docs: update README with UI improvements and Advanced Mode features

refactor: implement Priority 2 UI improvements

docs: add Priority 1 UI improvements summary

refactor: implement Priority 1 UI improvements from review

fix: route Advanced Mode progress messages to thinking output instead of summary

fix: remove legacy model info panels and initialize unified panel with default model

refactor: unify Model Information panel and deduplicate language selector

refactor: reorganize UI with single mode selector and stage-based Advanced Mode layout

fix: Advanced Mode bugs and enhancements

fix: Optimize LFM2-Extract models based on official docs

feat: Implement Advanced Mode 3-stage pipeline (extraction → deduplication → synthesis)

docs: Add Advanced Mode implementation plan

fix: download button now shows JSON file after generation

Improve UI/UX: Modern glassmorphism design, added Paste Text tab, and optimized visual hierarchy

docs: update README with new UI features and custom GGUF loading

style: comprehensive CSS overhaul for polished visual design

refactor: reorganize UI layout for better UX

fix: refresh Model Information panel after custom GGUF model loads

fix: change size_mb from string 'Unknown' to integer 0 to fix format error

fix: add missing custom_model_metadata state variable and update get_model_info to handle custom models

fix: ensure custom model metadata is properly returned and stored

fix: restore missing UI components (thinking_output, summary_output, info_output)

fix: populate Model Information for custom GGUF models

fix: correct package name for gradio_huggingfacehub_search

feat: redesign custom GGUF loader with native HF search

feat: add dynamic model search for custom GGUF loader

fix: make custom GGUF status messages visible

feat: add debug system prompt display and smart custom GGUF loader

Update three-model reasoning system with supports_reasoning field

fix: add file upload validation

fix: preserve Model Information layout during generation

feat: add DEFAULT_N_THREADS env var support + prettify model info

feat: add CPU thread configuration UI

feat: Add GPU detection with automatic CPU fallback

fix: Use N_GPU_LAYERS env var for GPU acceleration support

feat: Add GLM-4.7-Flash-30B (Original) IQ2_XXS 2-bit quantization

feat: Add Breeze 3B Q4 model (mradermacher/breeze-3b-GGUF)

docs: Update README with 22 models, reasoning modes, and GPU support

fix: Correct LFM2 2.6B Transcript context window to 32K

feat: Add 4 new models (1B-3B range) with various quantizations

feat: Add two non-reasoning model variants

feat: Reorder models by parameter count and move model dropdown in UI

feat: Add thinking headroom for max_tokens and Qwen3-30B model

feat: Add GLM-4.7-Flash-REAP-23B model with community settings

fix: Remove ERNIE 21B Q3 variant, keep only TQ1_0

fix: Remove misleading '(locked)' text from temperature display

feat: Auto-populate Advanced Settings from model config (Option A)

fix: Sync TQ1_0 settings with official ERNIE 4.5 specs

fix: Use official ERNIE 4.5 generation settings

feat: Replace ERNIE 21B with IQ3_XXS from mradermacher