Download the PHP package muzorix/ai-blog-engine without Composer

muzorix/ai-blog-engine

Fully automated, backend-only AI blog/article system for Laravel.

Discover topics, write full articles, humanize them, run fact-checking and SEO/AEO/GEO optimization, generate images and FAQs, manage internal linking and duplicate detection, and control the full lifecycle through a Filament admin panel. You bring your own frontend (Blade, Livewire, Inertia, or headless).

MIT licensed. Multi-provider throughout: Groq, Gemini, OpenAI, Claude.

Table of Contents

• What This Package Does
• Feature List
• Requirements
• Installation
• Configuration
• Admin Panel Guide
• The Article Pipeline
• Day-to-Day Usage
• Artisan Commands
• Scheduled Jobs
• Publishing Modes
• Topic Discovery
• AI Providers
• Image Providers
• Frontend Integration
• Models & Querying
• Programmatic Usage
• Database Tables
• Troubleshooting
• Package Structure
• License

What This Package Does

Most AI writing tools generate text and stop. This package treats article creation as a full lifecycle:

1. Discover and score topics from RSS, News API, or manual entry
2. Generate structured articles section-by-section
3. Humanize, fact-check, and optimize for SEO, AEO, and GEO
4. Generate FAQs, images, and internal links
5. Flag duplicates and route to review (or auto-publish)
6. Monitor freshness and refresh aging content

Everything is self-hosted. Your data, your pipeline, your providers.

Feature List

Requirements

• PHP 8.2+
• Laravel 11+ (tested on Laravel 13)
• Database MySQL, PostgreSQL, or SQLite
• Queue driver — database or redis recommended (QUEUE_CONNECTION=database)
• At least one AI text provider API key (Groq recommended — free tier available)
• Composer with ext-json and ext-mbstring

Installation

1. Install the package

For local path development (monorepo):

2. Run the install command

This will:

• Publish config/ai-blog-engine.php
• Publish and run database migrations
• Seed default topic sources (Manual, RSS, News API)
• Create the public/storage symlink (for images)
• Publish Filament CSS/JS assets

3. Publish Filament assets (if admin looks unstyled)

If the admin login page has no styling, run:

Important: Set APP_URL in .env to match how you access the site (http://127.0.0.1:8000 vs http://localhost:8000).

4. Configure environment variables

Copy the block below into .env and fill in your keys:

5. Enable Filament panel access on your User model

Your App\Models\User must implement Filament\Models\Contracts\FilamentUser:

6. Create an admin user

7. Start the queue worker

Article generation runs as queued jobs. Start a worker:

Or use the Laravel composer dev script if your project has it.

8. Start the scheduler (production)

Add to your server crontab:

The package registers these scheduled jobs automatically:

9. Open the admin panel

(Path is configurable via ABE_ADMIN_PATH.)

Configuration

Publish or re-publish config anytime:

Key config sections

Most settings can also be changed from AI Settings in the admin panel without editing files.

Admin Panel Guide

The Filament panel is at /abe-admin (panel ID: ai-blog-engine).

Dashboard

• Pipeline stats: queued topics, pending review, published count, freshness queue
• Recent articles table

Topics

Manage the topic pipeline:

Topic statuses: pending → scored → queued → processing → completed

Articles

Full CRUD for generated articles:

Review panel shows SEO score, readability, AEO score, duplicate flag, and fact-check flags.

Sources

Enable/disable and configure topic sources:

• Manual — topics entered by hand
• RSS — comma-separated feed URLs in config or config key-value in admin
• News API — requires NEWS_API_KEY and search query

Adjust performance_weight to boost sources that produce high-performing articles.

Freshness Queue

Articles flagged as needs_update appear here:

• Run Freshness Scan — scan all published articles
• Refresh — regenerate an updated draft through the full pipeline

AI Settings

Configure without touching .env:

• Site niche, description, author, URL
• Mode (review vs self-writing)
• Text and image providers
• Tone and target word count
• Humanization, fact-check, and inline image toggles

The Article Pipeline

When a topic is generated, it passes through these steps in order:

Each step logs to abe_ai_generation_logs with provider, tokens, and cost estimate.

Day-to-Day Usage

Workflow A: Fully automated (with review)

Workflow B: Manual topic → article

1. Go to Topics → Create
2. Enter title and summary (paste source content in summary for fact-checking)
3. Set status to queued
4. Click Generate on the topic row (or run php artisan ai-blog-engine:generate {id})
5. Review and approve in Articles

Workflow C: Self-writing (no human step)

Articles publish automatically when fact-check and duplicate checks pass.

Workflow D: Refresh aging content

Then use Freshness Queue in admin to refresh flagged articles.

Artisan Commands

Scheduled Jobs

Registered automatically when APP_ENV runs in console. Customize cron in config/ai-blog-engine.php under schedule:

Override via env: ABE_SCHEDULE_FETCH_TOPICS, ABE_SCHEDULE_FRESHNESS, etc.

Publishing Modes

Article statuses: draft → humanizing → fact_checking → seo_processing → pending_review → published

Also: scheduled, needs_update, rejected

Topic Discovery

Manual

Create topics in admin. Best practice: paste source article text in summary or store full content in raw_payload so fact-checking can verify claims.

RSS

Or configure feeds per-source in Sources → RSS → config key feeds as JSON array.

News API

Get a free key at newsapi.org.

AI Scoring

Each topic receives:

• Relevance score (0–100) — fit to your niche
• Opportunity score (0–100) — timeliness and content gap potential

Topics below ABE_MIN_RELEVANCE (50) or ABE_MIN_OPPORTUNITY (30) are not queued.

AI Providers

Text providers

Use one provider for all steps:

Multiple Groq keys (recommended for production / heavy usage):

When one key hits its rate limit, the package automatically switches to the next key — no job failure, no manual intervention.

You can still set GROQ_API_KEY for a single key; GROQ_API_KEYS takes precedence when set (comma-separated, no spaces required).

Keys are rotated round-robin for even load. On HTTP 429, the current key is skipped and the next is tried immediately.

Or override per pipeline step:

Programmatic usage

Image Providers

Images are stored on the disk configured in ABE_IMAGE_DISK (default: public) under abe-images/.

Frontend Integration

This package is backend-only. Query models from your own routes, controllers, or Livewire components.

Blade example

Livewire

Canonical URLs

Articles expose a url attribute using ABE_SITE_URL + ABE_BLOG_PATH:

Related articles

Models & Querying

Useful scopes

Publish manually

Programmatic Usage

Run the full pipeline

Generate with custom options

Dispatch as background job

Database Tables

Troubleshooting

Admin panel has no CSS / looks like plain HTML

Hard-refresh the browser (Ctrl+Shift+R).

Admin panel CSS loads but from wrong host

Set APP_URL in .env to exactly match your browser URL:

Articles stuck in processing

Ensure a queue worker is running:

Check storage/logs/laravel.log and the abe_ai_generation_logs table.

"API key not configured" errors

Verify the correct env key for your provider (GROQ_API_KEY, etc.) and run:

Images not displaying

Confirm ABE_IMAGE_DISK=public and files exist in storage/app/public/abe-images/.

No topics after fetch

• Check RSS feed URLs are valid and reachable
• Verify NEWS_API_KEY if using News API
• Lower ABE_MIN_RELEVANCE / ABE_MIN_OPPORTUNITY in config
• Topics with duplicate titles from the last 30 days are skipped

Fact-check flags everything as unverified

Paste the original source content into the topic summary or ensure raw_payload.content is populated when the topic is created.

Package Structure

License

MIT — free for personal and commercial use.