How wbern/community-assistant works
Project type and purpose
• This is a backend‐only, event-driven microservice built on Akka. It exposes HTTP endpoints to ingest customer emails and drives an agentic workflow using an LLM (OpenAI). Its goal is to collect prospect details for real estate inquiries via a series of AI-driven interactions, persisting the final customer data as events and sending follow-ups when needed.
File structure
• Top-level artifacts:
• pom.xml – single Maven module
• docker-compose.yml – local service dependencies (e.g., persistence store)
• CHAT_INTERFACE_ARCHITECTURE.md, ROADMAP.md, CLAUDE files – docs & CLI helpers
• Source packages under src/main/java:
• com.example.Bootstrap – application entrypoint and actor system setup
• community/… – generic sample of chat/email agent (not core for real estate)
• realestate/… – the real estate customer service agent:
– api/EmailEndpoint.java (Akka HTTP route)
– application/ProspectProcessingWorkflow, CustomerServiceAgent, EmailClient, ClientInfoEntity
– domain/ClientState, ClientEvent, ProspectState
• resources/application.conf – config for HTTP port, persistence plugins, OpenAI key
README for additional insights
• Main Akka components described:
• HTTP Endpoint (/emails)
• Workflow actor (manages state until all info collected)
• Agent (wraps LLM calls and memory; invokes “tools”)
• Event-sourced Entity (persists customer info as a stream of events)
• Timers (schedule/cancel follow-up reminders)