LiquidMind Core Engine/
LiquidMind AI Chat
TL;DR
The LiquidMind AI Chat is the conversational cockpit of FLOW: a context-aware assistant that knows where you are in the platform, reads the safe signals visible on that screen, selects the right tools for that exact moment, and grounds public explanations in the LiquidMind Knowledge Base.
How It Works
- 1
Screen Awareness: The assistant knows whether you are in Terminal, Transactions, Account, MindTrace, Positions, Reasoning, or Academy. It can work with the safe information already present on that screen: the active lesson, visible POIs, selected market, open positions, reasoning event, configuration summary, or performance snapshot.
- 2
Adaptive Tool Loadout: The chat is not one generic bot wearing different labels. Terminal questions can access POI summaries, trader configuration, open positions, and waiting orders. Transactions can filter, aggregate, compare, and analyze trade history. MindTrace can inspect funnel health, reliability, narrative bias, and semantic clusters. Reasoning can explain a specific decision chain. Academy can mentor the active lesson.
- 3
Knowledge Base Grounding: A global read-only documentation search tool is available from every tab. It searches approved LiquidMind docs, FAQ, glossary terms, Academy-linked topics, risk disclaimers, POI concepts, execution modes, agent limitations, and platform behavior so answers stay anchored to documented product truth.
- 4
Personal Trader Awareness: When the current context allows it, the assistant can reference your own account-scoped data: safe trader configuration, risk settings, active assets, active POIs, open positions, waiting orders, trade history, performance metrics, psychological profile, and selected account settings. Your data remains isolated to your account.
- 5
Streaming Tool Trace: When the assistant needs evidence, the widget can show tool-start and tool-done steps before the final answer. This is a visible operations trace: what LiquidMind checked, which observation came back, and how that shaped the response. It is transparency without exposing private prompts, secrets, raw credentials, or hidden system internals.
- 6
Voice and Follow-Up Memory: The same engine supports typed and spoken questions. Voice is transcribed, routed through the same context-aware system, and answered with the current page and recent session-scoped conversation in mind.
- 7
Safety Boundaries: The chat can help, explain, filter, summarize, redirect, and in selected Account contexts update allowed settings. It cannot reveal API secrets, exchange credentials, tokens, verification codes, private prompts, or other users' data, and it does not bypass permissions, approval mode, risk controls, or execution checks.
Think of the AI Chat as the human interface layer of LiquidMind Core. The engine produces signals, scores, guardrail decisions, traces, settings, and performance telemetry; the chat turns that machine state into a conversation you can actually use. It can explain why the Terminal looks biased long, why a setup was rejected, what your current risk configuration implies, how your closed trades are clustering, or how an Academy concept maps into live execution. It is not outside the system looking in. It is wired into the cockpit, reading the same safe instruments you see, calling the right tools for the active tab, and translating the platform's internal state into clear operational guidance.