vaquum-limen 2.5.5

Bitcoin-first research and trading platform.

Vaquum Limen turns Bitcoin market data into searchable signals, backtested outcomes, and decoder cohorts.

Limen • What Limen Is Not • Capabilities • First Experiment • Learn More

---

Limen — The Research Engine

Manifest-driven Bitcoin alpha research engine that turns market data into searchable signals, backtested outcomes, and decoder cohorts.

Limen unifies parameter search across machine learning and rule-based strategies, with built-in analytics that show not just what works, but why it works. It evolves from Talos, the hyperparameter optimization framework for TensorFlow and Keras cited in over 1,000 scientific papers with zero breaking bugs in six years.

What Limen Is Not

Limen is not:

• a trade execution system
• a downstream trade decision engine
• a generic multi-asset research platform

In the wider Vaquum architecture, Origo sits upstream as the data layer. Nexus, Praxis, and Veritas sit downstream for decisioning, execution, and oversight.

Capabilities

• Manifest-driven experiment pipelines
• Search across models, rules, features, targets, and hyperparameters
• Extensive built-in indicator and feature library for Bitcoin research
• Support for both machine learning and rule-based strategy research
• Bitcoin-native transforms, scaling, and target construction
• Leakage-safe train, validation, and test workflows
• Built-in backtesting, confusion analytics, and parameter diagnostics
• Decoder cohort construction and regime-diversified model pooling
• Reproducible runs with checkpointing, resumption, and retraining

First Experiment

The fastest first success is a small parameter sweep on the bundled BTC/USDT kline dataset with the built-in logistic-regression decoder.

1. Install the package:

pip install vaquum_limen

2. Load data and run a first experiment:

import limen

historical = limen.HistoricalData()
data = historical.get_spot_klines(kline_size=7200, n_rows=2000)

uel = limen.UniversalExperimentLoop(data=data, sfd=limen.sfd.logreg_binary)

uel.run(
    experiment_name="logreg-first",
    n_permutations=25,
    prep_each_round=True,
)

3. Inspect the core outputs:

• uel.experiment_log for the parameter sweep results
• uel.experiment_confusion_metrics for confusion analytics
• uel.experiment_backtest_results for backtest results

That path is the simplest way to get a real Limen run on your machine without relying on repo-local fixture files. If you want richer run directories, checkpoints, resumability, and stored round artefacts, continue into the UEL documentation below.

Learn More

• Start with the full docs hub in docs/README.md
• Define research units in docs/Single-File-Decoder.md, docs/Built-In-SFDs.md, and docs/Experiment-Manifest.md
• Run experiments in docs/Universal-Experiment-Loop.md and extend the artifact-rich path through docs/Advanced-Search.md and docs/Reducers-And-Feedback.md
• Analyze results in docs/Log.md, docs/Benchmark.md, and docs/Backtest.md
• Understand the model layer in docs/Reference-Architecture.md and the helper layer in docs/Utilities.md
• Promote finished runs into reusable outputs with docs/Trainer.md and docs/Regime-Diversified-Opinion-Pools.md
• Contribute through docs/Developer/README.md

Contributing

The simplest way to start contributing is by joining an open discussion, contributing to the docs, or by picking up an open issue.

Before contributing, start with docs/Developer/README.md.

Vulnerabilities

Report vulnerabilities privately through GitHub Security Advisories.

Citations

If you use Limen for published work, please cite:

Vaquum Limen [Computer software]. (2026). Retrieved from https://github.com/Vaquum/Limen.

License

MIT License.