AI Equity Research Platform

AI-native equity research platform for Indian equities (NSE/BSE). Provides institutional-grade research capabilities to retail investors via LLM-powered agents, RAG over financial filings, and automated analysis.

Medium Article

https://medium.com/@sanchiitvijay/meet-minerva-the-ai-native-equity-research-platform-revolutionizing-indian-markets-8f4447aa54b6

Services

Service	Port	Tech
Frontend	5173	React 19, TypeScript, Vite, Recharts
Backend	8001	FastAPI, LangGraph, SQLAlchemy, Qdrant, httpx (FMP, FRED, NewsAPI, Upstox, Kite)
PostgreSQL	5432	Companies, financials, users, portfolios
Redis	6379	Caching, Celery broker
Qdrant	6333	Vector search over filings

Quick Start

docker compose up -d                                          # infrastructure
python3 -m venv backend-ai/.venv && source backend-ai/.venv/bin/activate  # backend virtual env
pip install -r backend-ai/requirements.txt                    # backend dependencies
cp backend-ai/.env.example backend-ai/.env                    # configure backend env
python backend-ai/scripts/seed_db.py                          # seed database
cd backend-ai && python -m uvicorn src.main:app --port 8001 --reload &  # backend
cd ../frontend && npm install && npm run dev                  # frontend

Open http://localhost:5173 and switch to Live mode.

Features

• Iris Chat -- AI research copilot with persistent threads and mode-aware responses
• Company Analysis -- financial statements, ratios, risk flags, theme exposure
• Company Comparison -- side-by-side analysis of 2-5 companies
• Portfolio Management -- holdings tracking, concentration analysis, sector exposure
• News & Sentiment -- market headlines with bullish/bearish classification
• Document Upload -- upload PDFs/PPTs for AI analysis via RAG
• Screening & Discovery -- filter stocks by sector, industry, market cap
• Watchlists -- track companies of interest
• Personalized Alerts -- set rules for price, sentiment, filing triggers
• Timeline -- aggregated feed of filings and news
• Deep Research -- CLI-based multi-agent orchestrator for deep analysis

Documentation

See GETTING_STARTED.md for the full setup guide, feature documentation, API reference, and troubleshooting.

Repository Notes

• Python dependencies are maintained in backend-ai/requirements.txt.
• Operational backend scripts live in backend-ai/scripts/ (seeding, debug helpers, local model setup).

Tech Stack

• LLM: Groq / Ollama / OpenAI / DeepSeek (configurable)
• Agents: LangGraph with Router, 5 specialist agents, and Synthesis
• Database: PostgreSQL 15 with 19 SQLAlchemy models
• Vector Search: Qdrant for semantic search over company filings
• Frontend: React 19 with demo/live data modes
• Data APIs: FMP, FRED, NewsAPI, NewsData.io, Upstox, Kite Connect