Coding assistant cluster (Scala 3 + Pekko)

Distributed coding assistant demo: a developer client sends programming questions to a frontend router, which load-balances work across backend worker nodes.

Built with Apache Pekko Cluster for horizontal scale (Apache 2.0, production-friendly).

Stack

Layer	Choice
Language	Scala 3.6
JVM	Java 25 (--release 25)
Actors	Apache Pekko 1.1.3 (cluster + Artery)
Packages	com.pratyabhi.assistant.*

Modules

protocol/   CodingQuery, CodingResponse, WorkerRegistration
server/     CodingRouterApp (frontend) + CodingWorkerApp (backend)
client/     DeveloperClient — rotates sample coding prompts

Run

Requires JDK 25 and sbt 1.10+.

1. Router (frontend, port 2551) — start this first; it is the cluster seed node.

sbt "server/runMain com.pratyabhi.assistant.server.router.CodingRouterApp 2551"

2. Workers (backend) — join after the router is up.

sbt "server/runMain com.pratyabhi.assistant.server.worker.CodingWorkerApp 2552"
sbt "server/runMain com.pratyabhi.assistant.server.worker.CodingWorkerApp 0"

3. Developer client

sbt "client/runMain com.pratyabhi.assistant.client.DeveloperClient 127.0.0.1 2551"

The client cycles prompts such as Scala 3 val vs def, async/Future composition, Pekko scaling, refactors, git workflow, tests, and Kubernetes readiness.

Sample protocol

CodingQuery("Explain `val` vs `def` in Scala 3.", "dev-1730-0")
// → CodingResponse(..., answer, sessionId)

Workers return keyword-aware stub answers suitable for demos; swap CodingWorkerActor.stubAnswer for a real LLM client when ready.

References

• Pekko Cluster
• Pekko Scala 3