Sannvi
NAME
sannvi - Sannvi
SYNOPSIS
Robustness-first backtesting SaaS.
SCREENSHOT
DESCRIPTION
Build a trading strategy visually or in Python, run it against US stock and crypto data with realistic fills and costs, and get an Honesty Report that scores how likely the result is real and not luck.
Drag-and-drop blocks and Python both compile to the same execution path, so a visual strategy and its code export behave the same.
STACK
Python, FastAPI, React, TypeScript, Supabase, Stripe, Vercel, Railway
HIGHLIGHTS
- Honesty Report on every run: out-of-sample split, Monte Carlo resampling, a random-trader control, and a buy-and-hold benchmark
- Walk-forward analysis, strategy mining, and NSGA-II genetic search, all with deflated stats so more searching doesn't inflate the score
- Bar replay with manual orders through the real engine, scored against random traders and your own strategy
- Paper trading on a daily schedule. No broker integration and no real orders, ever
- Stripe subscriptions, Supabase auth with row-level security, Sentry, and PostHog