limen.sfd.rule_based

Canonical docs

Rule-Based Strategy — conditions, temporal modifiers and parameter references
Single-File Decoder

In the research projects

The Python project supplies its own execution signals. The Compiled project allocation scenario includes a declared rule-based primary. In both paths, the execution contract determines fills, costs and availability after a research decision has been formed.

What this package owns

Owns the rule-based decoder building blocks: the RuleBasedConfig schema (with leaf-condition validation) and build_predicate, which compiles a condition tree into an evaluable predicate over a frame. Does not own the SFD wiring that uses these (owned by the rule_based modules in limen.sfd.foundational_sfd / reference_architecture) or experiment execution.

Key entry points

Entry point	Use case	Notes
`RuleBasedConfig`	Validated rule-strategy configuration	Enforces required fields per leaf condition type
`build_predicate`	Compile a condition tree into a callable predicate	Evaluated against a `pl.DataFrame`

Adjacent modules

limen.sfd.reference_architecture.rule_based (RuleBasedStrategy) and limen.sfd.foundational_sfd.rule_based build on these.
limen.data.utils.split_data_to_rule_based_prep_output produces the data_dict variant that keeps the raw columns predicates reference.

Quick orientation

Module or family	Responsibility
`config.py`	RuleBasedConfig + leaf-condition validation
`predicates.py`	build_predicate + operator table

text · README.md

1rule_based/2├── README.md3├── __init__.py4├── _conditions.py5├── _temporal.py6├── common_strategies.py7├── config.py8└── predicates.py

Merged source PR · 62d79072Package layout · Limen 6.0.1 · current package validation · package source

Things to know

Selected Compiled rule pipelines can be reconstructed through Trainer and consumed by Sensor and binary Cohort. See the canonical selected-rule inference contract, including identical-window comparisons and caller-owned SQL causality.

Leaf conditions are validated by type: each of threshold, relative, crossover, slope, and sql_expr has its own required fields, and a missing field raises ValueError.
Logical nodes combine leaves with and, or, and not.
Leaf conditions may specify persistence_n or recency_n; see the canonical temporal-modifier contract.
Predicate values support {param}-style formatting, so thresholds can be parameterized from the manifest.