feat: integrate Gemini AI, free media engines, asynchronous task tracker, and PM2 auto-boot daemon

[randykaskuser]

Description

This pull request transitions the YouTube Automation Agent from a synchronous, OpenAI-dependent local runner into a production-ready, 24/7 background-daemonized platform. By introducing Google Gemini, free community media generation tools, asynchronous task tracking, and a glassmorphic real-time dashboard, the system can now run continuously on a remote VPS without risk of browser request timeouts or active client shutdowns.

---

🛠️ Summary of Changes

1. 🤖 Gemini AI & Free Content Engine Integration

• Credential Manager (utils/credential-manager.js): Updated to allow server initialization and startup if either openai or gemini keys are defined in the environment.
• Story Writer Agent (agents/script-writer-agent.js): Integrated gemini-2.5-flash to write rich, creative Indonesian children's stories tailored for the @ceritaku.singkat channel strategy.
• Asset Generator (utils/ai-video-generator.js):
  • Connected scene prompt descriptions to a free cartoon generator powered by image.pollinations.ai.
  • Implemented an Indonesian Google Translate Text-to-Speech (TTS) narration synthesizer that chunks long text naturally and stitches the audio files seamlessly via fluent-ffmpeg.
• Verification Script (test-gemini-flow.js): Created a standalone flow verification utility to validate Gemini responses, illustration downloads, and TTS assembly without running the full dashboard server.

2. ⚡ Non-Blocking Asynchronous Pipeline

• Dashboard Server (index.js):
  • Refactored /generate into a non-blocking asynchronous endpoint that starts the process in a background worker and returns 200 OK instantly to prevent HTTP timeouts.
  • Exposed a /generation-status polling endpoint to report progress, current pipeline stage, logs, and estimated remaining duration.
  • Added uptime metrics to /health to support dashboard health checks.

3. 🎨 Premium Glassmorphic Dashboard

• Monitoring UI (dashboard/index.html):
  • Resolved a client-side ReferenceError: process is not defined crash by retrieving platform uptime from /health.
  • Fixed card heights and removed persistent .loading classes to resolve card content overlap bugs.
  • Redesigned the entire interface into a stunning, responsive dark glassmorphic layout using the Google Fonts Outfit typography.
  • Added a dedicated Content Generation Hub control card featuring an active progress spinner, neon stage indicators (Queue, In Progress, Finished), and a live countdown timer showing estimated remaining seconds.

4. 🚀 VPS Daemonization & Uptime

• PM2 Setup (ecosystem.config.js): Added a production PM2 config to run the service as a background daemon with an automatic memory restart threshold at 1GB.
• Dependency Locking (package.json, package-lock.json): Added required utility libraries to handle text-to-speech downloads, folder sanitization, and concurrent execution.

5. 🔒 Security & Bloat Prevention

• Ignore Patterns (.gitignore): Added /data/ and related directories to ensure dynamic assets (temporary image downloads, JSON files, output MP4 videos) and credential key backups are never committed or exposed.
• Configuration (.env.example, README.md): Configured clear setup guides, missing variable examples, and deployment documentation to assist with manual configuration on fresh systems.

---

🚦 Verification Status

1. Standalone Validation (test-gemini-flow.js):
   Successfully tested script drafting, Pollinations cartoon download, and audio concatenation with zero errors.
2. Background Jobs (/generation-status):
   Verified that status endpoints return detailed JSON payloads mapping active sub-agent states:

   {
     "status": "generating",
     "currentStep": "Generating visual illustrations and synthesizing audio narration...",
     "steps": {
       "strategy": "completed",
       "script": "completed",
       "thumbnail": "completed",
       "seo": "completed",
       "production": "processing"
     },
     "estimatedSecondsRemaining": 115
   }