How tiledesk/tiledesk-server works
Project Type and Purpose
• Tiledesk-server is a monolithic Node.js/Express backend—an open-source conversational platform (live chat + chatbot engine).
• It exposes REST APIs, serves Jade/Pug views for the web widget, maintains WebSocket connections for real-time chat, and plugs into multiple messaging channels (WhatsApp, Telegram, email, etc.).
• It manages data in MongoDB, runs scheduled jobs, and loads “pubmodules” (plugin-style microservices) at runtime.
High-Level Component Breakdown
a. Clients
– Web widget / Admin UI (served from public/ and views/)
– Mobile apps (Android/iOS) via REST/WebSocket
– Third-party channels (Facebook, WhatsApp, Telegram, Chat21) via Webhooks & adapters
b. API Layer (Express)
– app.js + bin/www launch the HTTP + WebSocket server
– routes/ folder: defines REST endpoints & webhook receivers
– middleware/: auth, role checks, rate limits, etc.
c. Business Logic
– services/: orchestrate operations (chatbotService, messageService, emailService, scheduler, etc.)
– event/: event definitions for publish/subscribe flows
– pubmodules/: plugin modules loaded at bootstrap (routing-queue, transformer, voice-twilio, etc.)
– channels/: channel-specific SDKs & connectors (chat21, nativeauth, firebaseConnector, etc.)
d. Data Layer
– models/: Mongoose schemas & constants
– migrations/: database migrations
– config/: database, cache, email, widget, global settings
e. Real-time Engine
– websocket/: pubsub.js, subscription.js, webSocketServer.js
– services/queue & pubmodules/routing-queue: internal message bus for routing & triggers
f. Background Workers
– jobsManager.js & jobs.js: spawn worker processes for async tasks
– services/Scheduler + pubmodules/scheduler/: scheduled jobs (close unresponsive chats, etc.)
– Dockerfile-jobs: container image dedicated to job workers
g. Persistence & External Systems
– MongoDB (DATABASE_URI)
– External APIs: email SMTP, Twilio, Dialogflow/Rasa, Slack, etc.
– File storage via GridFS (services/fileGridFsService)
– Cache layer (config/cache.js + pubmodules/cache)
Relationships & Data Flows
• Client → HTTP Server → Routes → Services → Models → MongoDB
• Services → pubmodules (dynamic loading) → external systems (e.g., Rasa, email, SMS)
• Incoming webhook events (channels/…) → Routes → event emitters → Services → WebSocket → Client
• Services publish internal events → pubmodules/trigger → services or channels
• JobsManager/Scheduler → internal queue → Services → external APIs or DB updates
• WebSocketServer ↔ Clients for real-time message push/pull
Architectural Patterns & Design Principles
– MVC-style separation: routes (controller) / services (business logic) / models (DB).
– Plugin architecture: pubmodules managed by pubmodulesManager; channels folder as first-class connectors.
– Event-driven: event/*.js + pubmodules/trigger create a publish/subscribe mesh.
– Layered setup: HTTP API vs. WebSocket vs. background jobs vs. scheduler.
– Containerized deployment: multiple Dockerfiles for app, jobs, profiler; orchestrated via Docker Compose.
Key Details to Capture
– Exact folder-to-component mapping (e.g., routes/ → API controllers, services/ → business logic)
– Channels vs. pubmodules distinction
– WebSocket server separate flow for real-time updates
– Scheduler & jobs as separate worker processes
– Environment configuration (.env, Dockerfiles) driving container boundaries