escld

A social media app — posts, comments, likes, follows, moderation — with real-time messaging, WebRTC video/audio calling (mute, camera toggle, screen share, in-call chat, raise hand, and moderator-only call recording), and live streaming (RTMP ingest, HLS delivery via S3/CloudFront).

Built as eight services: a React SPA, a Java/Spring Boot API, two Rust services (WebRTC SFU + messaging, and RTMP live-streaming ingest), three small background workers (transcoding, feed fan-out, trending), and an AWS CDK infrastructure app. See docs/ARCHITECTURE.md for how they fit together.

Quick start

Requires Docker and Docker Compose.

docker compose up -d

First boot takes a few minutes (Elasticsearch and a couple of model/dependency-heavy image builds). Once everything's healthy:

• App: http://localhost:5173
• Backend API: http://localhost:8080
• Calling/messaging: ws-sfu on http://localhost:4000
• Trending API: http://localhost:4100

Sign up, confirm the email code, and you're in. Full local-dev details — what's real vs. emulated, the opt-in monitoring stack, running a single service outside Docker, test commands — are in docs/LOCAL_DEVELOPMENT.md.

Documentation

Doc	Covers
docs/ARCHITECTURE.md	System overview, service inventory, request/data flow, why the services are split the way they are.
docs/BACKEND.md	Spring Boot API — every endpoint, domain model, security, rate limiting, error handling.
docs/FRONTEND.md	React app — every route, auth flow, component/hook inventory.
docs/WS_SFU.md	The Rust messaging + calling service — every Socket.IO event, how call recording actually works.
docs/RTMP.md	The Rust live-streaming ingest service — protocol implementation, the publish→HLS pipeline, S3/CloudFront delivery, stream-key auth.
docs/BACKGROUND_SERVICES.md	The transcode worker, feed worker, and analytics service.
docs/DATA_ANALYSIS_AND_FEED_DESIGN.md	Proposed advanced analytics and personalized-feed architecture, event contracts, ranking model, and delivery roadmap.
docs/DATA_MODEL.md	Postgres schema, all five DynamoDB single-table designs, Elasticsearch, Redis.
docs/AUTH.md	Cognito, JWT validation, role mapping — end to end across all services.
docs/INFRASTRUCTURE.md	All 20 CDK stacks, deployment topology, manually-managed AWS resources.
docs/LOCAL_DEVELOPMENT.md	Running everything locally.

Repository layout

frontend/     React SPA
backend/      Spring Boot API
ws-sfu/       Rust WebRTC SFU + messaging (Socket.IO)
rtmp/         Rust live-streaming RTMP ingest + HLS delivery — see docs/RTMP.md
worker/       ffmpeg transcode worker (SQS-driven)
feed-worker/  Feed fan-out + search indexing worker (SQS-driven)
analytics/    Trending posts/hashtags service (Redis pub/sub)
infra/        AWS CDK — all production infrastructure
aws/          Manually-managed AWS resources (outside CDK) — see docs/INFRASTRUCTURE.md
config/       Local-dev observability config (Prometheus, Grafana, ElasticMQ)
docker-compose.yaml   Full local dev stack

Testing

Each service has its own test suite — see the table at the bottom of docs/LOCAL_DEVELOPMENT.md for exact commands per service (backend, frontend, worker, feed-worker, ws-sfu, infra).

Deploying

Infrastructure is defined in infra/ (AWS CDK). cdk synth is always safe to run and is how infra changes are verified; cdk deploy creates real AWS resources and should never be run casually. See docs/INFRASTRUCTURE.md.