Advanced Capabilities

FluxRules supports much more than basic condition matching. This page maps advanced capabilities to runnable examples so each claim is backed by executable code.

Capability Map

Capability	What it enables	Primary examples
Multiple engine implementations	Choose behavior and performance profile by workload	09_custom_engines.py, 10_pluggable_engines.py, 23_streaming_vs_stateless.py
Stateful evaluation	Keep and reuse context across evaluations	06_working_memory.py, 20_streaming_mode.py
Rules as data	Author and load rules from YAML	07_yaml_rules.py
Extensible actions	Register and execute custom actions	15_action_system.py, 21_action_registry.py
Complex decision modeling	Express nested logic and multi-fact scenarios	13_complex_rules.py, 24_cross_fact_joins.py, 25_nested_facts.py
Fact normalization pipeline	Transform heterogeneous inputs before evaluation	26_fact_loader.py, 27_factpipeline.py, 28_factpipeline_tier1.py, 29_custom_transforms.py
API and CLI integration	Embed rule evaluation in service and automation paths	11_cli_usage.py, 12_api_usage.py, 22_real_world_workflow.py

Advanced Evaluation Patterns

Engine mode as a runtime concern

Use different evaluation modes for different workloads. You can keep rule definitions stable while swapping the execution strategy.

• Use PhreakEngine (stateless, default) for high-throughput request-response evaluation.
• Use PhreakEngine(streaming_mode=True) when you need incremental deltas across a session.

See 10_pluggable_engines.py for switching evaluation modes while keeping the same rule intent.

Stateful versus stateless execution

FluxRules supports both one-shot evaluation and stateful flows.

• Stateless: evaluate facts independently.
• Stateful: keep working memory/session context and evaluate incrementally.

See 23_streaming_vs_stateless.py for a side-by-side comparison and 06_working_memory.py for state retention patterns.

Cross-fact and nested data reasoning

You can express nested condition logic and evaluate relationships between multiple facts. Engines evaluate flat key-value facts, so nested input must be pre-flattened into dotted (or otherwise flattened) keys before evaluation; engines do not traverse nested dictionaries at match time.

• Cross-fact joins: 24_cross_fact_joins.py
• Pre-flattened nested facts: 25_nested_facts.py

These patterns are useful in fraud, eligibility, and compliance workflows where single-row checks are not enough.

Extensibility Surface

Custom engines

You can implement custom engine behavior and still integrate with the surrounding FluxRules flow. See 09_custom_engines.py.

Action system and registry

You can define reusable action handlers and register them for rule-triggered behavior. See:

• 15_action_system.py
• 21_action_registry.py

Fact pipeline and transforms

For heterogeneous input data, normalize before evaluation using loader and pipeline stages:

• 26_fact_loader.py
• 27_factpipeline.py
• 28_factpipeline_tier1.py
• 29_custom_transforms.py

Deep-Dive Pages

These pages cover advanced areas in more detail:

• Action Registry
• Fact Pipeline
• Cross-Fact Rules
• Nested Facts
• Validation Framework
• Working Memory
• Custom Engines
• Custom Transforms

Verification Note

All examples listed in this page were executed successfully in repository verification on 2026-07-01.