Skip to main content

🧰 Tech Stack

Speech Coach is built with modern, production-grade technologies focused on scalability, real-time data flow, and developer efficiency.

💻 Backend​

  • FastAPI – Handles Telegram webhooks and RESTful API requests
  • FastStream – Manages audio processing tasks over Redis Streams
  • PostgreSQL – Stores users, questions, feedback, and session history
  • Redis – Used for queuing, caching (planned), and stream coordination
  • MinIO – S3-compatible object storage for voice messages

🔊 Audio & Feedback Processing​

  • OpenAI API –
    • Whisper for speech-to-text
    • GPT-4 for feedback generation and sample answers
  • Bytewax – Stream-based real-time analytics (e.g. sliding windows)
  • Redpanda + Debezium – CDC from PostgreSQL to analytics pipeline
  • ClickHouse – High-performance OLAP database for storing and querying metrics

🛠️ Infrastructure​

  • Docker + docker-compose – Containerized dev environment
  • Ansible – Partial service automation (in progress)
  • Terraform – Infrastructure-as-code for future scalability
  • HAProxy – Planned load balancing for multi-instance deployments
  • Patroni – Planned high availability setup for PostgreSQL

📊 Observability & Monitoring (Planned)​

  • Prometheus + Grafana – For metrics collection and visualization
  • Loki – Centralized logging backend

This stack reflects a balance of performance, resilience, and modern engineering principles — ready for production scaling and future experimentation.