint and unit tests with coverage on every push to main / master. Integration tests, production builds, Playwright E2E, and the AI quality evaluation are run locally before release. The last of these needs a reachable Ollama instance and costs roughly two minutes of inference, so it is not part of CI.

10.1 Frontend testing

ID Scenario Expected result Automated test Status
FE-01 Login with invalid email Inline validation error __tests__/login.test.tsx Pass
FE-02 Login with empty password Inline validation error __tests__/login.test.tsx Pass
FE-03 Register password mismatch Error shown __tests__/register.test.tsx Pass
FE-04 Register without terms Error shown __tests__/register.test.tsx Pass
FE-05 Dashboard renders for session Main sections visible __tests__/dashboard.test.tsx Pass
FE-06 Forums list renders Forum cards / heading __tests__/forums.test.tsx Pass
FE-07 Profile edit validation Field errors __tests__/profile-edit.test.tsx Pass
FE-08 Topbar search mount No hydration crash __tests__/topbar.test.tsx Pass
FE-09 Login page E2E smoke Page loads in browser e2e/smoke.spec.ts Pass
FE-10 Register page E2E smoke Page loads in browser e2e/smoke.spec.ts Pass
FE-11 Events page renders Calendar + create button __tests__/events.test.tsx Pass
FE-12 Profile setup (student) Education dropdown + skills __tests__/profile-setup.test.tsx Pass
FE-13 Admin panel Stats or forbidden message __tests__/admin.test.tsx Pass
FE-14 Protected routes E2E smoke No 500 on key pages e2e/smoke.spec.ts Pass

10.2 Backend & API testing

ID Endpoint / area Input / case Expected output Test file Status
API-01 POST /api/posts Unauthenticated 401 api-authorization.test.ts Pass
API-02 PUT /api/categories/:id Non-admin/non-mod 403 categories-authorization.test.ts Pass
API-03 GET /api/files/:id Non-owner 403/404 files-authorization.test.ts Pass
API-04 POST /api/posts/:id/like Toggle like 200 + count post-like.test.ts Pass
API-05 GET /api/categories List forums 200 + JSON integration/categories.int.test.ts Pass
API-06 POST /api/posts/:id/comments XSS in body Sanitized in DB integration/post-comments.int.test.ts Pass
API-07 GET /api/ai/summarize/:id Unauthenticated 401 ai-summarize.test.ts Pass
API-08 GET /api/ai/recommend/forums Authenticated 200 + forums ai-recommend-forums.test.ts Pass
API-09 POST /api/events Unauthenticated 401 api-events.test.ts Pass
API-10 GET /api/reports Student role 403 api-reports.test.ts Pass
API-11 GET /api/events Upcoming list 200 + DB rows integration/events.int.test.ts Pass
API-12 POST /api/collaboration Create space 201 + owner row integration/collaboration.int.test.ts Pass
API-13 POST /api/profile/setup Student education profileSetupComplete integration/profile-setup.int.test.ts Pass
API-14 POST /api/moderation/approve/:id Forum moderator 200 api-moderation-access.test.ts Pass
API-15 POST /api/moderation/ban/:userId Admin persists accountStatus 200 + BANNED api-moderation-access.test.ts Pass
API-16 GET /api/events/:id/attendees Authenticated 200 + Prisma rows api-events-attendees.test.ts Pass
API-17 GET /api/reports Forum moderator (ForumModerator) 200 api-reports.test.ts, integration/reports.int.test.ts Pass

10.3 Security testing

ID Attack / case Expected behavior Test file Status
SEC-01 XSS in message body Tags stripped before save security-critical.test.ts Pass
SEC-02 XSS in comment body Tags stripped (integration) integration/post-comments.int.test.ts Pass
SEC-03 Dangerous upload filename Rejected security-critical.test.ts Pass
SEC-04 Restricted user creates post 403 security-critical.test.ts Pass
SEC-05 Non-author views pending post Denied integration/post-visibility.int.test.ts Pass
SEC-06 AI rate limit exceeded 429 ai-rate-limit.test.ts Pass
SEC-07 Input sanitization helper HTML escaped security-critical.test.ts Pass

10.4 AI functionality testing

AI Feature 1: Content moderation (Ollama + heuristics)

ID Input Expected output Test file Status
AI-MOD-01 Clean academic text APPROVED ai-moderation.test.ts Pass
AI-MOD-02 Profanity / slurs FLAGGED or BLOCKED ai-moderation.test.ts Pass
AI-MOD-03 Ollama timeout / failure Heuristic fallback ai-moderation.test.ts Pass
AI-MOD-04 Malformed model JSON Safe fallback status ai-moderation.test.ts Pass
AI-MOD-05 Low-risk off-topic-only flag Auto-approved (downgraded) ai-moderation.test.ts Pass

Failure handling: lib/ai/post-moderation.ts falls back to profanity heuristics when Ollama is unavailable.

AI Feature 2: Thread summarization

ID Input Expected output Test file Status
AI-SUM-01 Valid post + comments Summary JSON ai-summarize.test.ts Pass
AI-SUM-02 Unauthenticated request 401 ai-summarize.test.ts Pass
AI-SUM-03 Hidden / pending post (non-author) 404 ai-summarize.test.ts Pass
AI-SUM-04 Rate limit exceeded 429 ai-summarize.test.ts Pass

Failure handling: app/api/ai/summarize/[postId]/route.ts returns cached summaries when available; on Ollama timeout or error it responds with 503 AI_UNAVAILABLE / 502 AI_ERROR, and the AiSummary UI shows a retry button.

AI Feature 3: Forum & thread recommendations

ID Input Expected output Test file Status
AI-REC-01 User with profile Ranked suggestions ai-recommend-forums.test.ts Pass
AI-REC-02 Ollama failure Heuristic fallback list ai-recommend-forums.test.ts Pass

Failure handling: Recommendation routes (/api/ai/recommend, /api/ai/recommend/forums) call Ollama with a short timeout and maxRetries: 0; on timeout, rate limit, or schema error they return heuristic scores instead with fallback: true, and the AiRecommend client can re-request with ?heuristic=1.

10.5 AI decision quality (moderation)

The tests in 10.4 check that the moderation pipeline behaves correctly: valid JSON in, correct status out, graceful fallback on failure. They say nothing about whether its judgments are any good. This section measures that.

npm run eval:moderation -- --runs=3   # ~2 min, requires a reachable Ollama instance

eval/run-moderation-eval.ts sends a labeled sample set through the same runPostModeration path used by POST /api/posts and scores the resulting statuses against ground truth. The full report, including the confusion matrix and every per-sample decision, is committed at eval/results/moderation-eval.md.

Dataset. 42 synthetic posts with an expected outcome and a written rationale each, in eval/moderation-dataset.ts:

  • 14 clean: ten in polished academic prose, plus four rough-but-legitimate posts (all-caps panic, broken non-native English, a bare code dump, Indonesian/English code-switching) that a tone-sensitive filter might wrongly reject.
  • 16 harmful: ten overt (targeted harassment, a threat of violence, doxxing, crypto and contract-cheating spam, keyword stuffing) and six evasive, written specifically to defeat a keyword or tone filter: digit-substituted abuse, veiled harassment with plausible deniability, sex-based discouragement in a warm register, defamation via "just asking questions", an essay mill in empathetic peer-support prose, and credential phishing formatted as a competent library notice.
  • 12 borderline: heated academic criticism, off-topic chatter, self-promotion, a past-papers request, idiomatic "this semester is killing me", and a user quoting abuse in order to report it.

Results (llama3.1:8b, 3 passes per sample, 126 decisions total):

Metric Result
Harmful posts withheld from the public feed 12 / 16 (37 / 48 individual passes)
Overt subset 10 / 10 (100%)
Evasive subset 2 / 6 (33%)
Clean posts wrongly withheld 1 / 14
Agreement on the 30 decisive samples 83.3% (82.2% scored per individual pass)
Precision / recall / F1 on the "withhold" class 0.923 / 0.750 / 0.828
Exact-status agreement across all 42 samples 57.1%
Median decision latency ~3.4 s

Precision and recall are computed only over the 30 clean and harmful samples; the borderline twelve have no single correct answer by construction, so scoring them would make the figures unfalsifiable. They are reported as a distribution instead. Recall is the metric that matters most here, because a false negative puts abusive content in front of students while a false positive only delays a legitimate post until a moderator clears the queue.

The main finding: the classifier reads tone, not intent. Every overt sample was caught, and two thirds of the evasive ones were not. The four that reached the public feed were the phishing notice written in correct institutional register (aiScore 0, labeled approve), the essay mill written as peer support, the sex-based discouragement written kindly, and the defamation framed as "just asking questions". The failure mode is consistent and worth stating plainly: harmful content survives if it is polite. The one false positive is the same mechanism inverted: a legitimate all-caps plea for help was flagged for "urgent tone and desperation". Both directions are documented in the report's Detection by difficulty section. An earlier version of this evaluation used only overt harmful samples and scored a perfect 1.000 recall; that number measured the dataset, not the system.

Why exact-status agreement is much lower than the headline. Ten of the sixteen harmful posts were withheld as FLAGGED rather than BLOCKED. mapAiResultToModerationStatus only auto-blocks when the model returns decision: "reject" and aiScore >= 0.9, and llama3.1:8b almost always prefers "flag". In this run two samples cleared the score threshold and still were not blocked. Auto-blocking therefore almost never fires and nearly everything harmful routes to the human queue. That is a moderator-workload cost rather than a safety one; the safety cost sits entirely with the four false negatives above.

Limitations. Ground truth is single-rater, assigned before running the pipeline, so there is no inter-rater agreement figure; the borderline labels in particular are one defensible reading rather than an objective answer. Forty-two synthetic samples is enough to expose a systematic failure mode but not enough for tight confidence intervals. The model is sampled rather than greedy: 33 of 42 samples returned an identical status on all three passes, and the report lists the nine that did not, several of them evasive samples flipping between APPROVED and FLAGGED. Headline figures therefore move by a point or two between runs; an earlier run of this same set scored 13/16 rather than 12/16. Ollama was healthy throughout, so the profanity fallback path was never exercised (0 / 126 decisions). This evaluates the moderation classifier only; summarization and recommendation quality remain unmeasured.

Coverage matrix

Area Unit/component Integration E2E
Auth (login/register/setup) ✅ ✅ setup ✅ smoke
Forums / threads ✅ ✅ categories ✅ smoke
Posts / comments ✅ ✅ comments + visibility ✅ smoke
Messages / chat ✅ partial ✅ ✅ smoke
Events ✅ ✅ list ✅ smoke
Files / uploads ✅ partial ✅ ✅ smoke
Collaboration ✅ ✅ create space ✅ smoke
Moderation / reports ✅ partial ✅ ✅
Admin ✅ ✅ ✅ smoke
AI endpoints ✅ ✅ ✅

Legend: ✅ covered · ⬜ add next · partial = some paths only

11. Deployment & Production Setup

11.1 Docker setup

  • Dockerfile — multi-stage: runner (Next.js), socket, migrator
  • docker-compose.yml — app, socket, minio, optional db-schema-sync profile
docker compose --env-file .env.production up -d app socket minio

11.2 Production environment

Secrets via .env.production (reconstructed from GitHub Secrets in CI).

Variable Purpose
DATABASE_URL / DIRECT_URL PostgreSQL (Neon)
BETTER_AUTH_SECRET Session signing
BETTER_AUTH_URL Server auth base URL
GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET Optional Google OAuth
NEXT_PUBLIC_SOCKET_URL Browser Socket.IO endpoint (e.g. https://socket-… or http://localhost:3100)
OLLAMA_API_BASE AI server
MINIO_* File storage

11.3 Live application URL

https://e2526-wads-b4ac-02.csbihub.id

12. GitHub Contribution Summary (individual)

Student Name: HES2209 (Harris Ekaputra Suryadi)

  • Features implemented: Project bootstrap (Next.js app structure, login/register pages, navbar, early dashboard), file sharing & collab spaces (collab pages/API, files library, MinIO Docker setup, presigned uploads, file share via DM), profile media (avatar upload/proxy + storage delete), auth migration (Better Auth integration + Firebase bridge route), Prisma/schema alignment (Category → ForumHub, enum/orderBy fixes, collab roles), upload reliability fixes (avatar/banner/files), collab-space chat UI bug fix, moderation logging + admin analytics enrichment.
  • API endpoints handled:
    • Posts: GET/POST /api/posts, GET/PATCH/DELETE /api/posts/{postId}, GET/POST /api/posts/{postId}/comments
    • Messages: GET/POST /api/messages, GET/POST /api/messages/{userId}
    • Files: GET/POST /api/files, GET/DELETE /api/files/{fileId}, POST /api/files/{fileId}/share, POST /api/files/scan
    • Collaboration: GET/POST /api/collaboration, GET/POST/DELETE /api/collaboration/{spaceId}
    • Storage: POST /api/storage/presign, avatar upload + storage delete
    • Users: GET /api/users/{userId}/avatar
    • Moderation hardening (approve/delete/warn/suspend/ban/queue/logs): implemented under /api/moderation/* and /api/reports/* with admin hooks under /api/admin/*
    • Admin: analytics enhancements under /api/admin/*
    • Broad security pass across routes under /api/forums, /api/events, /api/categories, /api/posts, /api/comments, /api/messages, /api/users
  • Tests written: Security and authorization test coverage and updates across middleware/security-critical paths; updated collaboration/files/messages/profile-edit tests (plus rate-limit and authz tests).
  • Security work: Primary security owner.
    • lib/security.ts — centralized security utilities (~800 lines)
    • middleware.ts / proxy.ts — security headers + route protection
    • lib/sanitization.ts — HTML stripping / XSS mitigation
    • lib/validation.ts — request body validation and upload filename/MIME/size checks
    • lib/rate-limit.ts — API rate limiting (write/auth/read buckets)
    • ClamAV virus scanning: lib/clamav.ts, POST /api/files/scan, plus Docker/compose integration
    • Moderation guards: restricted-account checks for posts/messages/comments
    • Socket server: security additions in socket-server/index.ts
    • Security documentation: CHECKPOINT_09_WEB_SECURITY.md
  • AI-related work: None identified in Harris’s commits (AI moderation/recommend/summarize implemented by Vickel’s work in shared commits).

Student Name: kevMkr (Kevin Makmur Kurniawan)

  • Features implemented: Events (events list/detail pages, RSVP actions, event creation flow, event API extensions, dashboard calendar widget, event banner upload, and forum–event integration); Forums UI (forum list/detail, new thread page, comment + like buttons on thread pages, forum seed script); Files UI (early files page shell); Public profiles (public profile page /profile/[userId], profile banner display, report-user modal/button); Profile UX (profile stabilization, profile edit improvements, chat profile popover); Landing & layout (login entry, responsive layouts, sidebar/topbar polish); Admin & roles (admin users page updates, role-aware forums/connections UI, admin role API tweaks); Notifications (initial in-app notification model and topbar notification UI); Storage/banners (profile banner upload, user banner serving, entity banner upload for events/collab spaces/forums).
  • API endpoints handled:
    • Authentication (legacy, pre–Better Auth): POST /api/auth/login, /api/auth/register, /api/auth/logout, /api/auth/refresh, GET /api/auth/me
    • Events: GET/POST /api/events, GET/PUT/DELETE /api/events/{eventId}, POST/DELETE /api/events/{eventId}/rsvp, GET /api/events/{eventId}/attendees
    • Users: GET/PATCH /api/users/{userId}, GET /api/users/{userId}/posts, GET /api/users/{userId}/events
    • Forums/categories: GET/POST /api/categories, category detail routes, forum-related post routes
    • Messages: DM route tweaks + GET /api/notifications
    • Storage/banners: POST /api/storage/upload-banner, GET /api/users/{userId}/banner, POST /api/storage/upload-entity-banner/{type}/{id}
    • Admin role management: PATCH /api/admin/users/{userId}/role
  • Tests written: No major dedicated test suites authored by Kevin in git history (only minor test/line-fix commits). Main test coverage for his areas was added by others (e.g. events.test.tsx, api-events.test.ts).
  • Security work:
    • Report user from profile (UI + report-profile modal)
    • Admin role management UI/API alignment
    • Not the primary security implementer; auth/events routes were later hardened by Harris and Vickel.
  • AI-related work: None identified in Kevin’s commits.

Student Name: Vickel317 (Vickelsteins August Santoso)

  • Features implemented:

    • Auth & onboarding: Migrated from Firebase to Better Auth (email/password + Google OAuth), unified API auth layer, profile setup flow, lecturer profile edit aligned with setup, mock OAuth for tests, OAuth error fixes
    • Dashboard & posts: Personalized dashboard with real connection counts, admin/user post CRUD, post likes, post visibility rules (pending/blocked), comment likes, comment tree sorting by engagement
    • Forums & moderation: Neon/Prisma ORM for forums/categories/comments, forum access control (only admins/lecturers create forums), moderator forum settings & member management, admin reports page + report modal, forum/thread/user reporting in DB and admin panel, streamlined admin nav (removed admin edit on posts/comments)
    • AI-powered UX: Async post moderation with profanity fallback, staff revert of AI decisions, thread summarization with DB cache, forum & thread recommendations with heuristic fallback, AI recommend UI, Reddit-style sidebar + polished navbar/topbar
    • Realtime messaging: Standalone Socket.IO server for DMs and collab chat, typing indicators, online status, message search modal, reconnect handling, group chat typing in collab spaces
    • Files & storage: MinIO upload routes (server + presigned), files UI, collab space file upload, avatar/banner/entity banner uploads, file sharing
    • Profiles & social: User profile/connections schema, follow model, privacy settings page, profile education fields, public profile view button, DM restriction rules (ALL, CONNECTIONS, LECTURERS, NONE)
    • Search: Global search across users, forums, and threads
    • Events & collaboration: Collab space API/UI polish, events UI integration (alongside teammates’ event work)
    • DevOps & docs: Docker + GitHub Actions CI/CD (app + socket server), deployment compose fixes, project overview/README/technology stack, full Swagger UI at /api-docs with OpenAPI spec for all endpoints, freshclam for ClamAV in Docker
    • UI polish: Thread/comment edit placement, admin/forums/files/collab/events UI refinements, removed notification feature (later cleanup)

  • API endpoints handled (high-level, implemented/extended under app/api/**):

    • Auth: POST /api/auth/sign-up/email, /api/auth/sign-in/email, /api/auth/sign-out, GET /api/auth/get-session, plus /api/auth/[...all]
    • Users & profile: GET/PATCH/PUT/DELETE /api/users/{userId}, /api/users, /api/users/connections, /api/users/{userId}/follow, /api/users/{userId}/posts, /api/users/{userId}/events, /api/users/{userId}/avatar, /api/users/{userId}/banner, POST /api/profile/setup
    • Search: GET /api/search
    • Forums: GET/POST/DELETE /api/forums/{forumId}/membership, GET/PUT /api/forums/{forumId}/members
    • Categories: GET/POST /api/categories, GET/PATCH/DELETE /api/categories/{id}
    • Posts/comments: GET/POST /api/posts, GET/PATCH/DELETE /api/posts/{postId}, GET/POST /api/posts/{postId}/comments, /api/posts/{postId}/like, GET/PATCH/DELETE /api/comments/{commentId}, /api/comments/{commentId}/like
    • Messages: GET/POST /api/messages, /api/messages/{userId}, /api/messages/space/{spaceId}
    • Events: GET/POST /api/events, GET/PUT/DELETE /api/events/{eventId}, RSVP + attendees endpoints
    • Files/storage: GET/POST/DELETE /api/files, /api/files/{fileId}, /api/files/{fileId}/share, POST /api/files/scan, storage upload/presign/delete/avatar/banner/entity-banner
    • Collaboration: GET/POST /api/collaboration, GET/POST/DELETE /api/collaboration/{spaceId}
    • AI: GET /api/ai/health, POST /api/ai/moderate, GET /api/ai/recommend, /api/ai/recommend/forums, GET /api/ai/summarize/{postId}
    • Moderation: queue/logs and approve/delete/revert/warn/suspend/ban/restore; per-forum mod auth via ForumModerator + lib/report-access.ts; admin users UI for sanctions; GET /api/events/{eventId}/attendees backed by Prisma
    • Reports: GET/POST /api/reports, GET/PATCH /api/reports/{reportId}, review + action
    • Admin: GET /api/admin/analytics, user management + role assignment
    • Docs: GET /api/swagger

  • Tests written:

    • Unit/component (Jest): login, register, profile, profile-edit, profile-setup, profile-education, admin, dashboard, forums, files, messages, collaboration, events, sidebar, topbar, post-comments, post-like, comment-sort, avatar-url, forum-access, categories-authorization, files-authorization, message-access, api-authorization, api-events, api-events-attendees, api-reports, api-moderation-access, api-search, rate-limit, security-critical
    • AI tests: ai-moderation, ai-summarize, ai-recommend-forums, ai-rate-limit
    • Integration (DB): categories, post-comments, post-visibility, profile-setup, collaboration, events
    • E2E (Playwright): e2e/smoke.spec.ts, e2e/login.spec.ts, e2e/register.spec.ts
    • CI: Jest + Playwright wired into GitHub Actions; Playwright artifacts ignored from git

  • Security work:

    • Unified Better Auth session cookies (httpOnly, sameSite: lax, secure in prod)
    • Role-based authorization (student/lecturer/admin + per-forum moderators) with tests in api-authorization.test.ts, categories-authorization.test.ts, files-authorization.test.ts
    • Input validation (lib/validation.ts) and XSS sanitization (lib/sanitization.ts) on messages, posts, and comments — covered in security-critical.test.ts
    • Rate limiting on write/auth/read and AI endpoints (lib/rate-limit.ts, lib/ai/rate-limit.ts)
    • Upload security: dangerous filename rejection, extension/MIME checks, size limits
    • ClamAV integration: freshclam in Docker; virus scan route (POST /api/files/scan)
    • Account restrictions: SUSPENDED / BANNED users prevented from posting, messaging, commenting, and uploading (lib/moderation.ts → isRestrictedAccount)
    • DM privacy: canSendDirectMessage rules by restriction setting
    • Removed hardcoded mock identity; API keys/secrets env-only

  • AI-related work:

    • Content moderation: Ollama-based async post moderation (lib/ai/post-moderation.ts), profanity heuristic fallback, moderation queue statuses, staff revert via POST /api/moderation/revert/{postId}
    • Thread summarization: GET /api/ai/summarize/{postId} with DB cache, refresh=1 bypass, model name from env (not hardcoded)
    • Recommendations: GET /api/ai/recommend (threads) and /api/ai/recommend/forums with heuristic fallback when Ollama is down or rate-limited
    • Prompts & schemas: lib/ai/prompts.ts, lib/ai/schemas.ts, ModerationResultSchema validation
    • Rate limits: per-route AI throttling (checkAiRateLimit)
    • Health check: GET /api/ai/health for Ollama availability
    • UI: AI recommend components, moderation status on posts, summary display on threads

13. AI Usage Disclosure (mandatory)

Tools used: Github Copilot and Cursor AI
Purpose: To help us with the code, answers questions, and explain errors that happens inside the code Parts assisted: Helps us in the API documentation, Testing, API routes, and CI/CD pipeline

14. Known Limitations & Future Improvements

  • Ollama quality depends on host GPU/CPU; heuristics used when offline
  • Moderation misses politely-worded abuse. Measured at 10/10 detection on overt harmful content but 2/6 on evasive content written in a calm, fluent register (see 10.5). Credential phishing formatted as an official notice was approved outright. Until this is addressed, the moderation queue should not be treated as a substitute for user reporting
  • E2E tests are smoke-level only (authenticated flows not fully automated)
  • Email verification not enabled for Better Auth email provider
  • Account sanctions: warn / suspend / ban are platform-wide (User.accountStatus); forum-scoped suspend is not implemented. Suspend duration is logged but not auto-expired (admin must Restore). WARNED is visible in admin only — does not block actions; SUSPENDED and BANNED block new posts, comments, DMs, and uploads
  • Shadow ban: isShadowBanned column exists but is not enforced in API checks
  • User-only reports (no linked post/comment/forum) are visible to platform admins only

15. Final Declaration

We declare that:

  • This project is our own work
  • AI usage is disclosed honestly
  • All group members understand the system

Signed: Harris Ekaputra Suryadi · Kevin Makmur Kurniawan · Vickelsteins August Santoso

16. SETUP (local development)

Prerequisites

  • Node.js 22+
  • PostgreSQL (local or Neon)
  • Optional: Ollama, MinIO, Docker

Steps

git clone <https://github.com/Vickel317/WADS_Final_Project.git>
cd academify
cp .env.example .env.local
# Edit .env.local — set DATABASE_URL, BETTER_AUTH_*, NEXT_PUBLIC_SOCKET_URL, etc.

npm ci
npx prisma migrate deploy
npx prisma generate

# Next.js + Socket.IO (one command)
npm run dev:all

# Or two terminals:
#   npm run dev
#   SOCKET_PORT=3100 npm run dev:socket

# Open http://localhost:3000

Realtime chat: set NEXT_PUBLIC_SOCKET_URL=http://localhost:3100 in .env.local and run the socket server on the same port (SOCKET_PORT=3100). Messages are still saved via REST; Socket.IO delivers live updates.

Google OAuth (optional)

  1. In Google Cloud Console, create a Web application OAuth client.
  2. Add authorized redirect URIs:
    • http://localhost:3000/api/auth/callback/google
    • https://e2526-wads-b4ac-02.csbihub.id/api/auth/callback/google
  3. Set GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET in .env.local (and GitHub Secrets for deploy).
  4. Ensure BETTER_AUTH_URL matches the site origin (no trailing slash).

17. DEPLOYMENT INSTRUCTIONS

  1. Push to main / master → GitHub Actions runs lint, unit tests, Docker build/push, and VPS deploy
  2. On success: Docker images pushed to Docker Hub
  3. Self-hosted runner on VPS: prisma migrate deploy + docker compose up -d
  4. Ensure GitHub Secrets match .env.example (no Firebase secrets needed)

Remove old Firebase secrets from GitHub

Delete these repository secrets if present (no longer used):

  • FIREBASE_WEB_API_KEY
  • NEXT_PUBLIC_FIREBASE_API_KEY
  • NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN
  • NEXT_PUBLIC_FIREBASE_PROJECT_ID
  • NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET
  • NEXT_PUBLIC_FIREBASE_MESSAGING_SENDER_ID
  • NEXT_PUBLIC_FIREBASE_APP_ID

Add instead (for Google login):

  • GOOGLE_CLIENT_ID
  • GOOGLE_CLIENT_SECRET

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages