Build a simple, polished AI Trader directly inside the existing JellyJelly section of my dashboard.
First, inspect my existing codebase and design, then implement the feature using the simplest, cheapest architecture that fits the project. Make the decisions yourself; don't overengineer it.
The experience should be extremely simple:
Connect and authenticate my crypto wallet once.
Press Start AI to let the agent research markets and trade autonomously within a limited, explicitly authorized allocation.
Keep running in the background until I press Stop.
Show a clean dashboard with current balance, total profit/loss, a performance graph, wins/losses, and recent trading activity.
Make it work for multiple users, with each user's wallet, permissions, trades, and data securely isolated.
Build the actual working feature, not just a mockup. The AI should evaluate opportunities, test strategies, learn from results, and execute trades through supported integrations. Start with paper trading if live trading cannot yet be enabled safely.
Use free and open-source tools wherever practical. Minimize hosting, AI inference, data-provider, and ongoing operating costs. Avoid paid services unless genuinely necessary. Reuse the existing dashboard, database, authentication, and wallet infrastructure instead of rebuilding them.
Security is essential: never ask for seed phrases or private keys. Wallet authentication alone is not trading authorization. Implement secure, revocable trading permissions and independent spending/loss limits. Keep signing authority isolated from the AI, and never claim trades or profits that have not been verified.
Make the interface beautiful, minimal, and mobile-friendly. Handle background execution, failures, and restarts reliably.
Implement everything possible in the repository, run the available tests, and fix errors. Don't stop at a plan. When finished, briefly tell me what works, how to run it, and anything I must configure before live trading is possible.