Real-time quantitative trading dashboard for B3 (Brazil's exchange) — built for HFT monitoring, technical analysis, and strategy backtesting.
QuantBR is a three-layer system that consumes B3 market data, computes technical indicators and trading signals in real time, and displays everything in a live browser dashboard.
Feed (C++20) ──ZMQ──► Backend (Python/FastAPI) ──WebSocket──► Frontend (React)
│ │
B3 UMDF (UDP) Redis pub/sub
B3 STEP (FIX) PostgreSQL
| Layer | Technology | Purpose |
|---|---|---|
| Feed | C++20, ZMQ, CMake | Consumes B3 UMDF (UDP multicast) + STEP (FIX 4.4); maintains L2 order book; publishes via ZMQ |
| Backend | Python, FastAPI, SQLAlchemy, Redis | Consumes ZMQ, computes indicators/signals, exposes REST + WebSocket API |
| Frontend | React 18, TypeScript, TailwindCSS, Vite | Live dashboard inspired by professional trading terminals |
- Visão Geral — Foreign/bank/retail flow panels, B3 indices (IBOV, IFIX, SMLL, IDIV), relative force gauges, aggression bars, DI futures curve, world markets, commodities, stock tape
- Tape & Order Flow — Live stock tape, flow chart (time series), signals panel, L2 order book, time & sales, order flow delta, volume profile
- Macro & Notícias — DI yield curve (DI1F26–DI1F33), sector heat map, economic calendar, live news feed (Infomoney, Valor Econômico, Reuters)
- QuantLab — Strategy sandbox with AST safety check + 5 s exec timeout; run/backtest; buy/sell/hold signal output
SMA, EMA, RSI (Wilder 14), Bollinger Bands (20, 2σ), Session VWAP, Order Flow Delta, Volume Profile, Book Imbalance
RSI Oversold/Overbought, Bollinger Band touches, VWAP Cross
Three specialist agents in a floating sidebar:
- Quant Developer — Explains indicators, reviews strategy code
- Quant Researcher — Market analysis, academic context, fundamental data
- Quant Operator — Real-time operational guidance, risk management
- JWT (HS256) authentication with 30-min access token + 7-day refresh token
- bcrypt password hashing (cost factor 12)
- RBAC with three roles:
viewer/trader/admin - Structured JSON audit log → PostgreSQL (
AuditLogmodel) - QuantLab AST sandbox — blocks
import,exec,eval, dunders - Rate limiting on AI agents (configurable per-user hourly cap)
See docs/setup.md for full instructions. Here is the minimum to run locally:
# 1. Clone
git clone https://github.com/lbaietti/quantbr.git
cd quantbr
# 2. Backend
cd backend
python3.11 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env # fill SECRET_KEY and DATABASE_URL
alembic upgrade head
uvicorn app.main:app --host 0.0.0.0 --port 8000
# 3. Frontend (separate terminal)
cd frontend
npm install
npm run dev # http://localhost:5173The C++ feed is optional for local development. The backend fetches market data directly from B3's public REST API when the ZMQ feed is not connected.
QuantBR/
├── feed/ # C++20 HFT feed (UMDF + STEP)
│ ├── include/ # Headers: parser, book, zmq publisher
│ ├── src/ # Implementations
│ └── CMakeLists.txt
├── backend/ # Python FastAPI
│ ├── app/
│ │ ├── api/v1/ # REST + WebSocket routes
│ │ ├── db/ # SQLAlchemy async engine + session
│ │ ├── feed/ # ZMQ subscriber
│ │ ├── indicators/ # SMA, EMA, RSI, Bollinger, VWAP, Delta, VolumeProfile
│ │ ├── market_data/ # B3 REST + BCB + news fetchers
│ │ ├── models/ # ORM models (User, Instrument, Trade, AuditLog…)
│ │ ├── quantlab/ # Strategy sandbox (AST check + exec timeout)
│ │ ├── schemas/ # Pydantic v2 schemas
│ │ ├── security/ # JWT, bcrypt, RBAC, audit
│ │ └── signals/ # Signal engine (RSI, BB, VWAP cross)
│ ├── alembic/ # Database migrations
│ ├── tests/
│ │ ├── unit/ # 37 unit tests (indicators, signals, sandbox AST)
│ │ └── integration/ # Auth integration tests
│ └── pyproject.toml
├── frontend/ # React 18 + TypeScript
│ ├── src/
│ │ ├── api/ # Axios client with JWT auto-refresh
│ │ ├── components/ # All panels + AgentSidebar
│ │ ├── hooks/ # useWebSocket, useAuth
│ │ ├── pages/ # LoginPage, DashboardPage (4-tab workspace)
│ │ ├── store/ # Zustand (authStore, marketStore)
│ │ └── types/ # TypeScript types
│ └── package.json
├── docs/ # Architecture, setup, ISO compliance
└── scripts/ # Utility scripts
| Document | Description |
|---|---|
| docs/setup.md | Installation, environment variables, first run |
| docs/architecture.md | System architecture and design decisions |
| docs/feed.md | C++ feed: UMDF parser, order book, ZMQ publisher |
| docs/backend.md | Backend API reference, indicators, signals, agents |
| docs/frontend.md | Frontend components, state management, WebSocket |
| docs/iso-compliance.md | Full ISO 25010 and ISO 27001 compliance mapping |
QuantBR was designed and implemented under ISO 25010 (software product quality) and ISO 27001 (information security management). See docs/iso-compliance.md for a full mapping of controls to code.
| Standard | Areas covered |
|---|---|
| ISO 25010 | Functional suitability, Performance efficiency, Reliability, Security, Maintainability, Testability, Portability |
| ISO 27001 | A.9 Access control, A.10 Cryptography, A.12 Logging & monitoring, A.13 Network security, A.14 Secure development |
QuantBR v1 was developed entirely using Claude Code (Anthropic's AI coding assistant), from initial planning through architecture design, implementation, security audit, and deployment. The development process involved providing domain context — B3 market structure, HFT requirements, ISO standards — and iterating with Claude Code to produce all layers of the system: C++20 feed, Python backend, and React frontend.
This is an intentional and transparent choice. The goal was to explore how far an AI coding assistant can take a complex, multi-layer, standards-compliant system from scratch — and to demonstrate that AI-assisted development can produce production-quality code when the right context and quality criteria are applied.
If you are curious about AI-assisted development at this scale, this project is an open reference.
Contributions are welcome. See CONTRIBUTING.md for guidelines.
This is an experimental project. If you are interested in extending it — new indicators, broker integration, deployment scripts, Docker setup — open an issue or a pull request.
- Docker Compose for full-stack local setup
- Real B3 UMDF feed connection (requires market participant credentials)
- More indicators: ATR, ADX, Stochastic, OBV
- Strategy persistence and backtesting with historical data
- Alerts (email / Telegram) from signal engine
- Multi-broker order routing via STEP (FIX 4.4)
- User management UI (admin panel)
MIT — see LICENSE.
QuantBR is an experimental educational project. It is not financial advice. Use of this software for actual trading is at your own risk. The authors assume no liability for financial losses.