| --- |
| title: SemantaAI |
| emoji: 🌐 |
| colorFrom: teal |
| colorTo: orange |
| sdk: static |
| pinned: true |
| license: mit |
| --- |
| |
| <p align="center"> |
| <h1 align="center">🌐 SemantaAI</h1> |
| <p align="center"><strong>Hybrid Sovereign World Intelligence OS</strong></p> |
| <p align="center"> |
| <a href="https://semanta.xyz">semanta.xyz</a> · |
| <a href="https://github.com/ivs-123/semantaai">GitHub</a> · |
| <a href="https://api-staging.semanta.xyz">API Staging</a> · |
| <a href="https://studio-staging.semanta.xyz">Studio</a> |
| </p> |
| </p> |
| |
| --- |
|
|
| ## What is Semanta? |
|
|
| Semanta treats data as **worlds** — structured, evolving domains — and provides an end-to-end pipeline from raw data ingestion through labeling, synthetic generation, world building, model training, and continuous observation. |
|
|
| **Canon:** v3.7 Final · **First Wedge:** Finance · **Mode:** Results-First |
|
|
| ## Core Capabilities |
|
|
| | Layer | Modules | Description | |
| |-------|---------|-------------| |
| | 🔬 Statistical | KS, Wasserstein, MMD, Energy, KL, Anderson-Darling | Full statistical validation battery | |
| | 📊 Retrospective | Regime detection, structural breaks, tail analysis | Auto-extract data characteristics | |
| | 🔮 Continuation | Seamless branch from endpoint, N scenarios | Time-series continuation engine | |
| | 🧬 Generative | VAE, mixture models, Student-t, SDE, GBM, Heston | Advanced synthetic data | |
| | 🔗 Coherence | Cross-domain FX+credit+equities+insurance | Multi-asset scenario generation | |
| | 📈 Econometrics | Granger causality, ARIMA, GARCH, Hurst, t-copula | Full econometric toolkit | |
| | 🧠 ML/DL | Random Forest, XGBoost, LightGBM, LSTM, Transformer | Real ML model wrappers | |
| | ⚡ Options | Black-Scholes, Greeks, Binomial, Asian, Barrier | Options pricing engine | |
| | 📉 Risk | VaR, CVaR, Stress Testing, Kelly, CPPI, Risk Parity | Enterprise risk management | |
| | 🔁 Backtesting | Walk-forward, Monte Carlo, Regime robustness | Strategy validation | |
| | 🏗️ Infrastructure | Logging, Profiling, CI/CD, Docker, K8s, Notifications | Production-ready | |
|
|
| ## Datasets |
|
|
| | Dataset | Type | Status | |
| |---------|------|--------| |
| | [semantaai-crypto_assets](https://huggingface.co/datasets/SemantaAI/semantaai-crypto_assets) | Crypto OHLCV + labels | Published | |
| | [semantaai-fx-majors](https://huggingface.co/datasets/SemantaAI/semantaai-fx-majors) | FX majors (EUR/USD, GBP/USD, etc.) | Published | |
| | [semantaai-fx-other](https://huggingface.co/datasets/SemantaAI/semantaai-fx-other) | FX crosses | Published | |
| | [semantaai-test-dataset](https://huggingface.co/datasets/SemantaAI/semantaai-test-dataset) | Test/sample data | Published | |
|
|
| ## Quick Start |
|
|
| ```python |
| from core.synthetic import retrospective_analysis, continue_from_endpoint |
| |
| # Load your data |
| prices = [1.08, 1.081, 1.083, ...] # EUR/USD prices |
| |
| # Analyze historical patterns |
| retro = retrospective_analysis(prices) |
| print(f"Regimes detected: {retro['regimes']['count']}") |
| print(f"Annualized vol: {retro['overall']['annualized_vol']}") |
| |
| # Generate future scenarios |
| cont = continue_from_endpoint(prices, n_scenarios=4, n_steps=30) |
| for s in cont['scenarios']: |
| print(f" {s['label']}: total_return={s['total_return']:.2%}") |
| |
| # Run statistical battery |
| from core.synthetic.statistical import run_statistical_battery |
| real = [{"v": prices[i]} for i in range(-100, 0)] |
| syn = [{"v": s['path'][-1]['price']} for s in cont['scenarios']] |
| battery = run_statistical_battery(real, syn, key='v') |
| print(f"Quality: {battery['conclusion']} ({battery['battery_passed']}/{battery['battery_total']})") |
| ``` |
|
|
| ## Stats |
|
|
| ``` |
| Tests: 834 (all passing) |
| Modules: 135 synthetic + 20 infrastructure |
| Functions: 907 |
| LOC: 26,904 |
| Production status: PRODUCTION READY |
| Stubs in own code: 0 |
| ``` |
|
|
| --- |
|
|
| <p align="center"> |
| <strong>Semanta v3.7 Final · semanta.xyz</strong><br> |
| Built with Codex + OpenCode · May 2026 |
| </p> |
|
|
