🔍 Detection Criteria

• Chat activity spikes (20% weight)
• Audio excitement levels (25% weight)
• Transcript content analysis (35% weight)
• Visual scene changes (20% weight)

The Self-Learning Engine continuously monitors system performance and automatically identifies areas for improvement.

How it works:

1. Collects performance metrics from all subsystems
2. Analyzes data patterns using LLM reasoning
3. Identifies bottlenecks, quality issues, and optimization opportunities
4. Proposes and applies safe auto-fixes
5. Monitors the impact of changes in subsequent cycles

Monitored dimensions: Video generation quality, clip detection accuracy, system response time, error rates, and user satisfaction patterns.

Automatic mode: The system runs a self-analysis every hour. Issues are detected and fixes are applied automatically.

Manual trigger: Click "Force Analysis Now" to run an immediate analysis cycle and see results in real-time.

Feedback loop: Each fix is tracked. The system learns from both successes and failures, improving over time.

Safety: All auto-fixes are logged and can be reviewed. Destructive changes require manual approval.