Update backend patch: add nvidia/nemotron + google/gemma model IDs

Restore patch_models.py from 5743f66

Fix: patch_models with openai/ prefix for OpenRouter models (owl-alpha, gemini-flash)

Fix backend: litellm openrouter/openrouter/owl-alpha + set as default

Fix backend: owl-alpha with openai/ prefix

Update backend models to match frontend fixes

Fix patch_models: replace kimi/minimax/glm with DeepSeek models in backend

Restore 3d39bbc + change: Kimi→DeepSeek-V4-Pro, MiniMax→deepseek-v4-flash(OR), GLM→DeepSeek-V4-Flash(HF)

Fix: patch existing dropdown items exactly as specified

Update all model IDs as requested

Fix: DeepSeek V4 Pro via OpenRouter + rename Claude to Owl Alpha

Fix: use openai/ prefix + api_base for OpenRouter models in litellm

Fix model IDs: openrouter/openrouter/owl-alpha format for litellm

Set DEFAULT_FREE_MODEL_ID to openrouter/owl-alpha

Fix title generation: use openrouter/owl-alpha for title

Fix title generation: add huggingface/ prefix for litellm

Replace Grok 4.3 with owl-alpha in backend models

Update patch_models.py with correct HF Inference model IDs

Restore 0c286f8: patch_models.py

Update backend patch: accept all new model IDs + HF Inference for Qwen3

Backend patch: accept all new model IDs + route Qwen3-235B via HF Inference

Fix patch: ensure models are added correctly to the list

Patch script: add DeepSeek/Gemini/Qwen3 to dropdown + fix title gen