Start here
From a research question to a Sensor introduces the decision. Choose Python Research for caller-owned data, preparation, fitting, backtesting and selection, or Compiled Research for a YAML definition executed through the CLI. Either project can be followed independently.
Obtain the matching wheel and recorded inputs from the project page before running an excerpt. The package installation does not install the example bundle. If final observations must remain excluded, choose the protected branch before research; Historical Data explains that first decision.
Read the book
| Decision | Chapter |
|---|---|
| Choose who owns the methodology | Begin the research |
| Establish the observations and their clocks | Historical data |
| Change the bar clock | Data bars |
| Define the information available to a model | Feature design |
| Define the observed outcome | Target design |
| Choose the learner and prediction meaning | Model choice |
| Express a strategy through predicates | Rule research |
| Define preparation, fitting and the objective | Define the experiment |
| Compare controlled changes to the definition | Perturbation and robustness |
| Choose the candidates to evaluate | Search |
| Change a running search deliberately | Reducers and feedback |
| Retain, inspect and continue a run | Run and recover |
| Evaluate successive periods | Walk-forward |
| Interpret predictions as trades | Trading behavior |
| Connect learning stages chronologically | Linked models |
| Interpret scores, alternatives, and selection reports | Research evidence |
| Carry selected fitted behavior forward | Selected behavior |
| Assess the observations reserved earlier | Final assessment |
| Inspect the complete chain | End-to-end workflow |
Reference
The public API index maps every declared export and public class member to its exact source contract. Use topic pages for the decision and semantics; this index closes aliases, methods and ownership.
Use historical source methods and search/recovery contracts for acquisition and queue detail. Use Manifest fields and the CLI for Compiled authoring. Use ordinary Python SFD functions and the UEL API for caller-owned execution. Shared preparation components have their own indicator, feature, target, transform, scaler and calibration contracts.
Log, Benchmark, execution, Trainer and Sensors, Cohort and guarded holdout preserve the exact outputs, evidence requirements and failure behavior. Optional native component integration is appendix material; it is not a Python step in the Compiled journey. Package READMEs identify their public entry points and adjacent ownership. The glossary fixes the terminology.
Developer
Start with Developer home to change, test, review, package, or release Limen. Edit and verify the book covers the reader, prose editor, source contributions, and publication. The documentation contract maintains the connection between projects, displayed examples, tests, merged source, and released wheel. Roadmap describes future direction; current behavior belongs in the book and reference.