Devzy v2.0.0 Release Notes
Semantic code understanding, graph-based analysis, and iterative review tracking
This release transforms Devzy from a code reviewer into an intelligent codebase-aware AI that better understands architecture and change impact.
New Features
Semantic Code Search
Our AI reviewer now understands more of your codebase context. Semantic indexing helps surface similar patterns across files and improves consistency checks.
Key Capabilities:
- Semantic Search - Find code by meaning, not just text matching
- Incremental Indexing - Only re-indexes changed files for speed
- Pattern Detection - Identifies when new code deviates from existing patterns
- Review Context - Injects relevant similar implementations into reviews
When reviewing a new authentication function, Devzy shows you how auth is implemented elsewhere in your codebase, ensuring consistency.
Code Graph for Architectural Intelligence
Understand dependencies and impact before you merge. A graph-based dependency model now maps relationships across your codebase.
Key Capabilities:
- Architecture Layer Detection - Define layers (UI → Services → Core) and catch violations
- Impact Analysis - See how many files are affected by a change
- Violation Detection - Flag imports that break your architectural rules
- Dependency Mapping - Visualize what imports what
Iterative Review System
Track issue resolution across commits. No more losing track of feedback! Devzy now remembers previous comments and tracks their resolution.
Key Capabilities:
- Resolution Detection - AI determines if issues were addressed
- Progress Tracking - See resolution rate per PR
- Acknowledgment Flow - Get notified when issues are fixed
- Auto-Cleanup - Clean up resolved comments on PR merge
Workflow: Push a fix → Devzy checks if it resolves the issue → Updates status automatically
Comment Optimization Engine
Smarter, cleaner review comments. No more duplicate or redundant comments. AI now consolidates and optimizes feedback.
Key Capabilities:
- Duplicate Detection - Removes redundant comments (internal + against existing PR comments)
- Pattern Consolidation - Groups similar issues (e.g., "Missing error handling" × 5 locations)
- AI Decision Engine - Intelligently decides what to consolidate vs. keep separate
- Smart Formatting - Clean, actionable comment format
Result: 40-60% reduction in comment noise while keeping all critical feedback.
AI Fix Prompt Generator
One-click prompts for your AI assistant. Every critical/high severity issue now includes a ready-to-use prompt for AI assistants.
Key Capabilities:
- Context-Rich Prompts - Includes file dependencies, similar patterns, and impact analysis
- Universal Compatibility - Works with Cursor, ChatGPT, Claude, Copilot, and more
- Collapsible Format - Clean UI in PR comments
- Severity Filtering - Only generates for critical/high issues
Improvements
Frontend Improvements
Severity Level Configuration:
- Filter by Severity - Choose which severity levels trigger comments (critical, high, medium, low)
- Custom Thresholds - Set per-framework severity thresholds
Architecture Settings (New!):
- Visual Layer Editor - Define your architecture layers with drag-and-drop
- Preset Templates - Quick-start with Clean Architecture, Layered, or Hexagonal patterns
- Import Rules - Specify which layers can import from which
- Path Pattern Matching - Map file paths to architectural layers
Compliance Framework:
- Framework Management - Enable/disable compliance frameworks
- Custom Rules - Add your own compliance rules with patterns
- Severity Mapping - Map rules to severity levels
Prompt Improvements
We've refined our review prompts to reduce false positives:
- Design Pattern Awareness - No longer flags getters, facades, or delegation as "duplication"
- Architectural Understanding - Distinguishes abstraction layers from copy-paste
- Smarter Duplicate Detection - Flags actual DRY violations, not composition patterns