Limen — 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. Built-in analytics record benchmark, backtest, and cohort artifacts for inspection. The project evolves from Talos, a hyperparameter optimization framework for TensorFlow and Keras.
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
- 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
- Perturbation-based robustness search across feature groups, scalers, and ablation
- Split-first train, validation, and test workflows
- Built-in benchmark, backtest, and parameter diagnostics
- Decoder cohort construction with pluggable selection
- Reproducible runs with checkpointing, resumption, and validated round replay
Run Experiments
Choose the route that fits how you define the experiment:
| Route | Use it when | Start here |
|---|---|---|
| CLI | Define a manifest in YAML and run it without Python orchestration. | First Experiment |
| Python API, backed by MSQ | Run a Python SFD with built-in or custom search strategies, including fully custom preparation and model functions. | Python/MSQ run |
Both routes use the Universal Experiment Loop (UEL) to execute rounds and the Mutable Search Queue (MSQ) to control parameter search. CLI saves an experiment directory automatically; Python callers enable persistence with experiment_dir. Adaptive feedback is optional in both routes.
First Experiment
The first runnable path is a YAML manifest executed through the limen CLI.
- Install the package:
pip install "vaquum-limen[data]"
Supported runtime: Limen requires Python >=3.10,<3.14; package metadata advertises Python 3.10-3.13 on macOS and Linux. The first experiment reads Arrow data, so it needs the data extra. The default install is intentionally light for API surfaces that do not load data. Use vaquum-limen[boosting] for LightGBM/XGBoost models, vaquum-limen[indicators] for TA-Lib comparison tooling, vaquum-limen[stats] for statistical helpers, or vaquum-limen[all] for the full research stack. Security support covers the latest released Limen version through SECURITY.md.
- Scaffold a starter manifest:
limen init logreg-first.yaml --template logreg_binary
- Validate, profile, and dry-run the manifest:
limen validate logreg-first.yaml
limen profile logreg-first.yaml
limen run --dry-run logreg-first.yaml
- Run it:
limen run logreg-first.yaml
- Inspect the result directory printed by the CLI:
- copied YAML manifest
metadata.jsonresults.csvround_data.jsonl
For a Python experiment, follow the Python/MSQ run. Supply an SFD and a search strategy; custom SFDs also receive your input dataframe explicitly.
Risk Boundary
Limen is research software. Benchmark and backtest outputs are not investment advice, trading advice, execution simulation, regulatory approval, or a promise of future performance. Past performance is not predictive, and trading digital assets can result in total loss of capital.
Learn more
- Start with the full documentation hub
- See planned direction in the Roadmap and the security posture in the Security Assurance Case
- Start with the YAML/CLI path in Command-Line Interface and Experiment Manifest
- Use Universal Experiment Loop for the Python/MSQ route and the engine shared with CLI
- Define extension research units in Single-File Decoder and Built-In SFDs
- Strengthen the research surface with Perturbation Strategies for robustness, Fractional Differentiation for stationary features, and Triple-Barrier Method for path-dependent labels
- Analyze results in Log, Benchmark, and Backtest
- Promote finished runs into reusable outputs with Trainer and Cohort
- Contribute through CONTRIBUTING.md and Developer docs
Contributing
Contribution starts through CONTRIBUTING.md, docs changes, or open issues.
Before contributing, start with the Developer docs.
Support
Use SUPPORT.md for support routes and scope boundaries.
Vulnerabilities
Report vulnerabilities privately through GitHub Security Advisories. Do not report vulnerabilities through public issues.
Citations
Published work should cite:
Vaquum Limen [Computer software]. (2026). Retrieved from GitHub.
Machine-readable citation metadata lives in CITATION.cff.
License
1 maintainer
Metadata
1 maintainer
Release files for vaquum-limen 5.26.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| vaquum_limen-5.26.0.tar.gz | 3.6 MB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| vaquum_limen-5.26.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 4.5 MB
Release files / vaquum_limen-5.26.0.tar.gz
| Download URL | vaquum_limen-5.26.0.tar.gz |
|---|---|
| Size | 3.6 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
43789c87186b6a4d0b1ad22f541ca306e0089710aa940877703d763483b1a646
|
|
BLAKE2b-256 checksum How to use checksums |
4920aef6936fd06f59fad851e8227a0a5c84a89d23887ee62ac3bc7a06f0c7bc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.13
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 14, 2026.
Transparency logRelease files / vaquum_limen-5.26.0-py3-none-any.whl
| Download URL | vaquum_limen-5.26.0-py3-none-any.whl |
|---|---|
| Size | 883.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
1c1eac4abe228a1e1c29039758fd4eb1ad9f78e834ee05659a4ce928040c048e
|
|
BLAKE2b-256 checksum How to use checksums |
df6d34a4c3f3a4770b73fbf4f44224c864fbdd013c0df3ea51c9a83abc40d78d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.13
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 14, 2026.
Transparency logSecurity for vaquum-limen 5.26.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for vaquum_limen-5.26.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 14, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
