AI Trader

What this project is

AI Trader is an active build: an AI-assisted trading research cockpit for reviewing market evidence, backtests, paper-risk context, and proposal workflows before any real-world trading decision.

The system is not an execution system. It is not broker-routed. It is not a promise of performance. The useful product shape is a research and review cockpit where market context, strategy notes, backtest metadata, risk views, and proposal history can be inspected before a human makes a decision.

Why I built it

Trading tools become dangerous when they blur research, confidence, and execution. I built AI Trader to slow that loop down: gather evidence, inspect risk, review proposals, and keep the human decision separate from the system's suggestions.

What it proves

AI Trader is meant to prove that market tools can be useful without pretending to trade for the user. The public value is the cockpit pattern: evidence cards, paper-risk context, proposal metadata, audit history, and review-first decision support.

What is already working

• AI Brain cockpit direction for market context, evidence review, and proposal workflows
• Market evidence cards for making research inputs easier to inspect
• Audit history for reviewing what the system surfaced and when
• Paper wallet and risk views for separating simulated context from real-world action
• Strategy and backtest metadata surfaces for understanding how an idea was evaluated
• Proposal-only paper loop that keeps suggested actions in review mode
• Proposal review metadata for tracking why a proposal exists and what context supports it
• Read-only risk preview so risk can be inspected without triggering execution
• Docker, runbook, and testing discipline around the build process

How it is designed

AI Trader combines a browser cockpit, backend services, research metadata, paper-risk views, and AI-assisted synthesis into a review-first workflow. The architecture is intentionally conservative: proposals are treated as review objects, risk preview is read-only, and paper context stays separate from any real-world trading decision.

The build discipline matters as much as the interface. Docker/runbook work, route checks, testable proposal flows, and explicit metadata are part of the credibility because market tools become dangerous when they blur uncertainty, execution, and advice.

Current boundaries

• This is a research and review cockpit, not an execution system.
• It does not route orders to a broker or claim autonomous execution.
• Backtests, paper proposals, and risk previews are decision-support context, not guaranteed outcomes.
• Personal trading P&L proof belongs on the trading proof ledger, not inside the AI Trader product page.
• OpenBB should only be described as working if it is separately verified in the current build.

What I am improving next

I am improving proposal review, risk-preview clarity, backtest metadata, and evidence quality before making stronger claims about decision support.

Proof/assets coming next

The strongest public-safe evidence is product evidence: cockpit screens, evidence-card structure, proposal review metadata, paper-risk views, test/build output, and runbook discipline.

If future screenshots are added here, they should show the system workflow rather than imply that AI Trader generated a profit result.

Proof slots: screenshot pending, architecture diagram pending, workflow demo coming, and proof will be added after verification.