Virtual Kimi Changelog

[1.0.8] - 2025-08-19

Changed

Improved fallback logic for LLM responses: now uses localized emotional responses if the LLM reply is empty or invalid.
Made emotional response selection dynamic and robust, based on available variants.
Enhanced error handling for missing API keys, network issues, and API errors, ensuring the user always receives a meaningful message.
Refactored code patching to avoid accidental code removal or misplaced edits.
Clarified and documented emotional response logic for maintainability.

[1.0.7] - 2025-08-19

Changed

Removed the global system prompt that caused issues and implemented per-character system prompts for each character.
Improved voice reading of messages for clearer and more natural audio playback.
Fixed various small bugs related to characters' personality traits.
Improved detection of words and phrases for memory recording to increase accuracy.

[1.0.6] - 2025-08-15

Added

Added 100+ videos for various contexts.

Changed

Optimized video preloading to improve speed on slow web servers.

Fixed

Fixed various minor bugs.

[1.0.5] - 2025-08-13 - "Personality & Language Sensitivity"

Added

Multilingual profanity/insult detection for negative context across 7 languages (en, fr, es, de, it, ja, zh)
Gendered variants support in negative keywords (fr, es, it, de) to improve accuracy (e.g., sérieux/sérieuse)
Extended personality keywords for Spanish and Italian (all traits) with gendered forms

Changed

Personality sync now completes missing values using character-specific defaults (with generic fallback)
Centralized side-effects on personality updates (UI/memory/video/voice) behind a single personality:updated listener
Sliders: generic handler only updates display; persistence and effects handled by specialized listeners
Trait updates preserve fractional progress (2 decimals) for smoother affection changes
Stats now use character-specific default for affection (with generic fallback) when missing

Fixed

Removed obsolete personalityUpdated listener to avoid duplicate processing
Unified KimiMemory affection default loading (removed conflicting double assignment and legacy default 80)
Minor cleanup and consistency improvements in utils and sync flows

[1.0.4] - 2025-08-09 - "Emotion & Context Logic Upgrade"

Added

Major improvements to emotion, context, and personality logic:
- Enhanced emotion detection and mapping for more nuanced responses
- Contextual keyword analysis for better understanding of user intent
- Refined personality trait system with dynamic adaptation
- Video selection logic now adapts to both emotion and conversational context
- Improved handling of multi-layered context (emotion, keywords, personality, situation)

Changed

Video playback and character reactions are now more tightly coupled to detected context and personality traits
Emotion and context logic refactored for clarity and maintainability
Keyword extraction and context matching algorithms improved for accuracy

Technical

Refactored core logic in kimi-emotion-system.js, kimi-logic.js, and kimi-memory-system.js
Updated video selection and playback logic in kimi-memory.js and kimi-memory-ui.js
Improved context propagation between modules

[1.0.3] - 2025-08-09 - "LLM multi-provider"

Added

LLM multi-provider UX enhancements:
- Dynamic API key label per provider (OpenRouter, OpenAI, Groq, Together, DeepSeek, Custom, Ollama)
- Visual "Saved" badge when a key is stored or after a successful test
- Localized tooltip explaining Saved vs connection test

Changed

OpenAI-compatible flow now reads llmBaseUrl/llmModelId and the correct provider key from KimiDB
Clears connection status message when provider/Base URL/Model ID/key changes for clearer feedback

[1.0.2] - 2025-08-09 - "Smoother Video"

Changed

Video playback and transition stability improvements:
- Lightweight MP4 prefetch queue (neutral + likely next clips) to reduce wait times during switches
- Earlier transition on canplay (instead of canplaythrough) for faster, smoother swaps
- Context-aware throttling to prevent rapid switching under load (speaking: ~200ms, listening: ~250ms, dancing: ~600ms, neutral: ~1200ms)

Fixed

Safe revert on failed play() during a switch to avoid frozen frames
Aligned event listeners to canplay and ensured proper cleanup to prevent leaks
Corrected prefetch cache initialization order (prevented undefined.has runtime error)
Removed unsupported <link rel="preload" as="video"> to eliminate console warnings

Technical

Front-end performance tweaks: GPU-accelerated fades with will-change: opacity and backface-visibility: hidden
Connection warm-up: added preconnect/dns-prefetch to the origin for faster first video start
Files updated: index.html, kimi-css/kimi-style.css, kimi-js/kimi-utils.js

[1.0.1] - 2025-08-08

Fixed an issue where the browser prompted to save the OpenRouter API key as a password. The input field is now properly configured to prevent password managers from interfering.
Added a waiting animation that appears between the user's message submission and the LLM's response, improving user feedback during processing.
Added a new section in the API tab: below the recommended LLM models, all available OpenRouter LLM models are now dynamically loaded and displayed for selection.

[1.0.0] - 2025-08-07 - "Unified"

Added

Intelligent Memory System: Automatic extraction and categorization of memories from conversations
Multiple AI Characters: 4 unique personalities (Kimi, Bella, Rosa, Stella) with distinct traits
Advanced Emotion Detection: Real-time emotion analysis with cultural awareness
Plugin System: Extensible architecture for themes, voices, and behaviors
Memory Management UI: Complete interface for viewing, searching, and managing memories
Enhanced Personality System: 6 dynamic traits that evolve based on interactions
Multilingual Support: Full localization in 7 languages with auto-detection
Production Health Check: Comprehensive system validation and monitoring
Performance Optimizations: Batch database operations and improved loading times
Security Enhancements: Input validation, sanitization, and secure API handling

Changed

Unified Architecture: Consolidated all emotion and personality systems
Improved Database: Enhanced IndexedDB implementation with batch operations
Better Error Handling: Centralized error management with fallback responses
Enhanced UI/UX: More responsive and accessible interface design
Optimized Video System: Smoother transitions and better emotion mapping

Fixed

Function export issues in module system
Memory leaks in event listeners
Cross-browser compatibility issues
Voice recognition stability problems
Database initialization race conditions

Technical

Migrated to unified emotion system
Implemented comprehensive validation layer
Added automated health monitoring
Enhanced plugin security validation
Improved mobile responsiveness

[0.0.9] - 2025-08-04 - "Enhanced"

Added

Advanced LLM model selection interface
Improved voice synthesis with better emotion mapping
Enhanced personality trait visualization
Better conversation export/import functionality

Changed

Upgraded database schema for better performance
Improved theme system with more customization options
Enhanced mobile interface responsiveness

Fixed

Various browser compatibility issues
Voice recognition accuracy improvements
Memory management optimizations

[0.0.8] - 2025-08-01 - "Evolution"

Added

Dynamic personality trait evolution
Enhanced emotion detection algorithms
Improved conversation context awareness
Better visual feedback systems

Changed

Redesigned settings interface
Improved conversation flow management
Enhanced error reporting system

Fixed

Database sync issues
Voice recognition edge cases
Theme switching problems

[0.0.7] - 2025-07-29 - "Immersion"

Added

Real-time video emotion responses
Enhanced voice interaction capabilities
Improved conversation context retention
Better visual theme system

Changed

Upgraded UI framework for better performance
Improved data synchronization mechanisms
Enhanced accessibility features

Fixed

Various stability improvements
Better error handling
Improved cross-platform compatibility

[0.0.6] - 2025-07-26 - "Connection"

Added

Multi-language support system
Enhanced conversation memory
Improved personality customization
Better audio/video synchronization

Changed

Redesigned conversation interface
Improved data persistence layer
Enhanced user experience flows

Fixed

Memory leak issues
Browser compatibility problems
Audio synchronization bugs

[0.0.5] - 2025-07-23 - "Rebirth"

Added

Complete application rewrite
Modern ES6+ JavaScript architecture
Responsive design system
Advanced AI integration capabilities
Comprehensive settings system

Changed

Modernized codebase with current web standards
Improved performance and reliability
Enhanced user interface design
Better data management system

Removed

Legacy jQuery dependencies
Outdated browser support

[0.0.4] - 2025-07-20 - "Stability"

Added

Enhanced voice recognition
Improved conversation flow
Better error handling
Enhanced visual feedback

Fixed

Various stability issues
Performance optimizations
Browser compatibility improvements

[0.0.3] - 2025-07-18 - "Polish"

Added

Improved user interface
Better conversation management
Enhanced customization options

Fixed

Various bugs and stability issues
Performance improvements

[0.0.2] - 2025-07-17 - "Improvements"

Added

Basic conversation memory
Improved personality system
Enhanced visual themes

Fixed

Initial bug fixes
Performance optimizations

[0.0.1] - 2025-07-16 - "Genesis"

Added

Initial release
Basic AI conversation capabilities
Voice recognition and synthesis
Simple personality system
Theme customization
Local data storage

Legend

Added: New features
Changed: Changes in existing functionality
Deprecated: Soon-to-be removed features
Removed: Removed features
Fixed: Bug fixes
Security: Security improvements
Technical: Internal technical changes

All notable changes to Virtual Kimi will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.