TradeSense

AI-powered trading dashboard that generates token signals from Pyth price feeds with clear TP/SL levels, gated by identity where needed, and stored for the user in a searchable history.

TradeSense gives users short, actionable trading signals (buy / sell / hold) for tokens using real-time, high-integrity price feeds from Pyth. Each signal includes calculated take-profit (TP1, TP2), stop-loss (SL), a confidence score, risk-level, and human-readable technical reasoning generated by an LLM. All generated signals are stored on the 0G chain, allowing users to review, audit, and re-run analyses in a verifiable and tamper-proof way. The app emphasizes transparency by displaying raw Pyth data and timestamps alongside every signal.

We built TradeSense as a web dashboard powered by Next.js. Pyth price feeds are pulled via Hermes and updated on-chain before being passed to an LLM, which generates structured trading signals (signal type, TP/SL levels, confidence, reasoning). These signals are then stored on the 0G chain to provide users with a verifiable and tamper-proof history. The frontend is optimized for clarity and speed, while backend serverless functions handle the Pyth fetch, inference, and storage pipeline. A hacky but effective optimization we added was batching the data pull and LLM inference into one flow to minimize latency and deliver signals quickly.