Aether: A Real-Time Space Weather Intelligence Platform Combining Stream Processing, Ensemble ML, and RAG-Based Conversational AI

Space weather monitoring is critical to the operation of satellites, power grids, and telecommunication systems. We present Aether, a reproducible local-first platform with real-time data ingestion, event-driven processing, hybrid storage, ensemble forecasting, anomaly detection, and natural language querying. Aether offers sub-second query latency on multi-year time series data with a Redis-ClickHouse hybrid storage model, processes over 850 events per second with Apache Kafka, and supports natural language querying with the Ollama local LLM without any cloud dependencies. We combine traditional LSTM and Prophet forecasting with Isolation Forest and Autoencoder-based ensemble anomaly detection with 87% precision and 82% recall. We evaluate the system with 30 days of operational data with all metrics meeting or exceeding the design specifications.