Summary

This v1.5.0 release of ReviewForge introduces substantial enhancements, transforming it into a more intelligent and adaptable code review assistant. Key updates include advanced AI review capabilities with teaching personas and multilingual support, alongside a new learning reports feature. The action has been refactored into a Composite Action for improved performance and maintainability. Users will also find more granular control over review rules with presets, custom rule support, and a new strict changes flag, ensuring reviews are tailored to specific project needs.

Breaking Changes

None explicitly identified in the commit messages for this release.

Features

Advanced AI Personas and Multilingual Reviews: Introduced teaching personas, multilingual review capabilities, and detailed learning reports for more comprehensive and educational feedback.
Customizable Review Rules: Added support for review rules presets and custom rules, allowing users to define and apply specific review criteria.
Default Concise Review Rules: Concise review rules are now the default, with an option to disable them.
Strict Changes Flag: Implemented a --strict-changes flag to enable lenient review verdicts based on changes.
Self-Review Workflow: A new workflow for self-review has been added.
Token Usage Tracking: Introduced token usage tracking and enforced English in learning reports.
Composite Action Conversion: The action has been converted to a Composite Action, improving its structure and integration.
Updated Default Gemini Model: The default Gemini model has been updated to 2.5-flash for enhanced AI performance.

Improvements

Reliability: Set INCREMENTAL to false by default for better reliability in review processes.
Review Configuration: Merged and forced full review configuration for thorough verification.
Professional Comment Format: Refactored AI review summaries to use professional headers and separators, and synced the professional comment format.
CI/CD Standardization: Standardized CI tooling with lefthook and goreleaser, and streamlined workflows by removing redundant reusable workflows.
Model Fallback: Added a fallback to gemini-2.0-flash for review verification.
Concise Review Rules: Strengthened concise review rules to override the base prompt and require summary sections (core changes, concerns, verdict).
Detailed Logging: Added detailed info logs for review configuration for better debugging and insight.
Persona Tone: Improved the Eli persona's tone.
Cleanup: Removed the obsolete Dockerfile after conversion to a composite action.

Fixes

Action Binary Name: Updated the action binary name to match goreleaser output.
Intentional Bug: Fixed an intentional bug introduced for testing purposes.
Error Handling: Improved GetPRDetails error handling.
Model Names: Corrected model names to use gemini-1.5-flash consistently.
Composite Action Issues: Fixed issues with the composite action, including explicitly passing repository and PR numbers, and resolving download problems with curl.
Review Fallback Order: Corrected the review fallback order.
Own-PR Review Fallback: Fixed the 422 fallback mechanism to downgrade event types on own-PR reviews.

Security

No specific security updates were identified in the commit messages.

Documentation

Updated README.md with the main reference, a new logo banner, and details about new features like teaching personas, multilingual reviews, learning reports, review rules presets, custom rules, strict changes flag, and concise review rules defaults.
Updated documentation related to the conversion to a Composite Action.

Updated (dependencies)

Standardized CI tooling with lefthook and goreleaser.
Removed Dockerfile as it's no longer needed after composite action conversion.

Technical Details

chore: remove reusable workflow and pass github-token to releases (#7) (5b144e4)
fix: update action binary name to match goreleaser output (#6) (2756e2c)
chore: standardize CI tooling with lefthook, goreleaser, and reusable workflows (#5) (b5fb174)
Update README.md with main reference (#4) (3f78dc3)
Set INCREMENTAL to false by default for better reliability (6a3d09b)
Prepare v1.5.0: Fix intentional bug, bump version fallback, and restore defaults (2b4ba6f)
Trigger review on PR 3 with non-incremental mode (d1a1f82)
Merge full review configuration (fa7c006)
Force full review for verification (48d0f7c)
Fix GetPRDetails error handling and introduce intentional bug in GetDiff (32db1e6)
Trigger final verification for professional comment format (7e64c78)
Sync professional comment format (17a3ed6)
Refactor AI review summary to use professional headers and separators (956ab60)
Remove obsolete Dockerfile after conversion to composite action (f659d9a)
Merge branch 'main' into test/sample-review (20f78bd)
Update default Gemini model to 2.5-flash (7cd14d2)
Fallback to gemini-2.0-flash for verification (9f9d7c2)
Fix model names: use gemini-1.5-flash (54ac410)
Fix composite action: pass repo and pr number explicitly (d1f3c46)
Fix composite action download with curl (8b0cb76)
Final Test 2: Add decorative banner (dee8d80)
Final Test: Add decorative banner and refactor PR details fetching (646b084)
Convert to Composite Action and update documentation (fe11d51)
Add self-review workflow (7330111)
Add token usage tracking and enforce English in learning reports (96da71d)
Require summary sections (core changes, concerns, verdict) in concise rules (352c3d3)
Strengthen concise review rules to override base prompt (ef25558)
Add detailed info logs for review configuration (d8aa540)
Remove reusable workflow from action repo (9e884d5)
Make concise review rules the default, add none to disable (d4c0a8a)
Add review rules presets and custom rules support (52ef833)
Add --strict-changes flag for lenient review verdicts (2c9e27e)
Fix review fallback order and improve Eli persona tone (037847d)
Add logo banner to README (8010804)
Add teaching personas, multilingual reviews, and learning reports (a36ea22)
Fix 422 fallback to downgrade event type on own-PR reviews (58db13f)
Initial implementation of ReviewForge (f1f38ff)

Contributors

Testing

Extensive internal testing and verification were conducted for new features like self-review, AI personas, and review rule configurations. Testing also covered the composite action conversion and various bug fixes to ensure stability and reliability.

Installation

Go Install

go install github.com/AxeForging/reviewforge@v1.5.1

Linux/macOS (AMD64)

curl -L https://github.com/AxeForging/reviewforge/releases/download/v1.5.1/reviewforge-linux-amd64.tar.gz | tar xz
chmod +x reviewforge-linux-amd64
sudo mv reviewforge-linux-amd64 /usr/local/bin/reviewforge

Linux/macOS (ARM64)

curl -L https://github.com/AxeForging/reviewforge/releases/download/v1.5.1/reviewforge-linux-arm64.tar.gz | tar xz
chmod +x reviewforge-linux-arm64
sudo mv reviewforge-linux-arm64 /usr/local/bin/reviewforge

Contributors

murilopmachado and machado144

Assets 12

01 Mar 20:03

github-actions

v1.5.0

8e2bc8b

v1.5.0

ReviewForge v1.5.0: High-Performance Refinement 🚀

This release marks a significant milestone in ReviewForge's evolution, transitioning to a High-Performance Composite Action architecture and introducing a more professional review experience.

Key Highlights

⚡ High-Performance Composite Architecture

Zero-Docker Overhead: Fully migrated from Docker to a Composite Action.
Fast Execution: The action now downloads pre-built binaries directly using curl, significantly reducing startup time (typically ~10s total execution).
Architecture Aware: Automatically detects runner OS (Linux/macOS) and Architecture (x64/ARM64) to fetch the correct binary.

👔 Professional Review Experience

Section-Based Summaries: PR summaries now use clear ### Headers and --- horizontal separators for a cleaner, more professional look.
Improved Verdicts: Summaries focus on Core Changes, Concerns, and a clear Verdict with justification.
Strict English Reports: Learning reports and AI insights are now strictly enforced in English (en-US) for consistency, even when the review itself is in another language.

🛡️ Safer Defaults & Reliability

Non-Incremental by Default: INCREMENTAL mode is now false by default. This ensures a comprehensive PR analysis for every push, reducing the risk of missed context in complex diffs.
Robust PR Detection: Improved detection of repository and PR numbers in composite environments.
Token Usage Tracking: Basic token usage metrics can now be appended to the review summary for better cost tracking.

What's Changed

Composite Action: Converted from Docker to binary-downloading composite runner.
Formatting: Refactored system prompts for header-based Markdown summaries.
Defaults: Set INCREMENTAL: false as the new default for better reliability.
Cleanup: Removed obsolete Dockerfile and streamlined repository structure.

Full Changelog: v1.4.0...v1.5.0

Assets 11