Project Argent: Difference between revisions

=== Agent Architecture ===

The system is composed of several specialized agents working in concert:

'''Observation Agent''': A data-ingestion engine that continuously monitors the public APIs (EIA, FRED, etc.) and computes the real-time EEM Score and its derivatives (e.g., rate of change).

'''Prediction Agent''': A suite of machine learning models (initially LSTMs, later potentially Transformer-based) trained to find correlations between EEM score dynamics and future price movements (ΔP) in target assets.

'''Decision Agent''': A core logic engine, likely based on Reinforcement Learning or a sophisticated rules-based system. It takes inputs from the Prediction Agent and the current portfolio state, calculates the "virtuousness" of a potential trade, and issues a final decision (Buy, Sell, Hold, Short).

'''Action Agent''': An execution module that interfaces with broker APIs (e.g., Alpaca, Interactive Brokers). It is responsible for placing orders, managing positions, and enforcing strict risk management guardrails (e.g., stop-losses, position sizing).

'''Qualitative Agent (Future)''': An LLM-based agent designed to augment the quantitative signals by analyzing news, political events, and market sentiment, providing a layer of human-like intuition.