Experiment Manager — Service

Stub. This is the most complex service. Plan to make this the gold-standard service doc — it's the test that the template works.

Overview

Temporal workflows + activities deployed on EC2 (planned move to AWS EKS). Triggered by EventBridge schedules created by the Backend when experiments are scheduled.

Tech stack

• Temporal
• Node.js (TypeScript)
• Postgres (RDS)
• AWS SDK (EventBridge integration)

Sections to fill

• Workflow vs activity boundaries
• Workflow catalog (per-experiment-type workflows)
• Scheduling model (EventBridge ↔ Temporal contract)
• Agent grading workflow specifics
• Split-optimization workflow specifics
• Retries, timeouts, idempotency conventions
• Deployment runbook (today on EC2, future EKS)
• Monitoring (Temporal UI access, alerts)
• Local development (Temporal dev server setup)
• Known issues