Insight - AI-Powered Market Research Tool

Insight is a full-stack market research app that turns a company name into a structured, source-grounded research brief. It combines live web search, financial market data, public filings, social sentiment signals, and an LLM-generated analysis into one interactive dashboard.

The app supports both public companies and private startups. Public companies receive financial metrics, stock history, analyst sentiment, earnings data, and SEC event context. Private companies receive startup-focused sections such as funding, investors, growth signals, milestones, and market traction.

Project Preview

Features

• AI-generated company briefs with summary, market size, positioning, SWOT, competitors, confidence score, and cited sources.
• Public/private company detection that switches the report structure based on available market data.
• Public market data including ticker, exchange, sector, revenue, market cap, ratios, earnings, annual financials, stock history, and analyst sentiment.
• Startup research view for funding rounds, investor backing, hiring signals, milestones, customers, partnerships, and traction.
• Source-grounded retrieval using Tavily search results and trusted business, finance, market, and startup domains.
• Competitor enrichment with ticker, logo, market cap, revenue, industry, and overlapping products when available.
• News and media signals from Tavily, Yahoo Finance news, StockTwits, Hacker News, Reddit, and source registries.
• Research Library with saved reports, favorites, and search history persisted in browser local storage.
• Interactive frontend dashboard built with React, Vite, and Recharts.
• PDF export through the browser print flow.
• Rate-limited FastAPI backend with CORS configuration for local and deployed frontends.

Tech Stack

Layer	Tools
Frontend	React, Vite, Recharts, CSS
Backend	FastAPI, Pydantic, Uvicorn, SlowAPI
AI	OpenRouter through the OpenAI-compatible API
Search/RAG	Tavily
Market Data	yfinance, Yahoo Finance endpoints, SEC EDGAR, optional Financial Modeling Prep
Deployment Ready	Vercel frontend config, FastAPI backend

Project Structure

.
├── backend/
│   ├── app/
│   │   ├── main.py          # FastAPI app and routes
│   │   ├── pipeline.py      # AI analysis pipeline
│   │   ├── schemas.py       # API request/response models
│   │   ├── market_data.py   # Financial data, tickers, ratios, history
│   │   ├── search.py        # Tavily search and source formatting
│   │   ├── edgar.py         # SEC EDGAR company/event helpers
│   │   ├── social.py        # StockTwits sentiment helpers
│   │   └── media.py         # Media/public-opinion helpers
│   ├── requirements.txt
│   ├── .env.example
│   └── test_openrouter.py
├── frontend/
│   ├── src/
│   │   ├── App.jsx
│   │   ├── App.css
│   │   └── main.jsx
│   ├── package.json
│   ├── vite.config.js
│   └── vercel.json
├── PRD.md
└── README.md

Prerequisites

• Python 3.10+
• Node.js 18+
• npm
• OpenRouter API key
• Tavily API key

Optional:

• Financial Modeling Prep API key for additional company profile data.

Environment Variables

Create backend/.env from the example file:

cd backend
cp .env.example .env

Required backend variables:

OPENROUTER_API_KEY=your_openrouter_api_key
OPENROUTER_MODEL=openai/gpt-4o-mini
TAVILY_API_KEY=your_tavily_api_key

Optional backend variables:

FMP_API_KEY=your_financial_modeling_prep_key
FRONTEND_URL=https://your-frontend-domain.com

Optional frontend variable:

VITE_BACKEND_URL=http://localhost:8000

If VITE_BACKEND_URL is not set, the frontend defaults to http://localhost:8000.

Local Development

1. Start the backend

cd backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Add your API keys to backend/.env, then run:

uvicorn app.main:app --reload

Backend runs at:

http://localhost:8000

API docs:

http://localhost:8000/docs

2. Start the frontend

In a second terminal:

cd frontend
npm install
npm run dev

Frontend runs at:

http://localhost:5173

API Endpoints

Method	Endpoint	Description
GET	/health	Health check
POST	/analyze	Generate a full market research report for a company
GET	/companies?q=	Search company suggestions/autocomplete
GET	/stock/{ticker}?period=1y	Fetch stock history
GET	/news/{company}?ticker=	Fetch recent company news
GET	/social/{ticker}	Fetch StockTwits social sentiment
GET	/media/{company}	Fetch media and public-opinion overview
GET	/preview?url=	Fetch Open Graph preview data for a source URL
GET	/debug/market?company=	Debug raw market-data lookup

Example /analyze request:

curl -X POST http://localhost:8000/analyze \
  -H "Content-Type: application/json" \
  -d '{"company":"Apple"}'

Useful Commands

Backend:

cd backend
uvicorn app.main:app --reload
python -m app.pipeline "Stripe"
python test_openrouter.py

Frontend:

cd frontend
npm run dev
npm run build
npm run preview
npm run lint

How It Works

1. The user enters a company name in the React frontend.
2. The frontend calls the FastAPI /analyze endpoint.
3. The backend fetches market data and determines whether the company is public, private, or unknown.
4. The pipeline retrieves current source material using Tavily and trusted domains.
5. Public-company reports are enriched with market data, stock history, analyst sentiment, earnings, and SEC filing context.
6. Private-company reports focus on funding, growth, traction, milestones, and investor signals.
7. OpenRouter generates a structured response using function/tool calling.
8. The backend validates the response with Pydantic and returns it to the dashboard.

Deployment Notes

The frontend includes a vercel.json rewrite configuration for client-side routing on Vercel.

For deployment:

• Deploy frontend/ to Vercel or another static frontend host.
• Deploy backend/ to a Python-friendly host such as Render, Fly.io, Railway, or an API server.
• Set VITE_BACKEND_URL in the frontend deployment to the backend URL.
• Set FRONTEND_URL in the backend deployment so CORS allows the deployed frontend.
• Add the backend API keys as environment variables on the backend host.

Notes

• API responses depend on upstream providers, so missing or rate-limited data may reduce report completeness.
• The project is designed for research assistance, not financial advice.
• The browser stores saved reports, favorites, and history locally; there is no user account system or database in the current version.

License

No license file is currently included. Add one before publishing if you want to define how others can use this project.

Product Demo

▶ Watch the full Insight product demo