An open-source asset valuation and investment intelligence agent for SMEs — DCF, IRR, NPV, Cap Rate, Monte Carlo scenario modeling, and AI-driven investment advisory across real estate, fixed income, and business projects.
NOVA is an open-source net asset and opportunity valuation agent. It registers multi-class asset portfolios, computes rigorous financial metrics (DCF, IRR, NPV, Cap Rate, Payback), models investment scenarios with Monte Carlo simulation, and delivers AI-generated investment narratives — all running locally through a seven-agent LangGraph orchestration layer.
O NOVA foi recentemente expandido (Evolution) para se tornar uma plataforma preditiva autônoma:
- Data Sync Engine: Ingestão autônoma de contas bancárias e contabilidade via integrações com outras plataformas do mercado, eliminando inserção manual de dados.
- Interactive Scenario Engine: WebSockets streamando simulações de Monte Carlo em tempo real para painéis dinâmicos.
- Sentinel Agent: Monitoramento contínuo de background que audita condições macroeconômicas (taxas de juros, mercado imobiliário) usando outras plataformas e dispara alertas de refinanciamento.
- Capital Matchmaker & AI Export Engine: Cruza os resultados das valuations com perfis de crédito SBA 7(a) ou Private Equity e exporta documentos em HTML/PDF no formato 'Pitch Deck'.
Accepts multi-class asset inputs: real estate properties, fixed income instruments, and business projects. Normalizes financial metadata — acquisition cost, current market value, income streams, depreciation schedules. Persists all records to a local SQLite database with zero cloud dependency.
Input: Real estate · fixed income · business project data
Output: Normalized asset records — registered, classified, persisted to SQLite
Runs asset-class-specific financial computations using numpy-financial and scipy. Real estate: Cap Rate, GRM (Gross Rent Multiplier), appreciation modeling. Fixed income: bond yield, duration, convexity. Projects: NPV, IRR, Payback Period, ROI. Enriches computations with live market data via yfinance. Runs Monte Carlo scenario simulations for uncertainty quantification.
Input: Asset portfolio records + market data
Output: DCF · IRR · NPV · Cap Rate · Monte Carlo scenario bounds
LangGraph orchestrates seven specialist agents across asset classes and decision types. Generates natural language investment narratives using Anthropic Claude (or offline rule-based heuristics). Produces comparative decision matrices across asset classes. Scores project viability and models what-if scenarios on demand.
Input: Computed valuations + scenario models
Output: Investment narrative · comparative analysis · project viability score · what-if scenarios
graph TD
A["Asset Input<br/>(Real Estate · Fixed Income · Projects)"] --> B["Asset Registry Agent<br/>app/agents/asset_registry.py"]
SYNC["Data Sync<br/>app/services/data_sync.py"] -.-> B
B --> C[("SQLite Database<br/>nova_assets.db")]
D["yfinance / Outras Plataformas"] --> E
C --> E["Real Estate Agent<br/>app/agents/real_estate.py<br/>Cap Rate · GRM"]
C --> F["Fixed Income Agent<br/>app/agents/fixed_income.py<br/>Yield · Duration"]
C --> G["Project Viability Agent<br/>app/agents/project_viability.py<br/>NPV · IRR · Payback"]
E --> H["LangGraph Orchestrator<br/>app/agents/orchestrator.py"]
F --> H
G --> H
H --> I["Scenario Modeling<br/>app/agents/scenario_modeling.py<br/>Monte Carlo · scipy"]
I -.->|"WebSockets"| M
H --> J["Comparative Decision<br/>app/agents/comparative_decision.py"]
H --> K["AI Advisor<br/>app/agents/ai_advisor.py<br/>Claude LLM / Offline"]
K --> MMATCH["Capital Matchmaker<br/>app/agents/capital_matchmaker.py"]
K --> EXPORT["Export Engine<br/>app/services/export_engine.py"]
H --> L["FastAPI Backend<br/>app/main.py · Port 8004"]
L --> M["REST API /api"]
M --> N["Glassmorphic Dashboard<br/>frontend/ · Chart.js"]
SENTINEL["Sentinel Agent<br/>Background Monitor"] -.->|"Alerts"| C
| Endpoint | Method | Description |
|---|---|---|
/api/assets |
GET / POST |
List or register assets (real estate · fixed income · project) |
/api/analysis |
GET / POST |
Run valuation analysis for a specific asset |
/api/compare |
POST |
Comparative decision matrix across multiple assets |
/api/system |
GET |
System status, AI mode (llm or offline), version |
| Component | Technology |
|---|---|
| Backend | FastAPI 0.115 (Python 3.11+) |
| Agent Orchestration | LangGraph 0.2.39 · LangChain 0.3.7 |
| Financial Math | numpy-financial · scipy · pandas · numpy |
| Market Data | yfinance 0.2.43 |
| AI Narrative | Anthropic Claude (optional) · offline heuristics fallback |
| Database | SQLite (zero-server, local-first) · SQLAlchemy |
| Dashboard | HTML + CSS + JavaScript · Chart.js |
| Testing | pytest · pytest-asyncio · httpx |
Rigorous financial math, not estimates. Every valuation uses deterministic financial formulas (numpy-financial) — DCF cash flows, IRR iteration, bond duration — not LLM-generated numbers. The AI advisor only narrates results it did not compute.
Monte Carlo, not single-point projections. Scenario modeling uses scipy-based Monte Carlo simulation to produce probability distributions over outcomes, not optimistic single-point forecasts.
Seven specialist agents, one orchestrator. LangGraph routes each asset type and decision task to the appropriate specialist agent, enabling modular expansion without architectural changes.
Zero-server dependency. SQLite requires no database server. The full valuation engine starts with python run.py.
NIST AI RMF 1.0 alignment. AI-generated investment narratives are logged with a clear separation from mathematically computed valuations, following NIST principles of transparency, explainability, and human oversight.
NOVA is built for SME owners, CFOs, real estate investors, and financial advisors who need rigorous asset valuation without enterprise software costs.
You do not need a financial engineering background. Register your assets, configure your scenarios, and run python run.py. NOVA handles the math, modeling, and narrative automatically.
git clone https://github.com/afild/NOVA.git
cd NOVA
pip install -r requirements.txt
python run.pyOpen http://localhost:8004/static/index.html in your browser.
LLM Mode:
export ANTHROPIC_API_KEY=your_key_here # Linux/macOS
set ANTHROPIC_API_KEY=your_key_here # Windows
python run.pyOffline Mode (default): All financial computations (DCF, IRR, NPV, Monte Carlo) operate identically. The AI advisor generates structured narratives using deterministic rule-based analysis.
- Python 3.11 or higher
- Git
git clone https://github.com/afild/NOVA.git
cd NOVA
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
python run.pypytest tests/ -v| NIST Function | NOVA Implementation |
|---|---|
| GOVERN | MIT License · computation-first AI narrative architecture · traceable agent decisions |
| MAP | Investment domain scoped to SME asset classes · documented formula assumptions |
| MEASURE | pytest suite · Monte Carlo confidence bounds · IRR convergence validation |
| MANAGE | Offline fallback · deterministic math separated from AI narrative · explainable scenarios |
The AI Advisor receives only computed valuation outputs (numerical results) — never raw asset data or market API responses.
NOVA/
├── app/
│ ├── main.py ← FastAPI app · lifespan · static serving
│ ├── config.py ← Pydantic settings
│ ├── agents/
│ │ ├── orchestrator.py ← LangGraph orchestration across 7 agents
│ │ ├── asset_registry.py ← Asset classification and registration
│ │ ├── real_estate.py ← Cap Rate · GRM · appreciation
│ │ ├── fixed_income.py ← Bond yield · duration · convexity
│ │ ├── project_viability.py ← NPV · IRR · Payback · ROI
│ │ ├── scenario_modeling.py ← Monte Carlo · scipy · uncertainty bounds
│ │ ├── comparative_decision.py ← Cross-class decision matrix
│ │ └── ai_advisor.py ← Investment narrative (Claude + offline)
│ ├── api/
│ │ ├── router.py
│ │ ├── assets.py
│ │ ├── analysis.py
│ │ ├── compare.py
│ │ └── system.py
│ ├── database/
│ │ ├── db_manager.py
│ │ └── schema.sql ← Tables: assets · valuations · scenarios
│ └── services/
│ ├── finance_math.py ← numpy-financial computations
│ └── market_data.py ← yfinance integration
├── docs/
│ └── images/ ← Architecture diagrams
├── frontend/
│ ├── index.html ← Glassmorphic dashboard
│ ├── styles.css
│ └── app.js ← Chart.js · API integration
├── tests/
│ ├── conftest.py
│ ├── test_agents.py
│ ├── test_api.py
│ └── test_math.py
├── .env.example
├── requirements.txt
└── run.py
Areas where contributions are most needed:
- Additional asset classes (private equity, commodities, crypto)
- Integration with real estate MLS data APIs
- Discounted cash flow sensitivity analysis dashboard
- Docker Compose setup for zero-dependency deployment
- Evolution Framework: Integração autônoma de dados e sincronização
- Motor Interativo Monte Carlo (WebSockets)
- Sentinel Agent para monitoramento contínuo em background
- Gerador de Pitch Deck e SBA Loan Annex (HTML)
- Agente Capital Matchmaker
- Remoção de dependências de terceiros nominais no README
- Full valuation pipeline: asset registry, financial computation, scenario modeling, AI advisor
MIT License — free to use, adapt, and redistribute.