StockDrop Is an autonomous AI hedge fund analyst aimed at solving one specific problem: Identifying "Buy the Dip" opportunities without the emotional baggage.
It continuously scans global markets for significant price drops in large-cap companies and deploys a Council of AI Agents to debate the fundamental, technical, and macro rationale before issuing a trade recommendation.
Unlike simple "screeners" or single-prompt GPT wrappers, StockDrop uses a multi-stage, multi-agent architecture to simulate a real investment committee.
Once a stock is identified (dropping >5% in 24h), a team of specialized agents gathers intelligence:
- 🕵️ News Agent: Scans thousands of headlines (Benzinga, Reuters, Seeking Alpha) and reads Earnings Transcripts to answer: Why is the stock down? Is it a structural issue or temporary panic?
- 📈 Technical Agent: Analyzes price action, support levels, RSI, and trend integrity.
- 🌍 Macro/Economics Agent: Triggered automatically if the company has high exposure to the US economy, fetching real-time data from the Federal Reserve (FRED) (Interest Rates, CPI, GDP) to assess headwinds.
- ⚔️ Competitive Landscape Agent: Identifies peers and checks if the drop is company-specific or sector-wide.
- 📰 Seeking Alpha Agent: Digs into specialized investor analysis to find contrarian viewpoints.
- 🧠 Market Sentiment Agent: Uses Google Search Grounding to gauge the real-time "pulse" of the internet and social sentiment.
Two distinct AI personas review the evidence from Phase 1 independent of each other:
- 🐂 The Bull (Value Investor): Constructs the strongest possible argument for buying, focusing on overreactions, value discrepancies, and misunderstood catalysts.
- 🐻 The Bear (Forensic Accountant): Constructs the strongest argument for avoiding the trade, highlighting risks, broken growth stories, and "value traps".
- ⚖️ The Portfolio Manager: Acting as the final decision-maker, this agent reads both arguments and the raw evidence. It is instructed to be risk-averse and verify claims using its own internet search tools. It outputs a final 0-100 Score and a decision:
STRONG BUY/BUY/HOLD/SELL/STRONG SELL
StockDrop doesn't just watch the S&P 500. It monitors:
- 🇺🇸 USA (NYSE, NASDAQ)
- 🇪🇺 Europe (STOXX 600 components)
- 🇨🇳 China (HKSE, Shanghai - Large Caps)
- 🇮🇳 India (Nifty 50 components)
We fuse reputable financial data with cutting-edge AI:
- Markets: TradingView & Alpaca (Real-time Prices & Screener)
- Alternative Data: DefeatBeta (Transcripts & Niche News)
- Economy: Federal Reserve Economic Data (FRED)
- Analysis: Seeking Alpha & Benzinga Pro
- AI Models: Google Gemini 3.5 Flash (Reasoning Core) & Gemini Deep Research
- Search: Google Search Grounding (Real-time fact-checking)
- Python 3.10+
- Google Cloud API Key (for Gemini)
-
Clone the repository:
git clone https://github.com/your-repo/Stock-Tracker.git cd Stock-Tracker -
Set up Virtual Environment:
python3 -m venv venv source venv/bin/activate -
Install Dependencies:
pip install -r requirements.txt
-
Configuration (.env): Create a
.envfile and add your keys:GEMINI_API_KEY=your_key_here ALPACA_API_KEY=optional ALPACA_SECRET_KEY=optional
Start the autonomous loop (scans every 2 hours):
python main.pyOr run the web dashboard to view reports:
uvicorn main:app --reloadView the dashboard at http://localhost:8000.
We analyze the performance of the AI Council's deep_research_verdict by comparing the price at recommendation vs. the price 1 week later.
| Verdict | Count | Mean | Median | Min | Max |
|---|---|---|---|---|---|
| HARD_AVOID | 6 | -15.08% | -3.24% | -99.84% | 19.65% |
| SPECULATIVE_BUY | 34 | 3.53% | 6.45% | -99.36% | 68.01% |
| STRONG_BUY | 3 | 6.29% | -1.69% | -8.94% | 29.49% |
| WAIT_FOR_STABILIZATION | 96 | 3.79% | 1.56% | -39.93% | 65.61% |
THIS SOFTWARE IS FOR EDUCATIONAL AND RESEARCH PURPOSES ONLY. It is NOT financial advice. The "decisions" made by the AI agents are simulations. Trading stocks involves significant risk of capital loss. Do not trade based on these outputs.