v1.0.6
🌐 Chinese LLM Provider Support
Added 7 new LLM providers covering major Chinese AI platforms and aggregators:
Kimi (Moonshot)
Aliyun (Alibaba Cloud / Qwen)
Volcengine (ByteDance / Doubao)
Zhipu GLM
MiniMax
OpenRouter
Ollama (Local)
Providers are grouped into OpenAI / Anthropic / Google / Popular in the dropdown.
🔄 Dynamic Model Fetching
Fetch available models from provider APIs in real time instead of relying solely on hardcoded lists
Gemini: filtered by supportedGenerationMethods — only models supporting generateContent are shown
Volcengine: filtered by modalities — only multimodal models with image input + text output
Model selector is gated behind API Key for providers that require one; Ollama, Bedrock, and OAuth types are exempt
Refresh button added next to Model ID for manual re-fetch
💾 Backup & Restore
Export all data (cards, knowledge base, insights, settings) as a .memex file
Save via system share sheet to iCloud Drive, Google Drive, or any location
Restore from a .memex file to recover all data
📂 Configurable Data Storage
Choose where Memex stores workspace data: app storage / custom device folder / iCloud
Custom folder and iCloud modes preserve data across app reinstalls
🔍 Knowledge Base Search
Search PKM knowledge base by filename or content keywords
🛠 UX Improvements
Model ID dropdown no longer auto-focuses the keyboard — tapping the arrow shows the list without triggering input mode
Form field order updated to Provider → Base URL → API Key → Model ID for a more natural configuration flow
Model fetching now logs request/success/failure details for easier troubleshooting