Rules Library
The Rule Library and Validation Engine is the foundational intelligence layer of Redline. It provides a centralized repository for creating and managing reusable detection rules, using a two-gate validation system that combines semantic matching with business logic validation.
Rules are party-neutral building blocks. Clause-linked rules leverage the clause library for semantic matching, while document rules enable document-level detection without clause dependencies. The two-gate architecture ensures precise, explainable recommendations by first identifying relevant clauses from the document, comparing it with your standard position (Gate 1) if the rule is associated with a clause from the Clause Libary then assessing favourability through configurable validation methods (Gate 2).
The Rule Library and Validation Engine's key capabilities include the following:
Centralized Rule Management: Single location to create, view, search, filter, and manage all detection rules with usage tracking across playbooks
Clause-Linked Rules: Rules associated with clause library entries that leverage semantic matching (Gate 1) with configurable thresholds (Flexible/Moderate/Strict) before applying validation logic
Document Rules: Document-level rules operating without clause library dependency for validating the conditions including multiple clauses and fields.
Two-Gate Validation Engine: Gate 1 (Semantic Matching) determines clause relevance; Gate 2 (Business Validation) assesses favorability and unfavorability based on defined natural language prompt.
Reusability: Same rule can be attached to multiple playbooks, ensuring consistent analysis across document types and party perspectives
| Clause-level rules | Document-level rules |
|---|---|
| Validation scope is restricted to a single clause (for example, Payment Terms or Limitation of Liability). | Document-level detection, flagging conditions like missing clauses, value thresholds, or prohibited terms without requiring a clause library match. |
| Used when the conditions being checked exist entirely within one clause. | Used when the conditions being checked involve relationships between different parts of the document (for example, contract value in one section and governing law in another). |
| Can reference clauses from your standard clause library, or custom clauses defined specifically for third-party documents. | Validation scope spans the entire document, across multiple clauses or fields. |
Match Range and Favorable/Unfavorable Conditions
Each clause-level rule includes two configuration elements that guide how the agent evaluates a clause:
Flexible: Captures broader variations in language and structure. Requires 0% to 60% match to sample language
Moderate: Balanced approach for most use cases. Requires 61% to 84% match to sample language
Strict: Requires above 85% match to sample language
Favorable and Unfavorable Conditions
Favorable position defines the language or terms in a clause that are acceptable per your organization's standards. These are extracted from the associated standard or sample clause text, using the Scan Clause Text feature, or entered manually. Favorable conditions are used for identification and redline recommendations only. They do not contribute to risk classification.
Risk conditions define the language or terms that indicate risk.
High Risk (deal-breakers requiring immediate escalation)
Medium Risk (negotiation points needing attention)
Low Risk (minor concerns and preferences).
If a risk condition from any of the above tier is triggered, Redline generates a redline recommendation for that clause. The overall risk category for the rule is determined by the highest tier triggered.
If no unfavorable conditions are found in a clause, the clause is classified as No Risk.
If no risk item is defined for a rule, the clause is classified as Undefined.