LiquidMind Core Engine/
AI Reasoning Agent
Concept Overview
When mathematical equations determine a setup is high-probability, control is handed over to the LLM-powered Reasoning Agent. The Agent operates in two core domains: evaluating initial Entry conditions, and performing ongoing Global Position Management for active trades.
LiquidMind's AI completely prevents algorithmic blindness. You can inspect the engine's entire thought process—including every tool call, observation, thesis validation, and logical deduction—by viewing the 'Mind Trace Reasoning' within the application UI. This trace provides complete transparency into why the orchestrator approved an entry or managed an active position.
The ReAct Framework: The AI strictly follows a 'Thought, Action, Observation' iterative loop. It examines the market context, decides which metric needs verification, executes a tool, and then adapts its next thought based on the fresh data.
Entry Analysis: Evaluates the setup across 4 Tool Stages. Stage 1 (Core): HTF Structure, POI validity, MCB momentum. Stage 2 (Detailed): Displacement, Volume Profile, opposing FVGs. Stage 3 (Market Data): CVD, Open Interest, Funding Rates. Stage 4: Raw chart data.
Position Management: The 'Global Market Sentinel' actively safeguards open trades. It continuously fetches fresh structure, momentum, and liquidity data to determine if the original Entry thesis is still valid, recommending to HOLD, DEFEND, REDUCE RISK, or EXIT IMMEDIATELY.
Early Invalidation & Logic Scoring: If severe threats are detected early (e.g., HTF trend misalignment or extreme divergence), the agent kills the trade instantly to save execution time. If valid, it produces detailed component scores out of 100.
Mind Trace Inspector
Step through the agent's internal reasoning for this mock execution. Select an event to inspect its full trace.