A personal Bloomberg-style financial terminal built for local use. Real-time market data, portfolio management, macro analytics, regime detection, and AI-powered analysis — all in a keyboard-driven dark UI.
Local-only by design. No authentication layer. Do not expose this to the internet without adding auth.
| View | Key | Description |
|---|---|---|
| Market | 1 |
Watchlist, interactive chart, regime detection (DCC-EWMA), TICK DATA cross-asset board |
| News | 2 |
Financial RSS feeds, Facebook social feed, Polymarket prediction markets |
| Market Movers | 3 |
Global indices table, sector/commodity/bond heatmap treemap |
| Clippings + AI | 4 |
Markdown notes viewer with local Ollama AI (summarize, translate, custom prompt) |
| Macro | 5 |
Yield curve, FRED indicators, Fed policy tracker, country comparison, allocation signals |
| Credit | 6 |
Credit spreads, stress indicators, consumer credit |
| Portfolio | P |
Positions, options, trade log, P&L, backtest, VaR/CVaR risk, paper trading |
| Tail Risk | T |
Tail-risk signals, VIX term structure |
TICK DATA board (Market view, right panel) — six collapsible sections in one scrollable table:
| Section | Rows | Source |
|---|---|---|
| RATES · US | 11 UST tenors (1M → 30Y) | FRED daily constant-maturity |
| RATES · JP | 15 JGB tenors (1Y → 40Y) | Japan MOF daily CSV |
| Americas / EMEA / Asia Pacific | Global equity indices | yfinance |
| FX | 20 currency pairs | yfinance |
Yield rows show the level as a percent (4.680%) but the move in basis points (+1.0bp) — a
percent-change on a yield is meaningless — and only the four tenors with a tradeable proxy
(^IRX ^FVX ^TNX ^TYX) can drive the chart. Crypto is reachable from global search
(BTC-USD), which also has the order-footprint indicator.
Analytics: Stop loss engine (ATR-adaptive + exceedance correlation regime), tail risk signals, DCC-EWMA correlation, sector rotation, country rotation, Fear & Greed index, Black-Scholes Greeks
Market quotes are served through a provider registry with automatic gap-fill failover:
- Yahoo Finance (primary) — full coverage, real-time
- Stooq (fallback) — keyless, end-of-day, US equities + major indices
For a mixed batch (e.g. PTT.BK Thai stocks + AAPL US), each provider fills only the symbols it can price — gaps from the primary are routed to the fallback automatically.
The active provider can be switched from the terminal header chip. Status is visible at GET /api/providers.
| Layer | Tech |
|---|---|
| Frontend | Next.js 16, React 19, TypeScript |
| State | Jotai + TanStack React Query |
| Charts | Recharts + custom candlestick |
| Styling | Tailwind CSS |
| Backend | Python FastAPI (46 routers) |
| Market data | yfinance |
| Macro data | FRED API + Alpha Vantage fallback |
| AI | Local Ollama + Claude API (Anthropic) |
| Prediction markets | Polymarket Gamma API |
| Crypto | Binance aggTrades API |
| Sovereign data | World Bank API |
| Thailand data | Bank of Thailand (BOT) API + SEC Thailand |
| Database | SQLite (backend/portfolio.db) |
| Options | Black-Scholes + Gram-Charlier correction |
- Python 3.11+
- Node.js 20+
- FRED API key (free) — required for macro data
- Other keys optional (see Environment Variables)
git clone https://github.com/YOUR_USERNAME/bloomberg-terminal.git
cd bloomberg-terminalcd backend
pip install -r requirements.txt
cp .env.example .env
# Edit .env — add FRED_API_KEY at minimum# from project root
npm install
cp .env.local.example .env.local
# PYTHON_API_URL=http://localhost:9317 (already set)One command (all platforms):
npm run dev:all # backend + frontend + Ollama
npm run dev:no-ollama # backend + frontend only (if Ollama not installed)Output is color-coded per process — Ctrl+C stops everything at once.
Pulled and nothing changed?
.env.localandbackend/.envare gitignored, so they stay exactly as this machine first set them up while the repo moves on — and an env var always beats the default in the code. Runnpm run doctorto see the drift (stale ports, missing or renamed keys, aPYTHON_API_URLleft exported in your shell) andnpm run doctor:fixto apply it. It also runs automatically beforedev, and after agit pullor branch switch.
On Ctrl+C,
uvicorn --reload's supervisor signals its worker mid-shutdown and asyncio used to print an alarming (but harmless)KeyboardInterrupt/CancelledErrortraceback. Every process still exited 0 and freed its port; the noise is filtered out inbackend/main.pyas of 2026-08-01.
Manual (separate terminals):
# Terminal 1 — backend
cd backend
python -m uvicorn main:app --port 9317 --reload
# Terminal 2 — frontend
npm run dev
# → http://localhost:9318Windows one-click: start.ps1 or start.bat launches both in separate windows.
Copy backend/.env.example → backend/.env. Only FRED_API_KEY is required.
| Variable | Required | Description |
|---|---|---|
FRED_API_KEY |
Yes | Get free key — macro indicators + US Treasury curve |
ANTHROPIC_API_KEY |
No | console.anthropic.com — portfolio AI chat |
ALPHA_VANTAGE_API_KEY |
No | Macro data fallback |
BINANCE_API_KEY |
No | Read-only — crypto order footprint |
OLLAMA_URL |
No | Default http://localhost:11434 — local AI for notes |
CLIPPINGS_DIR |
No | Path to your markdown notes folder |
THESES_DIR |
No | Path to investment theses folder |
FACEBOOK_ACCESS_TOKEN |
No | Facebook Graph API — social news feed |
BOT_API_TOKEN |
No | apportal.bot.or.th — Bank of Thailand |
SEC2_API_KEY |
No | secopendata.sec.or.th — SEC Thailand |
Frontend (.env.local):
PYTHON_API_URL=http://localhost:9317
bloomberg-terminal/
├── backend/
│ ├── main.py # FastAPI app init + router mounter
│ ├── config.py # Constants, indices, env vars
│ ├── db.py # SQLite schema + helpers
│ ├── greeks.py # Black-Scholes + Gram-Charlier
│ ├── routers/ # 46 modular routers
│ │ ├── market.py # Market data + heatmap
│ │ ├── stock.py # Quotes, history, dividends, earnings
│ │ ├── portfolio_v2.py # Portfolio CRUD (accounts, trades, dividends)
│ │ ├── risk.py # VaR, CVaR, stress test, risk parity
│ │ ├── stoploss.py # ATR-adaptive stop loss + regime
│ │ ├── macro.py # FRED macro indicators
│ │ ├── paper_trading.py # Paper trading engine
│ │ ├── rates.py # US Treasury + JGB curves (FRED + MOF)
│ │ └── ... # 38 more routers
│ ├── analytics/ # Quantitative models
│ │ ├── regime_calibration.py # MRS/HMM regime thresholds (Hamilton 1989)
│ │ ├── regime_v2.py # 4-state regime model
│ │ └── sector_*.py # BC / MOM / VAL / factor layers
│ └── .env.example
├── app/
│ └── api/ # Next.js proxy routes → Python backend
├── components/bloomberg/
│ ├── atoms/ # Jotai state atoms
│ ├── hooks/ # useTerminalUI, useMarketData, ...
│ ├── layout/ # Terminal shell, header, navigation
│ └── views/ # 7 view components + portfolio tabs
└── memory/reference/ # Architecture docs, API reference, data shapes
| Key | Action |
|---|---|
1–6, P, T |
Switch view |
/ |
Global symbol search |
Alt+1–Alt+7 |
Switch tab within current view |
Esc |
Close modal / search |
Adaptive stop loss using ATR (Wilder 1978) with exceedance correlation regime detection (Longin & Solnik 2001):
- ATR period adapts to VIX percentile (10–30 bars)
- Regime classified from Spearman correlation on tail returns across SPY/TLT/GLD/BTC
- Regimes: CRISIS (1.5×) → RISK-OFF (2.0×) → TRENDING (2.5×) → DIVERGENT (3.0×)
- Stop =
Price − ATR × dynamic_multiplier − ATR × buffer
GET /api/stoploss/compute?symbols=AAPL,TSLA
GET /api/stoploss/regime
GET /api/stoploss/atr?symbols=SPY
| Model | File | Reference |
|---|---|---|
| DCC-EWMA correlation | routers/tail_risk.py |
Engle (2002) |
| HMM / MRS regime calibration | analytics/regime_calibration.py, analytics/regime_v2.py |
Hamilton (1989); trained via scripts/train_hmm.py |
| Exceedance correlation regime | routers/stoploss.py |
Longin & Solnik (2001) |
| Black-Scholes + Gram-Charlier | greeks.py |
BSM |
| Ledoit-Wolf covariance | routers/risk.py |
Ledoit & Wolf (2004) |
| VaR / CVaR | routers/risk.py |
Historical + Parametric |
The backend must be running before the frontend. If using mobile hotspot, Yahoo Finance requests may be blocked or rate-limited by your carrier's DNS — switch to a regular WiFi connection.
# Verify backend is running
curl http://localhost:9317/api/market-dataThe project uses date-fns@4 alongside react-day-picker@8 which expects date-fns@^3. .npmrc sets legacy-peer-deps=true to resolve this. If you hit the error after a fresh clone, ensure .npmrc is present in the project root.
Volume Profile needs traded volume. Calculated indices (^VIX, ^OVX), Treasury yields (^TNX) and FX
(EURUSD=X) all report volume: 0 from Yahoo — there is no instrument trading behind the number — so the
button is disabled with a tooltip rather than drawing an empty profile. Cash indices like ^GSPC and ^DJI
do carry volume (Yahoo sums the constituents) and work normally. For a volume profile on volatility or
rates, chart a tradeable proxy instead: VIXY/VXX, TLT/IEF, FXE.
symbol_lists (indices, FX pairs, crypto) are seeded automatically from config.py on backend first start. If the market view shows no data after the backend starts, check the backend terminal for seed errors.
cd backend
python -m pytest tests/ -q # 291 tests — greeks, alerts, sync, portfolio, SEC, DCC
# from project root
npm run test:alerts # 44 tests — alert rule AST / normalize / labels
npm run test:chart # 13 tests — chart pane layout
npx tsc --noEmit # TypeScript type check- No authentication — designed for localhost only
backend/.envandbackend/portfolio.dbare gitignored — never commit these- Clippings directory is path-validated against a whitelist (no path traversal)
- Trade symbols are sanitized before any filesystem writes
- All 500 errors return generic messages (no internal paths or stack traces exposed)
MIT — personal/educational use. Not financial advice.