FinIntel

FinIntel is an Indian market intelligence platform for portfolio monitoring, sector analysis, ticker research, charting, provider-key management, and document risk review. The current build focuses on safer defaults, stronger type safety, modular backend services, and clearer compliance language.

What it does

• Market overview for Nifty, Sensex, India VIX, and USD/INR
• Market desk with equities, currencies, chart modes, alerts, and news tape
• Sector performance heatmap for major Indian indices
• Portfolio valuation and holding-level PnL tracking
• Heuristic research summaries and momentum/behavior signals
• Portfolio-aware research context using holdings, exposure, and available broker cash when read-only sync is enabled
• LLM provider key entry and verification modal for OpenAI-compatible endpoints
• Document clause scanning for common legal and risk keywords
• Compliance, privacy, and risk-disclosure surfaces in the product

Architecture

• Frontend: React, TypeScript, Vite
• Backend: FastAPI with modular routers, schemas, services, and SQLite persistence
• Data: Yahoo Finance market data and local application records
• Discovery: DuckDuckGo search for legal, news, blog, and public commentary inputs
• Session model: short-lived signed application sessions for protected write operations
• Provider keys: encrypted local storage for development verification, with production guidance to use environment secrets or an external secret manager

Security and compliance posture

• Broker credentials are not collected or persisted in the application UI
• Provider keys should be supplied through environment variables or an external secret manager
• LLM provider keys can be entered and verified from the settings modal, but production should still prefer managed secrets
• Zerodha integration, when enabled, is read-only portfolio intelligence only; FinIntel does not place, modify, or cancel orders
• Interactive broker credentials such as request-token exchange secrets are intentionally unsupported in the application
• Protected write routes require a signed session token issued by the backend
• The product includes investment-risk, privacy, and terms disclosures
• The app is an informational research tool, not a brokerage or personalized advisory service
• CI is configured with GitHub Actions for type checks, build verification, and backend tests

Local development

# Backend
pip install -r backend/requirements.txt
python -m backend.main

# Frontend
npm install
npm run dev

Or run both with:

python run.py

Environment

Use .env.example as the baseline. Important variables:

• VITE_API_BASE_URL
• FININTEL_APP_SECRET
• FININTEL_ALLOWED_ORIGINS
• FININTEL_DB_PATH
• FININTEL_ENABLE_BROKER_READONLY
• ZERODHA_API_KEY
• ZERODHA_ACCESS_TOKEN
• LLM provider settings are managed from the app modal and stored encrypted locally for development-only convenience
• ZERODHA_API_SECRET and ZERODHA_REQUEST_TOKEN should remain unset for this product because interactive trading-oriented broker flows are blocked

Project layout

backend/
  api/
  config/
  database/
  middleware/
  schemas/
  security/
  services/
  tests/
src/
  components/
  constants/
  hooks/
  services/
  types/

Validation

• Frontend type check: npx tsc -p tsconfig.app.json --noEmit
• Backend compile check: python -m compileall backend
• Backend tests: python -m unittest backend.tests.test_app
• CI workflow: .github/workflows/ci.yml
• API readiness probe: GET /ready

Known limits

• Market data quality depends on third-party sources and may be delayed
• Real-time here means refreshed frequently from public sources; it is not a paid institutional feed
• Research outputs are heuristic summaries, not analyst-grade recommendations
• Company due diligence uses external search results for legal, news, blog, and public-commentary context and should be manually reviewed
• Broker sync is read-only and intended to enrich analytics with holdings, investment value, cash, and PnL context
• The current auth model is application-session based and still needs a fuller user/account security model for a true production rollout