KOReader Assistant Plugin: v1.11 Release

This release focuses on optimizing performance, expanding model provider support, improving UI responsiveness, and fixing several edge-case bugs.

Key Highlights

OpenRouter Model Picker: Introduced a fully dynamic on-device UI model picker for OpenRouter with responsive sizing, page-forward shortcuts, and better model ID visibility.
DeepSeek: Added official support for DeepSeek's Thinking Mode parameter.
Google Gemma: Added a dedicated Gemma handler featuring automatic API detection and dynamic <thought> tag filtering for Gemma 4 models.
New APIs: Added integrated support for Perplexity, GigaChat, and Anthropic web search capabilities.

Gzip Compression: Implemented gzip compression across model fetching, update checking, and OpenRouter api workflows (leveraging libz), significantly decreasing data payload sizes and accelerating model list loading times.
Parallel API Requests: Improved localization/makefile processing with parallelized API calls.
Algorithm Upgrades: Integrated major optimizations for Lexrank to improve text summarizing efficiency.

KOReader Compatibility: Refactored the dictionary popup buttons to support the latest KOReader API (PR #15184+) while maintaining seamless backward compatibility with older versions.
Dialog Enhancements: Added a dedicated close button (top-right cross) to the "Ask Another" dialog and resolved query animation flashing issues.
Clipboard Sync: Added support for copying inputs to the clipboard for built-in system prompts.

Encoding Robustness: Fixed UTF-8 encoding issues post-text truncation by cleaning up orphaned continuation bytes via util.fixUtf8().
Reliability: Resolved a critical application crash that occurred when configuration.lua was missing.
Flow Correctness: Fixed a bug where follow-up questions were dropped after running X-Ray or Recap operations.

Full Changelog: v1.10...v1.11