01
Bind stable inputs
Rules reference approved field identities and exact Product Type revisions rather than arbitrary UI labels.
Rule Engine
Rules belong to the product, not scattered app code.
ConfigAI Rule Engine separates stable product data from executable configuration logic. Rules reference reviewed product fields and exact revisions so product changes and logic changes can evolve without losing traceability.
Product rules
Lounge Chair · revision 12
widthNumberrule input
arm_typeOptionrule input
base_typeOptionrule input
collectionReferencecontext
Readable rule
when width > 180 cm then require center support
Product inputs
Compatibility table
Assembly route
Configuration result
Two rule domains
Product decisions and physical assembly are related, but they are not the same problem.
ConfigAI keeps product rules and assembly rules separate so simple option logic does not have to carry scene topology, and physical relationships do not leak into every product field.
Product rules
Availability, compatibility, derived options, required selections, and other decisions based on reviewed Product Type fields.
if width > 180 → require center_support
Assembly rules
Connections, component relationships, placement, structural constraints, and configuration-scene logic for systems made of multiple parts.
middle_module → left_connection + right_connection
ZEN execution
A deterministic engine underneath an agent-friendly authoring layer.
ZEN gives ConfigAI a compact deterministic execution layer for rules. Agents can help author or explain a rule, but execution should still be reproducible from the reviewed definition and inputs.
02
Author visually or structurally
Tables, readable conditions, expressions, and assembly workflows represent the rule without burying it in application code.
03ZEN-backed
Execute deterministically
The engine receives a versioned product/configuration input and returns a reproducible result suitable for tests and later audit.
AI assistance
Let the agent propose the rule. Keep publication human.
An agent can inspect the product definition, source evidence, existing rule sets, and test failures to propose a change. It should not silently publish rule logic that changes a customer's valid configurations.
Agent-assisted authoring
Generate a first rule draft from product semantics
Explain which product fields a rule depends on
Suggest missing test cases
Propose migration when a semantic field revision changes
Review & publish
Run test scenarios
Inspect changed decisions
Review exact product-field bindings
Publish the reviewed rule revision
Test before publish
A rule editor without tests is just another place to hide logic.
The Rule Engine should make changed behavior visible before publication: explicit scenarios, invalid combinations, expected outputs, and later historical replay against saved configurations.
ScenarioExpectedStatus
180 cm · metal basevalidPassed
220 cm · no center supportinvalidPassed
middle module · missing right linkinvalidPassed
legacy configuration v7same resultReview
Deferred commercial layers
Pricing, BOM, and production should reuse the rule foundation, not distort it today.
The current product priority is a stable product definition, 3D workflow, and configuration logic. Commercial and production rule domains can be added later against the same reviewed product identity.
Roadmap
Pricing & quotation rules
Roadmap
BOM derivation
Roadmap
Production / ERP outputs
ConfigAI workspace
Make product logic visible, testable, and versioned.
ConfigAI keeps product semantics and executable rules connected without forcing business logic back into application code.