ProxMenux v1.1.9.2 (Beta)

This beta focuses on improving AI model compatibility and reliability, addressing issues where some models from provider APIs were not suitable for notification enhancement.

Main changes in v1.1.9.2

Verified AI Models System

The system now uses a curated list of verified models (verified_ai_models.json) that have been tested and confirmed to work for notification enhancement.

Hybrid Verification: When loading models, the system:

1. Fetches available models from the provider's API
2. Filters to show only models present in the verified list
3. Auto-selects the recommended model for each provider

Each provider has a pre-selected recommended model optimized for notifications (fast and cost-effective).

Per-Provider Model Memory

• Model Selection Persistence: Selected model is now saved per provider
• Seamless Switching: When changing providers, your previous model selection for each provider is preserved
• Automatic Migration: Existing configurations are automatically migrated to the new per-provider model storage

Model Compatibility Fixes

Removed incompatible models from each provider:

• Gemini: Excludes preview, experimental, image, embedding, TTS models
• OpenAI: Shows only GPT chat models, excludes whisper, dall-e, embeddings
• Groq: Excludes whisper, TTS, guard models
• OpenRouter: Excludes image, video, embedding models

Daily Verification: Background task verifies model availability and automatically migrates to a verified alternative if the current model becomes unavailable.

Verified Models by Provider