Limen docs

This page routes Limen docs by task.

Limen in one page

Limen is a Bitcoin alpha research engine for turning market data into experiments, logged analytics, backtests, and decoder cohorts. Its default operator path is a YAML manifest run through the CLI; Python APIs remain the extension layer for custom decoder and engine work.

Limen does not perform downstream trade decisioning or execution. In the wider Vaquum architecture, Origo sits upstream as the data layer, while Nexus, Praxis, and Veritas sit downstream for decisioning, execution, and oversight.

Start here

New to Limen

1. Read the product home page
2. Run the End-to-End Workflow
3. Learn the manifest contract in Experiment Manifest
4. Learn how data enters Limen in Historical Data
5. Review outcomes in Log, Benchmark, and Backtest
6. Continue to Universal Experiment Loop for the engine beneath CLI
7. Use Single-File Decoder, Built-In SFDs, and Glossary for Python extension work

Author experiments

1. Start with Experiment Manifest
2. Run manifests through End-to-End Workflow and Command-Line Interface
3. Review the shipped decoder patterns in Built-In SFDs
4. Use Single-File Decoder for custom Python experiment modules
5. Use Indicators, Features, Targets, Transforms, Scalers, Calibration, and Reference Architecture as the reference layer
6. Strengthen the research surface with Perturbation Strategies for robustness, Fractional Differentiation for stationary features, and Triple-Barrier Method for path-dependent labels
7. Adaptive search continues in Advanced Search and Reducers And Feedback
8. Inspect results in Log, Benchmark, and Backtest

Review finished runs

1. Start with Log
2. Compare model behavior in Benchmark
3. Evaluate trading behavior in Backtest
4. Review helper metrics in Standard Metrics Library and Reference Architecture
5. Continue to Trainer and Cohort for downstream promotion

Extend Limen

1. Start with Reference Architecture and Built-In SFDs
2. Continue to Universal Experiment Loop for direct engine integration
3. Continue to Advanced Search for SearchStrategy, ParamDomain, MSQ, and checkpoints
4. Continue to Reducers And Feedback for adaptive interventions
5. Use Utilities for the helper layer rather than the main workflow

Contribute or maintain

1. Start with Developer Guidelines
2. Read the docs contract in Documentation System
3. Use Pruning Strategies for reducer work
4. Use Contributing Foundational SFDs for SFD work
5. Use Technical Debt when assessing accepted known risk
6. Use Making Release and Semantic Versioning for maintenance work
7. Use Roadmap for planned direction and Security Assurance Case for the security posture

How Limen flows

1. Data enters through Historical Data or compatible external OHLC data.
2. Data can be reshaped with Data Bars when threshold bars are the right research surface.
3. Indicators, features, transforms, and scalers define the research surface. Targets define supervised labels; calibration adjusts probabilities and thresholds after model output.
4. An experiment is expressed as an Experiment Manifest and run through the Command-Line Interface by default.
5. Universal Experiment Loop is the engine beneath CLI execution; Single-File Decoder, Built-In SFDs, Advanced Search, and Reducers And Feedback are the Python extension layer.
6. Log, Benchmark, and Backtest explain what happened and why.
7. Trainer turns selected rounds into reusable sensors.
8. Cohort defines selector-driven ensemble inference for multi-member decoder aggregation.
9. Those outputs then move downstream into Nexus and the rest of the Vaquum stack.

Docs map

• Overview: Product Home, this docs hub, Roadmap
• Guides: End-to-End Workflow, Command-Line Interface, Experiment Manifest, Historical Data, Data Bars, Single-File Decoder, Built-In SFDs, Universal Experiment Loop, Advanced Search, Reducers And Feedback, Perturbation Strategies, Fractional Differentiation, Triple-Barrier Method, Log, Benchmark, Backtest, Trainer, Cohort, Conserved Flux Renormalization
• Reference: Glossary, Indicators, Features, Targets, Transforms, Scalers, Calibration, Standard Metrics Library, Reference Architecture, Utilities
• Developer: Developer Guidelines, Documentation System, Pruning Strategies, Writing Docstrings, Contributing Foundational SFDs, Making Release, Packaging, Release Policy, Security Assurance Case, Semantic Versioning, and Technical Debt
• Packages: package READMEs under /limen for data, experiment, inference, sfd, indicators, features, transforms, scalers, metrics, log, cohort, backtest, utils, calibration, cli, targets, and yaml, plus their nested package references

Product boundary

Limen owns

• experiment-oriented data access
• indicator, feature, transform, and scaler composition
• target construction, calibration, and CLI-driven YAML experiment execution
• manifest-driven and custom SFD-based research units
• parameter sweep and experiment logging
• benchmark-style analytics and backtesting
• round reconstruction and cohort construction

Limen does not own

• upstream source-of-truth market data infrastructure
• downstream trade decisioning
• execution and exchange operations
• system-wide oversight and audit

Read next

• For a first real run, continue to End-to-End Workflow, then Command-Line Interface
• For architecture and system boundaries, continue to Trainer and Cohort
• For the extension layer, continue to Built-In SFDs, Reference Architecture, and Advanced Search
• For contributor work, continue to Developer Guidelines