Poor financial decisions are often inspired by malicious actors on social media exploiting their audience and posing as experts. We wanted to allow for more transparency by providing real stats of what would happen if people were to follow said advice.
TradeTruth lets you upload a financial tip in a multitude of convenient formats such as a video file, a link to a clip online, or raw text and get real statistics of how it would perform. Our app runs a Monte Carlo simulation as a well as explicitly mark potential red flags. It outputs a list of warnings for flagged concepts (such as "Guaranteed Return"), alongside a series of visualizations that show the output of the simulation.
Next.js 15 full-stack app with MongoDB for persistence and NextAuth + Google OAuth for auth. Financial tips and video URLs are transcribed via OpenAI Whisper + yt-dlp, then parsed into structured trade objects by GPT-4o-mini. A custom Monte Carlo engine in pure TypeScript bootstraps 2,000 price paths from real historical market data (via Polygon.io) to generate P&L distributions. Mantine UI handles the frontend, and all sensitive logic lives in server-side API routes.
- Clone the repository
- Ensure NodeJS is installed
- Ensure
pnpmis installed - Run
pnpm installin the root directory - Put all the credentials into a
.env.localfile in the root directory
AUTH_SECRETa securely generated secret key for encrypting auth tokensAUTH_GOOGLE_IDandAUTH_GOOGLE_SECRETfor OAuth with GoogleMONGODB_URIfor the databaseOPENAI_API_KEYfor Whisper and ChatGPT use
- Run
pnpm run devand visit [http://localhost:3000] to view the site
