⚡ Performance: Improve memory management and add file size limits

[ManvithGopu13]

Performance Issues

Problems:

1. Global Model State

   • functions/image/image_local.py:7: Global _pipeline variable
   • Models stay in memory indefinitely
   • No cleanup mechanism
   • Could cause memory leaks in long-running processes

2. Large Base64 Responses

   • No file size validation before encoding
   • Large video/audio files encoded entirely in memory
   • Could cause memory exhaustion or slow responses

3. No Response Size Limits

   • Base64 encoded responses could be extremely large
   • No chunking or streaming for large files
   • Could timeout or fail for large outputs

4. No Caching

   • Models reloaded unnecessarily
   • No caching of generated content
   • Repeated requests with same parameters regenerate

5. Synchronous Operations

   • Long-running operations block the request thread
   • No async support for video generation (can take 15+ minutes)

Recommended Solutions:

• Implement model cleanup/unloading after idle period
• Add file size checks before base64 encoding
• Implement response size limits or streaming
• Add caching layer for models and generated content
• Consider async/background jobs for long operations
• Add memory monitoring and limits

Priority: Medium
Type: Performance