StockTrace

Project Name: StockTrace

Tagline: From Market Noise to Market Intelligence

Description

Most stock alert tools tell you what happened — they never tell you why. StockTrace is an AI-powered messaging agent that watches your watchlist around the clock, detects meaningful market and news events, traces the impact through a supply chain knowledge graph, and delivers a plain-English explanation directly into your messages — before you even have time to wonder.

The core insight behind StockTrace: markets are deeply connected, but those connections are invisible to most investors. A conflict in a lithium-mining region in Mali doesn't look like a Tesla problem — until you map the supply chain. StockTrace does that mapping automatically using a Neo4j graph database seeded with real company-supplier-material-region relationships. When a news event or price anomaly is detected, the system queries the graph, scores the impact, and generates a concise, neutral explanation using MiniMax's text generation model (MiniMax-M2.7). That explanation is then delivered into a real iMessage conversation via the Photon SDK — so your agent lives in your chat, not buried in a dashboard.

The backend is built with FastAPI and orchestrates five live data sources: Finnhub and Twelve Data for real-time price quotes and candlestick data, GDELT and NewsAPI for global event and sentiment detection, and MiniMax as the AI reasoning layer. Anomalies are flagged when a stock moves ≥2% or volume hits 2.5x its rolling baseline. From event detection to alert delivery, the full pipeline runs in seconds. The React frontend lets users manage their watchlist and review alert history. This is not a trading bot or a prediction engine — it's a context machine that tells you what's happening in your portfolio and why.

What problem does it solve? Investors are drowning in noise but starving for context. StockTrace cuts through by connecting real-world events to portfolio exposure through a live supply chain graph — something no existing consumer tool does.

How does it work? User adds a stock → backend polls Finnhub + Twelve Data every 60–120s → GDELT + NewsAPI scanned for relevant events → anomaly detector flags unusual price movement or supply chain headlines → Neo4j graph traces the dependency path (e.g. TSLA → battery → lithium → Mali) → impact scorer rates severity → MiniMax-M2.7 generates a 2–4 sentence neutral alert → Photon SDK delivers it into a real iMessage conversation.

Which Minimax APIs were used and why? We integrated MiniMax text generation (MiniMax-M2.7) as the core intelligence layer for every alert. The model receives a structured prompt containing the stock symbol, the triggering headline, the graph dependency path, and the impact score — and returns a concise, jargon-free explanation suitable for any investor. We chose MiniMax for its strong reasoning on structured financial context, its fast response times for real-time alert generation, and its clean API that made deep integration straightforward within the build window.

Photon Bonus Track StockTrace is fully integrated with the Photon SDK. Instead of alerts living on a web dashboard, the agent pushes directly into iMessage — making it feel like a real financial assistant texting you, not a notification from another app. The Photon integration handles alert formatting, delivery, and conversation threading so the experience feels native.

Built With

• ai
• auradb
• data
• fastapi
• finnhub
• gdelt
• javascript
• minimax
• neo4j
• newsapi
• photon
• python
• react
• sdk
• trae
• twelve