embeddedmarkets

AI agent-based simulation and risk-management tools for professional traders.

This is the pre-alpha developer release of the Embedded Markets Python toolkit. It exposes the planned public API surface for two products currently in private preview:

• LeveeBacktest™ Engine — multiverse backtesting powered by LoRA-fine-tuned AI agents, network contagion, and fat-tailed CGMY shock models.
• emPortfolioAnalyzer™ — regime-specific correlation analysis and hypothetical position-sizing aligned with return targets, drawdown tolerance, and risk budgets.

Pre-alpha status (v0.1.0a1). This release is intended for early evaluation only. Most engine and analyzer methods raise BetaUnavailableError while the private preview matures toward a public beta in Q2 2027. You can load portfolio CSVs, inspect module structure, and import the documented API paths. Join the beta notification list to be notified when features become available.

Why agent-based simulation?

Traditional walk-forward backtests replay a single historical path. Monte Carlo methods extend that by resampling returns but still treat market participants as passive noise. Embedded Markets' AI agent-based models (AABMs) go further: agents learn, imitate, herd, and adjust leverage endogenously, producing emergent feedback loops that conventional backtests miss — including the herding, model monoculture, and liquidity shocks introduced by autonomous AI trading agents.

Installation

Install from PyPI:

bash pip install embeddedmarkets

Dependencies

Requires Python 3.9 or later. The following dependencies are installed automatically:

networkx numpy scipy scikit-survival statsmodels

To install the latest pre-alpha directly from the repository:

bash pip install git+https://github.com/embeddedmarkets/embeddedmarkets.git For development, clone the repo and install in editable mode:

bash git clone https://github.com/embeddedmarkets/embeddedmarkets.git cd embeddedmarkets pip install -e .

What works today

In v0.1.x you can explore package structure and load portfolio holdings from a CSV file. All other engine and analyzer features are stubbed pending the private-preview integration.

python import embeddedmarkets as em

print(em.version)

0.1.0a1

Functional today: parse a holdings CSV into a Portfolio object.

portfolio = em.load_portfolio("holdings.csv") print(portfolio.holdings)

{'AAPL': 100.0, 'MSFT': 50.0, ...}

Calling preview-only features raises an informative error:

python import embeddedmarkets as em

try: sim = em.levee.run(strategy=my_strategy, agents=agents, market_data=data) except em.BetaUnavailableError as exc: print(exc) Planned API The snippets below show how the public API is expected to look once the beta opens. Method signatures and behavior are subject to change.

LeveeBacktest™ Engine python import embeddedmarkets as em

agents = em.levee.agents.from_config( path="configs/agent_archetypes.json", interaction_model="belief_similarity", # alternatives: "transaction_network", "herding_correlation" )

sim = em.levee.run( strategy=my_strategy, agents=agents, market_data=price_series, # sim extends it forward shock_model="cgmy", shock_params="configs/cgmy_baseline.toml", n_universes=100000, horizon="252D", # one trading year random_seed=42, )

summary = sim.tail_risk.summary() summary.plot_drawdown_distribution() emPortfolioAnalyzer™ python portfolio = em.load_portfolio("holdings.csv") # functional today

analyzer = em.PortfolioAnalyzer( portfolio=portfolio, objective="max_sortino_ratio", max_drawdown=0.15, # 15% max drawdown over the horizon stress_distribution=sim.tail_risk )

implied_weights = analyzer.evaluate() # raises BetaUnavailableError in pre-alpha print(implied_weights)

Target audience

Embedded Markets is designed for:

Proprietary trading firms and professional traders Asset managers building tail-risk frameworks Quantitative researchers studying agent-based market dynamics Technically oriented individual investors who want to stress-test strategies under counterfactual regimes

Roadmap

Milestone Status Description Pre-alpha package release Current — v0.1.0a1 Public API surface, portfolio loading, and informative stub errors Private preview engine In development LoRA-fine-tuned agents, network contagion, CGMY shock models Public beta Targeted Q2 2027 Full LeveeBacktest™ Engine and emPortfolioAnalyzer™ functionality Commercial & open-source tiers Post-beta Paid subscriptions plus select open-sourced components Feedback, support, and contributing This package will evolve rapidly as the private preview progresses toward beta. We welcome:

Bug reports and API feedback via GitHub Issues

General inquiries through our contact page Beta sign-ups at embeddedmarkets.com/#pricing Select components will be open-sourced; announcements will be made on the website and in release notes.

Important disclaimers

Embedded Markets is a software provider, not an investment adviser or broker-dealer. The Software and any outputs — including simulations, backtests, risk metrics, correlation analyses, position-sizing calculations, or hypothetical risk-adjusted position sizing suggestions — are for analytical purposes only and do not constitute investment advice.

All simulation and backtest results are hypothetical, have inherent limitations, and may differ materially from live-market outcomes. Do not use the Software's outputs as the sole basis for any trading or investment decision without independent validation by qualified professionals.

License

Use of this pre-alpha release is governed by the license included with the distribution and by the Terms of Use and Privacy Policy of Embedded Markets, Inc.

Embedded Markets, its logo, LeveeBacktest™, emPortfolioAnalyzer™, and related product names are trademarks of Embedded Markets, Inc.