A real-time macroeconomic intelligence terminal that transforms global data into actionable investment signals.
MacroPulse is a unified macro engine that aggregates global markets, macroeconomic indicators, real-world activity, and news sentiment into a single decision layer — delivering regime detection, dynamic risk scoring, and data-interpolated thesis generation for professional decision-making.
Global macro data is fragmented across incompatible sources, formats, and update cycles:
- Market data (equities, crypto, FX) arrives in real-time ticks
- Macroeconomic indicators (rates, inflation, yield curves) update monthly or daily
- Real-world signals (logistics, aviation activity) are buried in raw APIs
- News sentiment requires NLP interpretation to become tradeable
This fragmentation leads to delayed, inconsistent, and context-free decision-making — forcing analysts to manually reconcile data from 10+ sources before forming a single view.
Macro decisions drive trillions in capital allocation.
Yet most investors operate with fragmented, delayed, or noisy data.
MacroPulse compresses this complexity into a single decision layer.
MacroPulse solves this with a Unified Intelligence Layer that:
- Synchronises all data sources into a single canonical timeline
- Computes a real-time composite Risk Score with dynamic, regime-aware weights
- Detects the current Market Regime (Expansion → Recession) automatically
- Generates a multi-paragraph Macro Thesis from raw data — no human interpretation required
- Surfaces Proactive Alerts when critical macro signal combinations are detected
- Provides Actionable Positioning (Risk-On / Risk-Off / Defensive) as a direct output
The goal is not to show data. The goal is to generate insight.
🔴 CRITICAL — Systemic Stress Detected
VIX: 32 | Yield Curve: -0.75%
→ Historical pattern consistent with recession onset
The following core views provide everything from a high-level macroeconomic panorama to granular market signals.
Central monitoring of leading indicators, yield curves, and global inflation, enabling the automated identification of the current economic regime.
Real-time visualization of core market indicators across Equities, FX, and Crypto, featuring volatility tracking and statistical mean deviation.
Real-world activity monitoring via live iframes for aviation (FlightRadar24) and maritime shipping, providing raw data to validate global growth theses.
AI-driven insight engine that processes real-time News APIs through LLMs to generate sentiment scores and proactive macroeconomic alerts.
- ETL ingests global macro + market + activity data
- Data is normalized and stored locally
- Intelligence Engine computes risk, regime, and signals
- Thesis Engine generates macro interpretation
- Frontend displays actionable insights in real-time
| Feature | Description |
|---|---|
| 🌍 Unified Global Dashboard | All macro, market, and EM data on a single screen |
| Composite 0-100 score with regime-calibrated weights | |
| 🧠 Macro Thesis Generator | Data-interpolated research narrative, not generic labels |
| 🔄 Market Regime Detection | 5-state classifier: Expansion, Recovery, Overheating, Stagflation, Recession |
| 📊 Risk Breakdown Transparency | Full component-level attribution displayed in the UI |
| 🚨 Proactive Alert System | 6 rule-based triggers (VIX spike, yield inversion, BRL stress, etc.) |
| 🛩️ Real-World Activity Tracking | Aviation & maritime data as global trade proxies |
| 📰 Intelligence News Feed | Sentiment + causal implication + HIGH/MEDIUM/LOW impact scoring |
| 🇧🇷 Brazil Macro Intelligence | Selic, IPCA, Ibovespa, USD/BRL, Fear & Greed |
| 🔗 Cross-Asset Correlation Engine | Detects anomalies in historical relationships (e.g. VIX vs. Equities) |
| 🔍 Local Asset Search | Instant fuzzy search across all loaded series — zero API calls |
| 🕰️ Historical Backtesting | Validates Risk Engine against 2008, 2020, and 2022 crises |
| ⏱️ Automated ETL Scheduler | APScheduler runs ETL every 30 min — Celery-migration-ready |
┌─────────────────────────────────────────────────────────────┐
│ DATA SOURCES │
│ Yahoo Finance │ FRED │ BCB (SGS) │ OpenSky │ GDELT/NewsAPI │
└────────────────────────────┬────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ ETL WORKER LAYER │
│ backend/workers/etl_worker.py (APScheduler) │
│ • Fetches all sources in parallel │
│ • Validates & normalises (drops NaN, correct date types) │
│ • Runs Intelligence Engine │
│ • Evaluates alert conditions │
│ • Persists to DB + updates shared cache │
└────────────────────────────┬────────────────────────────────┘
│
┌──────────────┴──────────────┐
▼ ▼
┌─────────────────────┐ ┌─────────────────────────────┐
│ SQLite / PostgreSQL│ │ In-Memory Cache │
│ time_series_data │ │ (Redis-ready interface) │
│ macro_status │ │ backend/api/cache.py │
└─────────────────────┘ └──────────────┬──────────────┘
│
▼
┌─────────────────────────────┐
│ FastAPI — READ ONLY │
│ backend/api/main.py │
│ /intelligence/global │
│ /timeseries/global │
│ /backtest │
└──────────────┬──────────────┘
│
▼
┌─────────────────────────────┐
│ Frontend Terminal UI │
│ Cyberpunk aesthetic │
│ Vanilla JS + Alpine.js │
│ TradingView Charts │
└─────────────────────────────┘
Key architectural principle: The ETL worker is the only layer that touches external APIs and mutates state. The FastAPI layer is strictly read-only — ensuring deterministic, low-latency responses.
| Capability | MacroPulse | Traditional Dashboards |
|---|---|---|
| Data aggregation | ✅ | ✅ |
| Real-time synthesis | ✅ | ❌ |
| Regime detection | ✅ | ❌ |
| Actionable positioning | ✅ | ❌ |
| Automated thesis generation | ✅ | ❌ |
| Component | Technology |
|---|---|
| API Framework | FastAPI (async, OpenAPI docs built-in) |
| ORM | SQLAlchemy |
| Scheduler | APScheduler (Celery-migration-ready) |
| Language | Python 3.11+ |
| Source | Data |
|---|---|
Yahoo Finance (yfinance) |
Equities, crypto, FX, volatility |
FRED (fredapi) |
Fed Funds rate, CPI, T10Y2Y yield spread |
| Banco Central do Brasil | Selic rate, IPCA inflation |
| OpenSky Network | Real-time aviation activity |
| GDELT / NewsAPI | Global news sentiment |
| Alternative.me | Crypto Fear & Greed Index |
| Component | Technology |
|---|---|
| Core | Vanilla JS (ES6+), HTML5, CSS3 |
| Reactivity | Alpine.js (lightweight state, no build step) |
| Charting | TradingView Lightweight Charts v4 |
| Icons | Lucide Icons |
| Fonts | VT323, Geist Mono, Inter (Google Fonts) |
| Environment | Database |
|---|---|
| Development | SQLite (zero-config, local) |
| Production | PostgreSQL (swap DATABASE_URL in .env) |
The ETL worker (backend/workers/etl_worker.py) is the single source of truth for data mutation. The API layer (backend/api/main.py) is 100% read-only, serving from cache. This enables independent scaling and eliminates race conditions.
ETL Worker → DB + Cache ← FastAPI (read-only)
Risk Score weights are not fixed. They adapt based on the detected market regime:
| Regime | VIX | Macro | Sentiment | Activity |
|---|---|---|---|---|
| Expansion | 25% | 20% | 30% | 25% |
| Recession | 50% | 35% | 10% | 5% |
| Overheating | 30% | 40% | 20% | 10% |
| Stagflation | 35% | 30% | 20% | 15% |
| Recovery | 20% | 20% | 25% | 35% |
In a Recession, VIX and Macro dominate. In Recovery, physical activity is the leading signal. This mirrors how professional macro desks actually reprice risk.
The asset search bar operates entirely on the locally cached time-series data — zero extra API calls. Sub-1ms results, instant navigation, and resilience to network outages.
The BCB delivers Selic as a daily rate. MacroPulse automatically annualises: ((1 + r/100)^252 - 1) × 100. This ensures all rates are comparable and human-readable for professional users.
build_macro_thesis() is deterministic — no external AI calls, no latency, no hallucinations. But the function signature is designed for a drop-in LLM replacement: pass the structured data dict to any LLM API and return its output. Interpretability is maintained.
"Liquidity tightening combined with declining aviation activity suggests early-stage industrial slowdown, despite stable equity markets. Risk remains moderate but skewed to the downside."
Multi-source ETL inevitably produces sparse series (weekends, holidays, different update cadences). MacroPulse drops all NaN rows before DB insert and logs the count — preventing NOT NULL constraint failures silently.
- Data without context is noise
- Indicators without synthesis are useless
- Speed without structure leads to bad decisions
MacroPulse exists to solve all three.
The composite Risk Score (0–100) aggregates four components, each scaled to 0–100 before weighting:
Risk Score = (VIX_score × w_vix) + (Yield_score × w_yield)
+ (Sentiment_score × w_sent) + (Activity_score × w_act)
| Component | Raw Signal | Risk Logic |
|---|---|---|
| Volatility (VIX) | CBOE VIX Index | Higher VIX → higher risk. Capped at VIX=50 → 100 |
| Economic Inversion | T10Y–T2Y spread | Negative spread → inversion → higher risk |
| News Sentiment | Bullish % of articles | Low bullish % → higher risk |
| Physical Activity | Flight/vessel counts | Lower activity → higher supply chain risk |
Two-factor classifier — S&P 500 momentum × VIX level:
| Regime | VIX | S&P 500 Return | Positioning |
|---|---|---|---|
| Expansion | < 15 | Positive | 🟢 Risk-On — overweight equities, crypto |
| Recovery | < 20 | Negative | 🔵 Accumulation — value stocks, high-yield |
| Overheating | > 20 | Positive | 🟡 Cautious — commodities, inflation hedge |
| Stagflation | > 20 | Negative | 🟠 Hedge — gold, TIPS, real assets |
| Recession | ≥ 25 | Negative | 🔴 Risk-Off — Treasuries, USD cash |
Six rule-based triggers, evaluated on every ETL run:
| Alert | Condition | Severity |
|---|---|---|
| Systemic Stress | VIX > 30 AND Yield Curve < 0 | 🔴 CRITICAL |
| High Risk Score | Risk Score > 65 | 🟡 WARNING |
| Recession Regime | Regime = Recession | 🟡 WARNING |
| Deep Yield Inversion | T10Y2Y < −0.50% | 🟡 WARNING |
| BRL Weakness | USD/BRL > 5.80 | 🔵 INFO |
| Stagflation Detected | Regime = Stagflation | 🔵 INFO |
| Metric | Value |
|---|---|
| API Response Time | < 50ms (served from cache) |
| ETL Cycle | Every 30 minutes (configurable) |
| Data Sources | 6 external providers |
| Asset Coverage | 12+ series across 4 asset classes |
| Historical Depth | 2005–present (after backfill) |
| Alert Conditions | 6 rule-based triggers |
| Risk Score | 0–100 composite with full breakdown |
| DB Reliability | NaN-safe insert with logged drops |
- Python 3.11+
- Git
git clone https://github.com/your-username/MarcopPulse.git
cd MacroPulse
python -m venv .venv
.venv\Scripts\activate # Windows
# source .venv/bin/activate # macOS/Linux
pip install -r requirements.txt
uvicorn backend.api.main:app --reloadCreate a .env file in the project root:
# Required for news intelligence (get free key at newsapi.org)
NEWS_API_KEY=your_newsapi_key
# Required for US macro data (get free key at fred.stlouisfed.org)
FRED_API_KEY=your_fred_key
# Optional — for real aviation data (opensky-network.org)
OPENSKY_USERNAME=your_username
OPENSKY_PASSWORD=your_passwordNote: The platform works without all keys. Missing keys gracefully fall back to alternative sources or cached data.
uvicorn backend.api.main:app --reloadThe ETL scheduler starts automatically and runs every 30 minutes. On first startup, the cache is warmed from the database.
# POST request — fills DB with 2005-present data (covers 2008 & 2020 crises)
curl -X POST http://localhost:8000/etl/historical-ingestOpen frontend/index.html in your browser via Live Server (VS Code) or any static server.
# Or use Python's built-in server
python -m http.server 8080 --directory frontend
# Visit: http://localhost:8080# Runs immediately in background — doesn't block response
curl -X POST http://localhost:8000/etl/runAfter ingesting historical data, validate the engine against known crises:
curl http://localhost:8000/backtestExpected response:
{
"summary": {
"total_events": 4,
"passed": 4,
"accuracy_pct": 100.0
},
"results": [
{
"event": "2008 Financial Crisis",
"verdict": "PASS",
"avg_risk_score": 71.4,
"peak_risk_score": 89.2
},
{
"event": "2020 COVID-19 Crash",
"verdict": "PASS",
"avg_risk_score": 68.1
}
]
}# Intelligence snapshot
curl http://localhost:8000/intelligence/global
# Time-series data
curl http://localhost:8000/timeseries/global
# Swagger docs
open http://localhost:8000/docsMacroPulse validates its Risk Engine against four historical reference events:
| Event | Period | Expected Signal |
|---|---|---|
| 2008 Financial Crisis | Sep 2008 – Mar 2009 | Risk Score ≥ 60 · Recession Regime |
| 2020 COVID-19 Crash | Feb 2020 – Apr 2020 | Risk Score ≥ 65 · Recession Regime |
| 2022 Fed Rate Shock | Mar 2022 – Oct 2022 | Risk Score ≥ 50 · Overheating Regime |
| 2019 Bull Market | Jan 2019 – Dec 2019 | Risk Score ≤ 35 · Expansion Regime |
The backtester runs the full calculate_global_risk_index() pipeline on each historical day and scores the result against expected thresholds — ensuring the model has genuine predictive power, not just retrospective curve-fitting.
- Unified Global ETL pipeline
- Dynamic regime-calibrated Risk Score
- 5-state Market Regime Detection
- Data-interpolated Macro Thesis Engine
- Proactive Alert System (6 conditions)
- APScheduler background ETL (Celery-ready)
- Cross-asset correlation anomaly detection
- Historical ingestion (2005–present)
- Backtesting engine (4 crisis events)
- Alpine.js reactive UI
- News feed with HIGH/MEDIUM/LOW impact scoring
- Local asset search (instant, zero-latency)
- Brazil Macro integration (BCB API)
- LLM Integration — OpenAI/Gemini for generative research reports (interface already defined)
- Celery + Redis — Production task queue migration (migration path documented in code)
- PostgreSQL — Production database (one config change)
- Risk Score Sparkline — 7-day trend mini-chart
- WebSocket Real-time Feed — Push updates instead of polling
- User Watchlists — Saved asset sets per user
- Export to PDF — Research report download
| User | Use Case |
|---|---|
| Macro Trader | Real-time regime signal to size equity/bond exposure |
| Quantitative Researcher | Cross-asset correlation data and risk factor decomposition |
| Hedge Fund Analyst | Morning briefing — thesis + positioning from a single terminal |
| Retail Investor | Understand whether markets are risk-on or risk-off without reading 10 reports |
| EM Specialist | Brazil-specific macro metrics (Selic, IPCA, BRL) integrated with global context |
MacroPulse aims to become a next-generation open-source macro intelligence terminal — bridging the gap between raw data and decision-making for professional and retail users alike.
Bloomberg Terminal costs $24,000/year. MacroPulse is not a dashboard.
It is an attempt to democratize macro intelligence — bringing institutional-grade analysis to anyone with a browser.
The thesis is simple: macro intelligence should not require a Bloomberg subscription. With modern APIs, deterministic ML, and a well-designed pipeline, the same quality of market regime analysis is achievable at zero marginal cost.
The next frontier is AI-augmented synthesis — where the heuristic thesis engine is enhanced by LLMs, enabling the platform to generate full-length, contextually-aware research reports indistinguishable from those written by senior macro analysts.
MacroPulse/
├── backend/
│ ├── api/
│ │ ├── main.py # FastAPI — read-only API layer
│ │ ├── cache.py # Shared in-memory cache (Redis-ready)
│ │ └── schemas.py # Pydantic response models
│ ├── db/
│ │ ├── database.py # SQLAlchemy session management
│ │ └── models.py # TimeSeriesData, MacroStatus
│ ├── etl/
│ │ ├── global_macro.py # Yahoo Finance + FRED data fetch
│ │ ├── news_engine.py # GDELT/NewsAPI + impact scoring
│ │ ├── activity_tracking.py # OpenSky aviation data
│ │ ├── historical_ingest.py # 2005-present backfill script
│ │ └── load.py # NaN-safe DB persistence
│ ├── services/
│ │ ├── macro_intelligence.py # Risk Score + Regime + Thesis
│ │ ├── alert_engine.py # 6 proactive alert conditions
│ │ ├── correlation.py # Cross-asset anomaly detection
│ │ └── backtest_engine.py # Historical crisis validation
│ └── workers/
│ ├── etl_worker.py # APScheduler background ETL
│ └── __init__.py
├── frontend/
│ ├── index.html # Terminal UI structure
│ ├── style.css # Cyberpunk theme + animations
│ └── app.js # Fetch, state, charts, search
├── instructions.ai # Living architecture document
├── requirements.txt
├── .env.example
└── README.md
MIT License — free to use, modify, and distribute.
| Provider | Data | Link |
|---|---|---|
| Yahoo Finance | Market prices, crypto, FX | yfinance |
| FRED | US macro indicators | fred.stlouisfed.org |
| Banco Central do Brasil | Selic, IPCA | bcb.gov.br |
| OpenSky Network | Aviation activity | opensky-network.org |
| GDELT Project | Global news | gdeltproject.org |
| NewsAPI | English news | newsapi.org |
| Alternative.me | Fear & Greed Index | alternative.me |
| TradingView | Lightweight Charts | github |
| Alpine.js | Reactive UI | alpinejs.dev |
Built by a macro-obsessed developer who believes insight should be accessible.
MacroPulse — From Data to Decision.