v2.8.0 - Windows Git Bash support

What's New

🪟 Windows Support

• Git Bash (MINGW64/MSYS2) now officially supported
• Explicit OS detection with detect_os() helper
• Platform-aware install paths ($HOME/bin), cache ($LOCALAPPDATA), and config ($APPDATA)
• WSL was already supported (it's Linux)

🔧 GitHub Contribution Infrastructure

• Issue templates (bug report, feature request) with pre-flight checks and dropdowns
• PR template with linked issue requirement, type labels, and automated checks table
• PR validation workflow: issue reference, status:approved, type:* label checks
• CONTRIBUTING.md rewritten to strict issue-first workflow
• 11 new labels: type:*, status:*, priority:*

🤖 Project Skills (for AI agents)

• 6 project-local skills: branch-pr, commit-hygiene, docs-alignment, issue-creation, shellcheck-standards, testing-coverage
• AGENTS.md index for automatic skill loading

Upgrade

brew update && brew upgrade gga
# or
curl -fsSL https://raw.githubusercontent.com/Gentleman-Programming/gentleman-guardian-angel/main/install.sh | bash

v2.7.3

Bug Fix

• Fix missing pr_mode.sh in install script: install.sh was not copying lib/pr_mode.sh to the install directory, causing --pr-review to fail for users installing GGA directly (not via gentle-ai).

v2.7.2 — Clarify Antigravity independence from VS Code

Fix

• Clarified that Antigravity is Google's AI-first IDE, independent from VS Code configuration
• GGA works through git hooks — no IDE-specific configuration needed for any IDE

v2.7.1 — VS Code & Antigravity Integration Docs

What's New

VS Code / Antigravity Integration

• New integration section: setup instructions for using GGA with VS Code and Antigravity's Source Control panel
• Tips for output visibility, bypass with --no-verify, and Antigravity's built-in Gemini
• Added Antigravity to the provider list in intro

Troubleshooting

• New section: "GGA not running from VS Code Source Control panel" with 4 diagnostic steps

Provider Examples

• Added Antigravity/VS Code usage comment in provider examples

Note: No code changes — all 7 providers were already fully implemented. This release adds documentation for IDE-based workflows.

v2.7.0

What's New

⏱️ Timeout & Progress Feedback (#35)

AI provider calls now have a configurable timeout with visual progress indicator.

• TIMEOUT config option (default: 300s / 5 minutes)
• GGA_TIMEOUT env var override
• Spinner animation in TTY mode, periodic text updates in CI/pipes
• Exit code 124 on timeout with helpful troubleshooting suggestions
• Based on PR #20 by @ramarivera

# In .gga config
TIMEOUT="600"  # 10 minutes for large reviews

🤖 GitHub Models Provider (#36)

Access GPT-4o, DeepSeek R1, Grok 3, Phi-4, LLaMA, and more through GitHub's model inference API.

• Provider format: github:<model>
• Auth via gh auth token — no extra API keys needed
• Uses python3 for safe JSON handling (no jq dependency)
• Based on PR #3 by @Kyonax

# In .gga config
PROVIDER="github:gpt-4o"
PROVIDER="github:deepseek-r1"
PROVIDER="github:grok-3"

🔍 PR Review Mode (#37)

Review all files changed in a PR, not just the last commit.

• --pr-mode: review all files changed compared to base branch
• --diff-only: with --pr-mode, send only diffs (faster and cheaper)
• Auto-detects base branch (main/master/develop)
• PR_BASE_BRANCH config option for manual override
• Based on PR #30 by @Jose-cd

gga run --pr-mode                # Full PR review
gga run --pr-mode --diff-only    # Diffs only (saves tokens)

📊 Test Coverage

• 174 unit tests, 0 failures
• 61 new tests across the 3 features

Upgrade

brew upgrade gga
# or
curl -fsSL https://raw.githubusercontent.com/Gentleman-Programming/gentleman-guardian-angel/main/install.sh | bash

Full Changelog: V2.6.2...v2.7.0

v2.6.2

Bug Fixes

• fix: cross-platform hash support for Linux (sha256sum fallback for shasum)

v2.6.1 - Relaxed STATUS parsing

What's Changed

Bug Fix: Relaxed STATUS parsing (#18, PR #19)

AI providers with system-wide instruction files (like AGENTS.md or CLAUDE.md) often prepend acknowledgment text before the actual review response:

# 📋 Instructions loaded!
- /path/to/AGENTS.md
---
**STATUS: PASSED**

This was causing "ambiguous response" failures because GGA expected STATUS on line 1.

Changes

• ✅ Search for STATUS in first 15 lines instead of requiring line 1
• ✅ Accept markdown formatting (**STATUS: PASSED**)
• ✅ 14 new tests for STATUS parsing edge cases
• ✅ 161 tests total

Upgrade

brew upgrade gga
# or
curl -fsSL https://raw.githubusercontent.com/Gentleman-Programming/gentleman-guardian-angel/main/install.sh | bash

Full Changelog: v2.6.0...v2.6.1

v2.6.0

What's New

Commit Message Validation

• gga install --commit-msg - Install commit-msg hook instead of pre-commit
• Commit message is automatically included in AI review when using commit-msg hook
• No config needed - behavior is automatic based on context
• Perfect for validating conventional commits, issue references, etc.

Index Corruption Fix (#15)

• Now reads from staging area (git show :file) instead of working directory
• Fixes race conditions when files are modified after staging
• Works correctly with lint-staged, prettier, and other tools that modify files

Other Improvements

• Signal handling for graceful cleanup on Ctrl+C
• gga uninstall now handles both pre-commit and commit-msg hooks
• 147 tests (17 new)

Contributors

• @ramarivera - Original commit-msg hook implementation (#11)

Full Changelog

v2.5.1...v2.6.0

v2.5.1 - Provider CLI fixes

Bug Fixes

• fix(gemini): Use -p flag for non-interactive prompt passing. Gemini CLI requires the -p flag for explicit prompt passing in non-interactive mode - fixes exit code 41 issue.

• fix(opencode): Use positional argument instead of stdin pipe. OpenCode CLI accepts prompt as positional argument per documentation: opencode run [message..]

Both providers now use the correct CLI invocation method per their official documentation, fixing issues in CI/non-interactive environments.

Tests

All 130 tests passing (12 skipped - Ollama integration tests require local Ollama instance).

v2.5.0 - OpenCode Provider

What's New

Features

• OpenCode provider support (PR #4 by @ramarivera)
  • PROVIDER="opencode" for default model
  • PROVIDER="opencode:model_name" for specific models (e.g., opencode:anthropic/claude-opus-4-5)

Documentation

• CONTRIBUTING.md - Comprehensive guide covering:
  • Project philosophy and design decisions
  • Development workflow and testing
  • Release process
  • PR review guidelines

Provider Examples

# Use OpenCode with default model
PROVIDER="opencode"

# Use OpenCode with specific model
PROVIDER="opencode:anthropic/claude-sonnet-4"

Tests

• 130 tests total (12 new for OpenCode)

Install/Upgrade

# Homebrew
brew update && brew upgrade gga

# Direct install
curl -fsSL https://raw.githubusercontent.com/Gentleman-Programming/gentleman-guardian-angel/main/install.sh | bash

Contributors

• @ramarivera - OpenCode provider implementation