Klar — Feature List (as of 25 August 2026)

Accounts & Auth

• Registration (username, email, password) with email verification link
• Login / logout, JWT access tokens (15 min) + refresh tokens (30 days)
• Cross-site auth support: tokens in localStorage + Authorization: Bearer, with httpOnly cookies as a same-site fallback (handles klarsocial.eu vs klarsocial.de being different top-level domains)
• Password reset (forgot password → email link → reset)
• Resend verification email
• Argon2 password hashing
• Change password (requires current password, invalidates all other sessions)
• Account deletion (cascades to posts/comments/messages/media; cleans up stored files)
• Self-service data export (GDPR Art. 15/20) — full JSON download of everything Klar holds about you

Profile

• Username (case preserved as entered, case-insensitive uniqueness, 14-day change cooldown)
• Display name, bio, avatar (upload, resized/processed server-side)
• Public profile page: avatar, stats (posts/followers/following), bio
• Private accounts: toggle in Settings; when on, posts are hidden from non-followers
• Follow requests for private accounts: request → pending → accept/reject (not instant-follow)
  • Accept/decline available in the notification dropdown and directly on the requester's profile page
  • Dedicated /follow-requests management page

Posts & Media

• Create post: caption + optional photo
• Automatic EXIF stripping on upload (location, device info, timestamp removed)
• Three generated sizes per image (thumbnail/medium/full), real width/height captured for correct aspect-ratio rendering
• Edit caption, delete post (with real cascade cleanup of media files)
• Post permalink pages (/posts/[id]) — shareable, deep-linkable
• Post detail modal: comments, likes, delete (from feed, profile, or permalink)

Feed

• Personal feed: strictly chronological, no algorithm, no ranking — deliberate product stance
• Fan-out-on-write (feed_items) for fast reads; backfilled on follow, cleaned up on unfollow
• Discovery feed: global, cross-user, cursor-paginated
• Private-account posts are excluded from the discovery feed and profile grids unless you're the owner or an accepted follower (enforced backend-side in three places: single post view, profile posts list, and the discovery feed itself)

Social Graph

• Follow / unfollow (instant for public accounts, request-based for private ones)
• Followers / following lists
• Block / unblock (blocked users can't follow, like, or comment on your posts)

Comments & Likes

• Post likes, comment likes
• Comments with threaded replies
• Edit/delete own comments; post owners can also delete comments on their posts

Notifications

• Types: post like, comment, comment like, follow, follow request, follow accepted
• Persisted history (GET /notifications) + real-time delivery via SSE
• Cross-replica delivery via Redis (Upstash) pub/sub — works correctly however many backend instances are running, not just replica-count-1
• Notification previews: actor avatar (follows/requests) or post thumbnail (likes/comments)
• Clicking a notification navigates to the relevant post or profile
• Follow-request notifications have inline Accept/Decline buttons, not just a link

Direct Messages (Chat)

• 1:1 conversations, gated to mutual followers only
• Send, edit, delete messages; reply-to-message threading
• Emoji reactions (8-emoji picker; can react to your own or the other person's messages)
• Read receipts
• Real-time message + reaction delivery (same Redis/SSE pipeline as notifications)
• Unread-message badge on the Chat icon, correctly clears when actively viewing a conversation (no false positives from your own open chat)
• Conversation list shows the true most-recent activity — a plain message, a reply, or a reaction — with proper phrasing ("Me: ...", "X replied: ...", "X reacted 👍 to your message: ...")
• Avatar in conversation list/chat header links to the other person's profile; rest of the row opens the chat

Search

• User search by username or display name

Moderation & Reporting

• Report posts, comments, or user accounts (POST /reports) with a fixed reason taxonomy: spam, harassment, hate speech, violence/graphic content, self-harm, sexual content, CSAM, impersonation, other
• Auto-moderation on submission: a CSAM report hides the content immediately, zero tolerance, no threshold; violence/self-harm/sexual-content reports flag the content (shown behind an interstitial rather than removed outright — a single report shouldn't have unilateral takedown power)
• Admin review queue (/admin/reports), critical reports sorted first: dismiss (restores visibility) or remove content outright, each with an optional internal review note
• Admin access gated via an ADMIN_EMAILS allow-list, requiring the matching account's email to also be verified (prevents someone from claiming an admin address by registering it first)
• Not yet built: DSA-mandated automated "Statement of Reasons" notification to affected users, appeal/counter-notice flow, UrhDaG rights-holder copyright takedown portal

Privacy & Compliance

• Impressum, Datenschutzerklärung (privacy policy), Nutzungsbedingungen (ToS), and a plain-language Transparenz page explaining data handling in everyday terms
• Legal pages carry a version-controlled "Stand:" date, auto-updated by a GitHub Actions bot commit whenever the underlying legal page file changes
• ToS + privacy consent checkbox at registration, enforced server-side (POST /auth/register rejects the request with 400 if accept_terms isn't true, and 422 if it's missing entirely) — not just a client-side gate anymore. Acceptance is timestamped (users.terms_accepted_at) and included in the self-service data export
• No tracking/advertising cookies — only functional auth storage
• Documented sub-processors: Bunny.net (hosting/CDN/storage and self-hosted Postgres, German DC), Scaleway (transactional email, EU), Upstash (real-time notification relay, US — SCCs apply)

Pre-Launch

• Site-wide passcode gate (/welcome) via Next.js proxy.ts — blocks the whole site except the gate page and legal pages until a shared passcode is entered; disabled automatically if no passcode is configured (e.g. local dev)

Frontend/UX Details

• Consistent top navigation bar across all primary pages (Feed, Discovery, Search, Chats, Profile) with active-state icon highlighting
• Mobile-correct viewport handling (dvh instead of vh) so nav bars don't get clipped by mobile browser chrome
• Root URL smart-redirects to Feed (logged in) or Login (logged out)
• Global cursor-pointer fix for all interactive elements

Backend/Infra

• Rust (Axum) backend, self-hosted PostgreSQL 18 in a Bunny Magic Container (Frankfurt) — TLS enforced even on loopback, connection pool retries through pod-startup races
• Nightly automated Postgres backups (db-backup sidecar → private Bunny Storage Zone, no CDN pull zone); durability handled by off-pod storage, so the sidecar itself needs no persistent volume
• UUIDv7 primary keys, denormalized counters (follower/following/post/like/comment counts), hash-partitioned likes/notifications, monthly-partitioned interaction event log (post_events — logging foundation for possible future recommendations, not used for ranking today)
• Bunny.net: application hosting (Magic Containers), CDN, S3-compatible object storage
• Redis (Upstash, TLS) for cross-replica pub/sub
• Multi-provider transactional email abstraction (currently live on Scaleway TEM)
• Per-route rate limiting (stricter on auth endpoints)
• GitHub Actions CI/CD: separate backend/frontend pipelines, Docker builds, auto-deploy to Bunny on push to main
• Health check endpoint

Not yet built / explicitly deferred: DSA Statement-of-Reasons user notifications & appeal flow, UrhDaG rights-holder copyright portal, birth-date/16+ age verification enforcement, uptime monitoring/alerting, notifications for follows-of-a-reply/DM-specific push beyond what's listed above, ClickHouse-based ranking (data collection foundation exists, ranking layer doesn't).