- How do I install semitexa/laravel-ai-verify in a Laravel application?
- Install it with composer require semitexa/laravel-ai-verify. Laravel package auto-discovery registers the service provider, then publish the config with php artisan vendor:publish and set your AI driver credentials in .env. Run your test suite with the package's fakes enabled so installation does not require live API calls.
- Which Laravel and PHP versions are supported by laravel-ai-verify?
- The package is built for modern Laravel applications and relies on Laravel-native service providers, facades, and config publishing. The composer.json file is the authoritative source for the exact illuminate/support and PHP constraints, so check it before upgrading. The repository's test matrix and release history are the best signal for ongoing maintenance and compatibility.
- How do I configure the AI provider and credentials for laravel-ai-verify?
- Publish the package config and choose a driver; credentials are read from environment variables so you can keep secrets out of source control. The driver-based design lets you swap providers without changing your verification calls. In production, run php artisan config:cache after setting the environment values.
- Can I fake AI verification responses in Laravel tests?
- Yes. The package ships with test-friendly fakes and mocks so your CI pipeline does not make live API calls. Fake the facade or bind a mock driver to return deterministic pass, fail, confidence, and reason values. That keeps verification tests fast and stable while still exercising your validation flow.
- What does an AI verification verdict contain and how do I use it in validation?
- A verdict includes a pass or fail result, a confidence value, and a reason you can log or show to users. You can inspect those fields directly or wrap them in a custom Laravel validation rule. Treat the verdict as structured data and decide whether to reject, flag, or allow the request based on your own policy.
- What are the ai-verify, ai-graph, and ai-verify-receipt MCP tools?
- They are the MCP tool surfaces exposed through Laravel Boost in v0.6.0. ai-verify returns a verdict or gate result, ai-graph handles structural inspection, and ai-verify-receipt supports auditability. The tool payloads mirror the CLI --json output so existing parsers and dashboards keep working.
- How do I enable or disable Laravel Boost MCP integration for laravel-ai-verify?
- The MCP tools register only when laravel/mcp is installed and the package is included through Boost's boost.mcp.tools.include list. Disable the integration with the AI_VERIFY_MCP_BOOST=false environment variable or the ai-verify.mcp.boost config key. Use the environment variable for CI and air-gapped environments where you want a per-environment opt-out without code changes.
- Does a failed AI verification throw an exception or return a verdict?
- In the core package a failed check is returned as a structured verdict you can act on, not an uncontrolled exception. In the MCP path, a fail verdict is intentionally an answer rather than a tool error so agents do not retry, swallow, or misinterpret it. Your application code should convert the verdict into a validation error or response according to your policy.
- What timeout should I use for MCP verification calls on large Laravel suites?
- Laravel Boost runs each MCP tool call in a fresh process, so every verification pays the full application bootstrap cost. The Boost default of 180 seconds can be too short for large suites, which is why the package exposes a timeout parameter. Tune it against your suite size and confirm that timeouts return a structured result or error you can handle.
- How does laravel-ai-verify compare to CAPTCHA, honeypot, and rule-based spam checks?
- It is an AI-powered verification layer, not a replacement for Laravel's built-in validation rules. Unlike CAPTCHA or honeypot checks, it can evaluate text and structured data and return a confidence score and reason for the decision. Because drivers and responses are configurable and fakeable, you can combine it with existing spam screening and content moderation approaches.