🧠 System Overview

Speech Coach is built with a modular microservices architecture designed for real-time audio processing, AI feedback, and language analytics.

The system is divided into logical layers:

• User & Frontend – Interaction through Telegram, routed via NGINX
• Application Layer – FastAPI + Redis Stream for async task distribution
• Processing Layer – Workers handle transcription and feedback via OpenAI
• Storage Layer – PostgreSQL for data, MinIO for audio
• Analytics Layer – CDC via Debezium → Redpanda → Bytewax for streaming
• Monitoring & Visualization (Planned) – Grafana, Loki, Prometheus, ClickHouse

This architecture supports future scaling via FastAPI/Worker clusters, HA PostgreSQL, and a frontend dashboard.

For deeper insight into component communication, see Data Flow.