A personal AI-driven research tool for Indian retail investors to gain actionable investment insights.
Project Repository: Mosaic-fund-agent
To empower retail investors with institutional-grade analysis by combining live portfolio data, LLM-driven sentiment, risk scoring, and quantitative ML forecasting.
Addressing the challenge of fragmented financial data by centralizing holdings, news, and market signals into a single dashboard. It automates complex analyses (like ETF premium/discount tracking and volatility position sizing), freeing users from manual data engineering so they can make confident, data-driven decisions.
- Portfolio: Live holdings via Zerodha Kite MCP.
- Market: Live ETF iNAV and prices from NSE.
- Commodities: Pre-market signals via COMEX and Gold API.
- Institutional: FII/DII cash and F&O participant OI data.
- News: Global and local news from NewsAPI and GNews.
- Historical: Mutual fund history for backtesting.
- Time-Series: Historical price action and volatility stored in ClickHouse.
- Portfolio Metadata: Sector breakdowns, holding weights, and risk scores.
- Unstructured Text: News headlines processed by LLMs for sentiment.
- Quantitative Signals: 25+ alpha features for LightGBM forecasting.
- Risk Metrics: Regime detection and anomaly residuals.
- Streamlit Data Hub: Web UI for data import, SQL exploration, and visual dashboards.
- Terminal/CLI: Robust command-line interface with 13+ commands for analyzing portfolios, asking questions via LLM, and running imports.
- React Dashboard: Auto-refreshing (every 5 minutes) React-based HTML dashboard generated from JSON reports for a quick overview of risk and sentiment.
- Live portfolio integration (Zerodha).
- LLM-driven sentiment analysis.
- Advanced risk scoring and GARCH-based anomaly detection.
- Dashboard for centralized insights.
For detailed setup and installation instructions, please refer to the Mosaic-fund-agent Repository.