git-explain

Understand what changed, why it changed, and what risks it introduces — powered by AI.

Instead of +520 -380, git-explain tells you what actually changed, infers why from repository context, highlights regressions and affected modules, and suggests commit messages, PR descriptions, and test cases.

Installation

npm install -g git-explain

Quick Start

On first run, the setup wizard walks you through selecting a model:

$ git explain

Welcome to Git Explain!
  ✓ Git detected
  ✓ Ollama detected

Select AI model:
  1. Qwen3-Coder       (recommended)
  2. DeepSeek V4 Flash
  3. GPT-5              (requires API key)
  4. Claude             (requires API key)

Local models run on your machine. Ollama models (Qwen3-Coder, DeepSeek V4 Flash) are downloaded automatically.

Usage

git explain                    # Full analysis of working tree changes
git explain --staged           # Staged changes
git explain HEAD               # Last commit
git explain abc1234            # Specific commit hash

git explain -m                 # Commit message only
git explain -r                 # Risk analysis only

git explain --offline          # Force offline mode
git explain --model qwen3-coder:latest
git explain --setup            # Re-run setup wizard

Sample Output

$ git explain --staged

  Files: 7   +36 / -51

  Change Analysis
──────────────────────────────────────────────────

  📌  Summary
    Switches the default AI model from deepseek-v4-flash
    to qwen3-coder, refactors git diff parsing, and
    separates .npmrc from version control.

  🎯  Why This Change Was Made
    The primary motivation is switching the default AI model
    to qwen3-coder for better availability. Supporting changes
    fix a bug where staged diff stats were incorrectly
    calculated, and clean up configuration management.

  ⚠️  Potential Regressions
    - Users expecting deepseek-v4-flash by default may
      need to manually configure it
    - Modified parseDiffResult may have edge cases in
      complex Git scenarios

  🔗  Affected Modules
    - src/config.ts — Default model change
    - src/git.ts — Fixed staged diff stats calculation
    - src/llm/ollama.ts — Enhanced type safety
    - src/setup.ts — Updated model selection defaults
    - .gitignore / .npmrc.example — Config separation

  💾  Suggested Commit Message
    feat: switch default model to qwen3-coder and
    improve Ollama integration

  📝  PR Description
    This PR switches the default AI model to qwen3-coder,
    refactors the Ollama provider for better type safety,
    fixes staged diff statistics, removes unused code, and
    separates .npmrc from version control for security.

  🧪  Test Cases
    1. Verify qwen3-coder is set as default in config
    2. Test setup wizard initializes with correct default
    3. Validate Ollama model existence checking
    4. Ensure git diff stats work for staged + unstaged
    5. Confirm .npmrc.example contains correct config

Output Sections

Section	Description
Summary	2-3 sentence overview
Why This Change Was Made	Motivation inferred from diff + repository context
Potential Regressions	Specific regressions with file references
Affected Modules	Directly and indirectly impacted modules
Suggested Commit Message	Conventional commit format
PR Description	3-5 sentence PR summary
Test Cases	Specific test scenarios to cover

Configuration

Config is stored at ~/.git-explain/config.json. Environment variables override at runtime:

Variable	Used For
OPENAI_API_KEY	GPT-5 / OpenAI-compatible
ANTHROPIC_API_KEY	Claude
OLLAMA_URL	Custom Ollama endpoint (default: http://localhost:11434)

Requirements

• Node.js 18+
• Git
• Ollama (for Ollama models)

License

Apache 2.0