akka-chat

A real-time corporate chat built on the Akka actor model with event-sourced persistence and a WebSocket transport. Co-workers message each other from a shared contact list, with delivery, online/offline presence and typing indicators. Originally built around 2016 as an exploration of Akka Persistence and actor-based system design.

This is a historical project on a 2016-era stack (Scala 2.11, sbt 0.13, Akka 2.4 / Akka HTTP 10.0, AspectJ + Kamon metrics). It is preserved here as an architecture reference rather than a maintained, build-on-modern-toolchains app. Company-specific names and hosts have been removed.

Layout

Path	What	Stack
backend/	Chat server: WebSocket API, actor-per-conversation, event-sourced journal, admin/CRM token auth.	Scala · Akka · Akka HTTP · Casbah/MongoDB · Kamon
frontend/	Admin web client (contact list, conversations, admin panel).	AngularJS · Bootstrap · Grunt
perf-tests/	Actor-based load generator that simulates many concurrent users/admins.	Scala · Akka

Design notes

• Actor topology — connections, conversations and the per-user dialog list are each modelled as supervised actors (client_connection/, chat_api/). A ConversationSupervisor routes messages to a Conversation actor per dialog; a DialogsListSupervisor maintains each user's dialog list.
• Event sourcing — conversations persist their events through Akka Persistence (MongoDB journal + snapshot store), so state is rebuilt by replaying events.
• Auth — admin/CRM tokens are validated against an external CRM endpoint (CRM_TOKEN_URL); a MockCrmToken is provided for local runs.
• Metrics — Kamon (StatsD) with AspectJ weaving for Akka actor metrics.

Running

Toolchain is pinned with mise:

mise install                 # Java 8, sbt, Node

mise run backend:run         # chat server (WebSocket on :9100)
mise run frontend:serve      # admin client
mise run perf:run            # load test

Backend config lives in backend/application.conf (MongoDB URL, bind host, CRM auth mode — set env = "MOCK" to skip the external CRM).