Codemetry

Codemetry analyzes a Git repository and produces a metrics-based "mood proxy" (bad/medium/good) for each day or time window. It does not infer emotions. Instead it estimates quality/strain/risk using measurable signals such as churn, scatter, follow-up fixes, and optional static-analysis totals, and reports a confidence score plus human-readable reasons.

Documentation | Getting Started

Requirements

• PHP 8.2+
• Git

Installation

Codemetry is a development/analysis tool and should be installed as a dev dependency.

Laravel adapter

composer require codemetry/laravel --dev

Publish the configuration file:

php artisan vendor:publish --tag=codemetry-config

WordPress (WP-CLI)

wp package install codemetry/wordpress
# or
composer require codemetry/wordpress --dev

Core (framework-agnostic)

composer require codemetry/core --dev

Usage

Laravel

# Table output (one row per day)
php artisan codemetry:analyze --days=7

# JSON output
php artisan codemetry:analyze --days=7 --format=json

# Filter by author or branch
php artisan codemetry:analyze --days=7 --author="Jane Doe" --branch=main

# Enable AI-powered explanations (requires API keys in config)
php artisan codemetry:analyze --days=7 --ai=1

WordPress (WP-CLI)

# Table output (one row per day)
wp codemetry analyze --days=7

# JSON output
wp codemetry analyze --days=7 --format=json

# Filter by author or branch
wp codemetry analyze --days=7 --author="Jane Doe" --branch=main

# Enable AI-powered explanations
wp codemetry analyze --days=7 --ai=1

# View current configuration
wp codemetry config

Core PHP API

use Codemetry\Core\Analyzer;
use Codemetry\Core\Domain\AnalysisRequest;

$analyzer = new Analyzer();
$request = new AnalysisRequest(
    days: 7,
    branch: 'main',
);

$result = $analyzer->analyze('/path/to/repo', $request);

foreach ($result->windows as $mood) {
    echo "{$mood->windowLabel}: {$mood->moodLabel} ({$mood->moodScore}/100)\n";
}

Signal Providers

Built-in providers that generate metrics for each analysis window:

Provider	Signals
ChangeShape	Additions, deletions, churn, commit count, files touched, churn per commit, scatter
CommitMessage	Fix/revert/wip keyword counts, fix ratio
FollowUpFix	Commits touching the same files within a configurable horizon, fix density

Optional tool-based providers (best-effort, skip gracefully if tools are unavailable):

Provider	Tool
PhpQuality	phpstan / psalm
JsTsQuality	eslint / tsc
CssQuality	stylelint

Extending

Add a custom signal provider by implementing the SignalProvider interface:

use Codemetry\Core\Signals\SignalProvider;
use Codemetry\Core\Domain\RepoSnapshot;
use Codemetry\Core\Domain\SignalSet;
use Codemetry\Core\Signals\ProviderContext;

class MyProvider implements SignalProvider
{
    public function id(): string
    {
        return 'my_provider';
    }

    public function provide(RepoSnapshot $snapshot, ProviderContext $ctx): SignalSet
    {
        // Compute and return signals
    }
}

Register it in the provider list — no pipeline changes required.

AI Engines

AI integration is opt-in and metrics-only. Engines never receive raw code or diffs.

Supported engines:

• OpenAI
• Anthropic (Claude)
• DeepSeek
• Google (optional)

Configure in config/codemetry.php:

'ai' => [
    'enabled' => true,
    'engine' => 'openai', // openai, anthropic, deepseek, google
    'api_key' => env('CODEMETRY_AI_KEY'),
],

When AI is requested but unavailable (missing keys, API failure), analysis continues normally with an ai_unavailable confounder added to the result.

Privacy

• The AI engine abstraction receives aggregated metrics only — never raw source code, full diffs, or file contents.
• All analysis runs locally via Git commands against your repository.
• No data is sent to external services unless AI engines are explicitly enabled.

License

MIT