System Architecture
Real backend infrastructure — RSS ingestion → 9-agent AI pipeline → personalization → learning loop. Every node reflects actual system state.
Architecture
Ingestion Layer
Personalization Layer
Learning Loop
Agents & Engines
ASMO is not a single LLM call. Each stage is handled by a specialized agent with defined inputs, outputs, and responsibilities.
Event Analysis Agent
Extracts event type, sentiment, confidence and market reasoning from an article.
Market Prediction Agent
Generates directional market forecasts with impact ranges and confidence scores.
Confidence Review Agent
Validates and calibrates prediction confidence using historical accuracy context.
Scenario Intelligence Agent
Builds if/then forward-looking scenarios with trigger and invalidation conditions.
Intelligence Decision Agent
Synthesizes all analysis into a single act / review / hold decision with market bias.
Alert Agent
Generates alert candidates from the intelligence decision for active market signals.
Reflection Agent
Reviews prediction outcomes and decision quality, producing structured reflection notes.
Memory Agent
Synthesizes the full pipeline run into a persistent intelligence memory record for future context.
Portfolio Intelligence Agent
Evaluates portfolio-level risk and impact of the intelligence event across user holdings.