Maple is built as a modern web application, with a deliberate emphasis on determinism and transparency in how results are computed.
The research engine
Backtesting, scoring, and validation in Maple are handled by a deterministic calculation engine — the same strategy and the same historical data always produce the same result. This matters because it means Maple's numbers are reproducible and auditable, not the output of an opaque model that could give a different answer each time it's asked.
Market data
Maple sources historical and live market data from established third-party providers. Requests for data that require an API key are routed through a server-side proxy, so those provider credentials are never exposed in the application a user's browser downloads.
Accounts and storage
Maple uses Supabase (a hosted Postgres platform) for account authentication and for storing signed-in users' saved strategies and feedback, protected by row-level security policies that scope every user's data to their own account. A guest mode is also available that keeps all activity local to a single browser, with nothing sent to a server.
AI assistance
Maple's AI-assisted research (turning a plain-English idea into a structured strategy, and generating plain-language commentary on results) is separate from the deterministic engine that actually computes results — the AI narrates and assists, it does not decide what a backtest's numbers are. Conversational AI chat, where available, is powered by a language model and is clearly labeled as AI-generated commentary, not a computed result.
Hosting
Maple is a standard web application, hosted on modern edge infrastructure, with no proprietary trading connectivity, no order-routing systems, and no brokerage integration of any kind.