Validation Framework

NetConverter AI uses a comprehensive multi-tier validation system to ensure accurate and reliable translations. Each validation tier builds on the previous, ensuring quality at every step.

Validation Tiers

Tier	Name	Trigger	Purpose
1	Syntax Check	Always	Validates basic syntax structure, brackets, quotes, keywords
2	Structural Check	Always	Validates configuration structure, completeness, and logical consistency
3	Semantic Review	When quality threshold not met	AI-powered semantic validation and quality assessment
4	Retranslation	When additional corrections needed	Retranslates with corrections based on validation feedback

Tier 1: Syntax Validation

Always runs - This is the first line of defense against syntax errors.

What It Validates

• Basic syntax structure (brackets, quotes, parentheses)
• Command/statement format correctness
• Vendor-specific keyword validation
• No obvious parse errors

Tier 2: Structural Validation

Always runs - Validates the logical structure and completeness of the configuration.

What It Validates

• Configuration completeness (all required sections present)
• Object references (addresses, services, zones exist)
• Logical consistency (no conflicting rules)
• Feature completeness (all source features represented)

Tier 3: Semantic Review

Triggers when: Quality score below threshold or critical issues detected

What It Validates

• Semantic correctness of translations
• Vendor-specific best practices
• Security policy accuracy
• Configuration intent preservation

Tier 4: Retranslation

Triggers when: Quality score below threshold after Tier 3

What It Does

• Uses validation feedback to identify issues
• Retranslates problematic sections
• Applies corrections based on review
• Re-validates the corrected configuration

Validation Flow

The validation system follows a progressive approach:

1. Syntax validation runs first on all translations
2. Structural validation ensures completeness
3. If quality thresholds are met, translation is approved
4. If not, semantic review is performed
5. Additional corrections are applied if needed
6. Final validation confirms quality

Quality Scores

Each validation produces quality scores that measure:

• Syntax Score: How well-formed the configuration is
• Structural Score: How complete and consistent the structure is
• Semantic Score: How accurately the intent is preserved
• Overall Score: Combined quality assessment

Validation Results

Each validation returns detailed results including:

• Pass/fail status for each tier
• Quality scores
• List of issues found (with severity levels)
• Recommendations for fixes
• Total validation time

Best Practices

• Review validation results - Check the quality scores and any issues identified
• Address critical issues - Fix any syntax or structural problems before deployment
• Monitor quality scores - Higher scores indicate better translations
• Use validation feedback - The system provides recommendations for improvements