ChatSpeed v1.1.23

[1.1.23] - 2026-01-22

🚀 New Features

• Model-Specific Custom Body Parameters: Added support for configuring custom JSON parameters (e.g., enable_thinking) at the specific model ID level.
• Dynamic Group Switching (/switch): Introduced a unified /switch route prefix that dynamically resolves requests based on the currently "Active" proxy group.
• System Prompt Keyword Replacement: Implemented KV-based replacement rules in proxy groups to automatically correct System Prompts before forwarding.
• Proxy Group Batch Update: Supports synchronizing prompt injection, text replacement, and tool filtering across multiple groups in one transaction.

🪄 Improvements

• Refined Parameter Injection Hierarchy: Established a strict "Model Patch > User UI Intent > Engine Default" priority system.
• Smart Type Conversion: Automatically converts string values like "true"/"false" to Booleans and numeric strings to Numbers for provider compatibility.
• UI Refactoring: Upgraded the Proxy Group dialog to a tabbed layout, matching the AI Engine settings style and fixing button visibility issues.
• Enhanced Robustness: process_custom_params now intelligently handles various nested data structures and edge cases.

🐞 Bug Fixes

• ModelScope/Qwen Compatibility: Automatically forces enable_thinking to false for non-streaming requests to prevent 400 Bad Request errors.
• UI State Persistence: Fixed an issue where internal chat might lose real-time parameters (e.g., temperature, stop sequences) during protocol conversion.
• Element Plus Adaption: Resolved deprecation warnings for el-radio component APIs.
• Security Hardening: Removed HTML tags from i18n messages to eliminate XSS warnings.

---

[1.1.23] - 2026-01-22

🚀 新特性

• 模型级自定义 Body 参数：支持在具体的模型 ID 级别配置附加参数（如魔塔的 enable_thinking）。
• 动态分组切换 (/switch)：新增统一的 /switch 路由前缀，代理会根据当前“激活”的分组动态转发请求，无需在 IDE 中频繁修改路径。
• 提示词关键词替换：在代理分组中支持 KV 形式的文本替换规则，在请求发出前自动修正系统提示词（System Prompt）。
• 代理分组批量更新：支持一键同步多个分组的提示词注入、文本替换及工具过滤设置。

🪄 改进

• 参数注入优先级金字塔：确立了“模型补丁 > 用户 UI 设置 > 引擎默认值”的严密注入逻辑。
• 智能类型转换：自定义参数支持自动将字符串 "true"/"false" 转换为布尔值，数字字符串转换为数值，确保供应商兼容性。
• UI 重构：代理分组编辑界面升级为选项卡（Tab）布局，风格与“AI 引擎”设置统一，解决了长列表遮挡按钮的问题。
• 增强鲁棒性：process_custom_params 现在可以自动识别并处理多种嵌套级别的数据结构。

🐞 修复

• 魔塔/Qwen 兼容性补丁：自动在非流式请求时强制关闭 enable_thinking，规避 400 报错。
• UI 状态同步：修复了内部聊天模块在转换过程中可能丢失 UI 实时参数（如温度、停止符）的问题。
• Element Plus 适配：修正了 el-radio 等组件的过时 API 用法警告。
• 安全增强：移除了国际化消息中的 HTML 标签，解决 XSS 安全警告。

ChatSpeed v1.1.22

Chatspeed v1.1.22 Release Notes

This release introduces a major new feature: the Intelligent Sensitive Information Filtering System. We are committed to protecting your privacy while maintaining a seamless AI interaction experience. The system now automatically identifies and masks sensitive data before it ever leaves your local device.

✨ Security & Privacy Enhancements

• Automated Data Masking: Automatically detects and masks universal PII such as Emails, IP Addresses, and Credit Cards before sending messages to AI providers.
• Deep Context Recognition (ZH & EN):
  • Chinese Context Optimization: Specialized detection for Chinese Resident IDs, Mobile/Landline numbers, Social IDs (WeChat/QQ), Banking info, Chinese Names, Addresses, Company names, and Project titles.
  • English Context Optimization: Specialized detection for SSNs, English Names, Addresses, Company names, and other English-specific sensitive data.
• Flexible List Management:
  • Custom Blocklist: Manually add specific keywords (e.g., project internal codenames) to be always filtered.
  • Allowlist: Define exceptions (e.g., your public business email or localhost IP) that should never be masked.
• Local-First Security: The entire filtering logic is executed in Rust on your local machine. Only sanitized data is stored in the database and sent to AI models.

🔧 Interaction & UX Improvements

• Quick Toggle Button: Added a filter icon in the chat and assistant header. Toggle filtering state with one click, with real-time sync across all windows.
• Multi-language UI Support: While deep detection is optimized for Chinese and English, the replacement placeholders and UI are localized for 10 languages (EN, ZH, JA, KO, DE, FR, ES, PT, RU).

🐞 Bug Fixes

• Code Cleanup: Removed redundant exports and unused functions to improve system performance.

---

Chatspeed v1.1.22 发布日志

本次更新带来了一个重磅功能：智能敏感信息处理系统。我们致力于在不牺牲 AI 交互体验的前提下，最大限度地保护您的隐私。现在系统能够自动识别并掩盖消息中的敏感信息，确保数据在发送给 AI 供应商之前已完成脱敏。

✨ 隐私与安全增强

• 全自动敏感信息过滤: 在消息发送给 AI 之前，自动识别并掩盖邮箱、IP 地址、信用卡等通用隐私信息。
• 深度适配中英文场景:
  • 中文深度优化: 针对中国大陆地区的身份证号、手机号、座机号、微信号/QQ号、银行卡/支付宝、中文姓名、地址、公司名、项目名、大写金额等进行了专项适配。
  • 英文深度优化: 支持识别 SSN (社会安全码)、英文姓名、国际地址、公司名等英文环境下的敏感信息。
• 灵活的名单管理:
  • 自定义黑名单 (Blocklist): 允许用户添加特定词汇（如项目代号）强制过滤。
  • 白名单 (Allowlist): 允许用户添加例外词汇（如公开邮箱、本地 IP）豁免过滤。
• 极致安全策略: 脱敏逻辑完全在本地 Rust 端执行，且数据库中仅存储脱敏后的安全数据。

🔧 交互与体验优化

• 新增快捷过滤开关: 在主聊天窗口和助手窗口顶部新增 filter 按钮，支持一键切换脱敏状态，并实现跨窗口实时同步。
• 多语言 UI 支持: 虽然深度脱敏目前主要针对中英文，但我们为脱敏后的占位符 (Placeholders) 提供了 10 种语言（中、英、日、韩、德、法、西、葡、俄）的本地化显示支持。

🐞 问题修复

• 内部清理: 清理了冗余代码和未使用的函数，优化了程序运行效率。

ChatSpeed v1.1.21

🔧Optimizations

Optimized window visibility logic; after the change, pressing the hotkey will only hide the window if it has focus, otherwise it will bring the window to the front.

🐞 Bug Fixes

• Fixed an issue in model settings where the baseUrl input field became uneditable when switching API protocols.

ChatSpeed v1.1.19

Chatspeed v1.1.19 is an important release focused on system stability improvements and data security enhancements. This update significantly improves troubleshooting efficiency through comprehensive optimization of error handling mechanisms. Additionally, it introduces streaming encryption technology to solve memory consumption issues when backing up large database files, ensuring stable system operation even when processing massive amounts of data. Furthermore, we have fixed multiple key issues affecting user experience, including chat interface parsing errors, new model compatibility, and cross-platform window display issues, providing you with a smoother and more reliable AI interaction experience.

🔧 Technical Optimizations

• Error Serialization Optimization: Unified and optimized error serialization mechanisms across all modules. Error messages now include specific message content for easier troubleshooting.
• Window Management Enhancement: Improved window management mechanisms with debounced shortcut support and better error handling.
• Web Search Tool Enhancement: Optimized the implementation of the WebSearch tool, improving the stability and performance of search functionality.
• Data Backup Optimization: Optimized database backup logic with new database backup encryption.

🐞 Bug Fixes

• Optimized chat interface: Fixed reference parsing errors and resolved redundant chat content caused by incorrect parsing of step information provided by some vendors.
• Added compatibility for claude-sonnet-4-5 model, supporting cases where thinking.budget_tokens is greater than or equal to max_tokens.
• Optimized the maximum height of settings window to prevent it from exceeding screen display range.
• Removed proxy settings for crawler webview on Windows platform. All platforms now rely on system proxy for webview creation, resolving potential issues with crawler windows on Windows 11.
• Optimized data backup logic, removed support for old databases, and added backup support for mcp_sessions, schema, shared, static, and other directories.

ChatSpeed v1.1.18

Chatspeed v1.1.18 further enhances application startup performance and user experience, especially on the Windows platform. By optimizing the environment initialization process, it resolves startup freezes and unexpected command line window pop-ups, and corrects the accuracy of shortcut descriptions in the release notes.

🔧 Technical Optimizations

• Optimized Environment Initialization, Resolved Startup Freezes and Panic: Refactored the environment initialization process to be asynchronous, resolving potential startup freezes and the "runtime within a runtime" panic on Windows.
• Error Unification: Unified the error handling approach across the application to ensure consistency and readability of error messages.

🐞 Bug Fixes

• Fixed Windows Startup Command Line Popup: Applied CREATE_NO_WINDOW | DETACHED_PROCESS flags to child processes executing shell commands during Windows environment initialization, completely eliminating unexpected command line window pop-ups during startup.
• Corrected Shortcut Description in Release Notes: Corrected the description of main window positioning shortcuts in the release notes, clarifying that they only support horizontal movement (left and right).
• ccproxy Fix: Fixed an issue in the ccproxy module where duplicate content might appear during streaming output in tool compatibility mode.
• Conversation Fix: Fixed an issue where conversation list tool call parameter parsing errors caused the list to fail to load.

ChatSpeed v1.1.17

Chatspeed v1.1.17 focuses on enhancing cross-platform stability and user experience by addressing critical bugs across Windows and Linux environments, optimizing environment variable loading, and resolving UI state management issues in key settings components. This release also incorporates all features and fixes from the unreleased v1.1.16, including new window positioning shortcuts and unified custom event formats.

✨ Enhancements

• Added Main Window Positioning Shortcuts: Added main window center, left and right positioning shortcuts to improve window management efficiency
• Unified Custom Event Format: Unified all custom events to start with cs://, standardizing the application's internal event system

🔧 Technical Optimizations

• Global Shortcut Upgrade: Upgraded main window positioning shortcuts (to left and right) to global shortcuts, allowing triggering without application being active

🐞 Bug Fixes

• Fixed Windows ARM64 Crash Issue: Resolved the issue where the Windows ARM64 version crashed after installation in a PD virtual machine, by explicitly specifying the sqlite3 library path.
• Fixed Windows Command Line Popup Issue: Fixed the problem of an unexpected command line window popping up when starting the MCP service on the Windows platform.
• Fixed Windows MCP Startup Error: Corrected the MCP service startup error on Windows, where directly running the npx script instead of wrapping it with cmd /c led to a %1 is not a valid Win32 application error.
• Optimized Log Visibility: Improved the log output level for shell command failures (e.g., shell not found or internal command failure) when searching for npx and similar commands in Linux environments, making them visible in production.
• Fixed Form State Overwrite Issue: Addressed a bug in Agent.vue and Skill.vue components where editing an item and then adding a new one would overwrite the previous item due to un-reset form data.
• Fixed Initial Language Setting Issue: Fixed the issue where initial installation language was incorrectly set to Simplified Chinese; now it prioritizes the user's native language with fallback to English if unsupported
• Fixed Shortcut Key Validation: Fixed error notifications when shortcut keys are left empty during setting
• Fixed Shortcut Key Binding Mechanism: Fixed issue where unset shortcuts caused module shortcut binding failures
• Fixed Settings Window Switching: Fixed issue where switching between setting windows via menu fails after opening a settings window
• Removed Unavailable Vendor: Removed built-in Pollinations vendor as it was mostly unusable after testing
• Fixed Ubuntu Window Display: Fixed issue where window could not be shown via shortcut after being minimized on Ubuntu

🕰 In Progress

Part of the ReAct logic has been completed, and development of this workflow module will continue

ChatSpeed v1.1.15

v1.1.15 focuses on enhancing the user experience of tool calls and system stability, while optimizing the protocol conversion capabilities of the ccproxy module.

✨ Enhancements

1. Optimized the tool call display interface, where multiple tool calls within the same conversation turn are now displayed as a single session
2. Optimized message deletion, where deleting a message will remove all tool call information for the same conversation turn at once
3. Optimized the conversation entry point, where all conversation errors are now notified to the frontend
4. Optimized error messages, returning status codes for HTTP request-related errors to facilitate frontend adaptation based on the situation
5. Optimized Markdown parsing in conversations, using a streaming parsing mechanism which greatly improves parsing efficiency and user experience
6. Added ALT + ←/→ hotkeys to the main window for quickly moving the window to the bottom-left and bottom-right corners

🔧 Technical Optimizations

7. Optimized the conversation history in the ccproxy module, preprocessing conversation records and tool call records before protocol conversion to make them as suitable as possible for various protocols
8. Optimized the ccproxy module, now supporting * and ? wildcards by defining model injection conditions
9. Optimized database upgrade logic for better version update management in the future

🐞 Bug Fixes

10. Fixed tool call interruption issues caused by non-standard end flag sending in some models (e.g., gtp-oss), improving system compatibility
11. Fixed an issue in the ccproxy module where tool structures were incorrect during complex scenarios when converting from gemini protocol to claude protocol

🕰 In Progress

Part of the ReAct logic has been completed, and development of this workflow module will continue

ChatSpeed v1.1.14

This update primarily introduces proxy alias wildcard support, significantly enhancing the flexibility and convenience of model configuration. You can now use wildcards like * and ? to define proxy aliases, adapting to changes in model names without frequent configuration modifications. Additionally, we have optimized the search scraper's filtering mechanism, fixed a bug where some tool calls in the chat interface could not be expanded, and optimized the model configuration process.

✨ Enhancements

• Optimized the search scraper to automatically filter multimedia websites and files, providing more precise results as requested. The WebSearch tool's default number of results has also been adjusted to 5, with a maximum of 30, offering more flexible search control.
• Added proxy alias wildcard support, now allowing * for multiple characters and ? for single characters.

🔧 Technical Optimizations

• Adjusted API documentation to adapt to the new MCP proxy.
• Optimized the model configuration process by changing the import button from invisible to disabled when adjusting model settings, improving user experience.

🐞 Bug Fixes

• Fixed a bug where some tool calls in the chat interface could not be expanded.

---

本次更新重点引入了代理别名通配符支持，极大地提升了模型配置的灵活性和便捷性。现在，您可以利用 * 和 ? 等通配符来定义代理别名，无需频繁修改配置即可适应模型名称的变化。此外，我们还优化了搜索爬虫的过滤机制，修复了对话界面工具展开的 bug，并对模型配置流程进行了优化。

✨ 功能增强

• 优化了搜索爬虫，现在会自动过滤多媒体网站和多媒体文件，返回结果将按要求更精准。同时，WebSearch 工具的默认返回结果数量调整为 5 条，最大支持 30 条，以提供更灵活的搜索控制。
• 新增了代理别名通配符支持，现在支持通过*匹配多字符和?匹配单字符

🔧 技术优化

• 调整了 api 文档，以适配新的 MCP 代理
• 调整模型配置时，导入按钮从不可见变为禁用状态，以优化时间

🐞 Bug 修复

• 修复了对话界面有些工具无法展开的 bug

ChatSpeed v1.1.13

This update focuses on improving the stability and feature expansion of the MCP proxy, while optimizing assistant features and content extraction capabilities. Regarding the MCP proxy server, we have introduced the Streamable HTTP protocol to solve session loss issues caused by network disconnections and computer hibernation. For the scraper module, we have fixed critical stability issues and adopted Readability to improve content area recognition accuracy, reducing unnecessary token consumption from junk information.

✨ Enhancements

• Added openai reference information extension field: Added an extension field for reference information to the OpenAI-compatible interface, providing richer reference data.
• Optimized assistant quick send information feature: Optimized the experience of quickly sending information with the assistant (default shortcut ALT+Z) to improve operational efficiency.
• Added Streamable HTTP proxy server protocol: Added the Streamable HTTP proxy server protocol with endpoint /mcp/http, providing a more flexible connection method.
• Implemented Streamable HTTP Session Persistence and Automatic Recovery: By introducing an advanced session management mechanism, we have achieved disk persistence and automatic recovery for Streamable HTTP sessions. This completely resolves the MCP service session ID loss issue (manifesting as 410 or 401 errors) caused by network disconnections, computer hibernation, or proxy server restarts, significantly enhancing connection stability and user experience.
• Optimized scraper module content extraction: Optimized the scraper module content extraction algorithm. The current version can accurately identify content areas when extracting web content, with a hit rate improved to 90% ~ 95%.
• Refactored AI conversation mechanism: Refactored the AI conversation mechanism, making the interface better able to display the tool calling process; optimized system prompts to make AI better able to use tools to complete daily conversation tasks.
• Updated documentation links: Updated the blog link and English documentation link in README to ensure users can access the latest documentation resources.

🔧 Technical Optimizations

• Cleaned up invalid i18n: Cleaned up a large amount of invalid internationalization resources, reducing application size and improving performance.
• Assistant window size adjustment: Adjusted the assistant window size and implemented automatic restoration based on user size adjustments, enhancing user experience.
• Archived Rust-implemented ReAct module: Based on workflow module adjustments, archived the original Rust-implemented ReAct module to prepare for the new workflow architecture.
• Adopted Readability to improve content recognition accuracy: The scraper module adopted the Readability algorithm to improve content area recognition accuracy, reducing unnecessary token consumption from junk information.

🐞 Bug Fixes

• Fixed scraper module event listener crash issue: Fixed the issue where scraper module event listeners could cause the scraper module to crash, significantly improving scraper functionality stability.
• Removed unused imports and dependencies: Cleaned up unused imports and dependencies in the project, optimizing project structure and build efficiency.

---

本次更新专注于提升 MCP 代理的稳定性和功能扩展，同时优化了助手功能和内容提取能力。MCP 代理服务器方面，我们引入了 Streamable HTTP 协议解决了网络断连和电脑休眠导致的会话丢失问题。爬虫模块方面，我们修复了影响爬虫模块稳定性的关键问题，内容提取采用了 Readability 来提高内容区域识别准确性，减少垃圾信息带来的 token 不必要消耗。

✨ 功能增强

• 新增 openai 引用信息拓展字段：为 OpenAI 兼容接口增加了引用信息的拓展字段，提供更丰富的引用数据。
• 优化助手快速发送信息功能：优化了助手快速发送信息的体验（默认快捷键 ALT+Z），提升操作效率。
• 新增 Streamable HTTP 代理服务器协议：新增 Streamable HTTP 代理服务器协议，入口端点为 /mcp/http，提供了更灵活的连接方式。
• 实现 Streamable HTTP 会话持久化与自动恢复：通过引入先进的会话管理机制，实现了 Streamable HTTP 会话的磁盘持久化和自动恢复。这彻底解决了因网络断开、电脑休眠或代理服务器重启导致的 MCP 服务会话 ID 丢失（表现为 410 或 401 错误）问题，显著提升了连接的稳定性和用户体验。
• 优化爬虫模块内容提取：优化了爬虫模块内容提取算法，当前版本在提取网页内容时可以精准识别内容区域，命中率提升至 90% ~ 95%。
• 重构 AI 对话机制：重构了 AI 对话机制，现在的界面可以更好地展示工具调用过程；优化了系统提示词，使得 AI 可以更好地利用工具完成日常对话任务。
• 更新文档链接：更新了 README 中的博客链接和英文文档链接，确保用户可以访问最新的文档资源。

🔧 技术优化

• 清理无效的 i18n：清理了大量无效的国际化资源，减少应用体积并提升性能。
• 助手窗口大小调整：调整了助手窗口大小，并实现自动根据用户调整大小进行还原，提升用户体验。
• 归档 Rust 实现的 ReAct 模块：基于工作流模块调整，归档了原先 Rust 实现的 ReAct 模块，为新的工作流架构做准备。
• 采用 Readability 提高内容识别准确性：爬虫模块我们采用了 Readability 算法来提高内容区域识别准确性，减少垃圾信息带来的 token 不必要消耗。

🐞 Bug 修复

• 修复 scraper 模块事件监听崩溃问题：修复了 scraper 模块事件监听可能导致爬虫模块崩溃的问题，显著提升了爬虫功能的稳定性。
• 移除未使用的导入和依赖：清理了项目中未使用的导入和依赖，优化了项目结构和构建效率。

ChatSpeed v1.1.12

v1.1.12 is a release focused on core stability and key user experience optimizations. We have fixed underlying issues that could cause application crashes and improved the assistant and application startup workflows to be smoother and more intuitive.

✨ Enhancements

• Optimized Assistant Shortcut Workflow: The assistant shortcut (default Alt+Z) now directly sends text from the clipboard for processing, eliminating the need for an extra confirmation step (like pressing Enter) and significantly improving operational efficiency.
• Improved First-Launch Experience: To provide a more welcoming initial interface, the application now automatically loads the default conversation on first launch instead of displaying a blank page.

🐞 Bug Fixes

• Fixed Assistant Shortcut Double-Trigger Issue: Resolved a bug where the assistant shortcut could trigger twice in quick succession under certain conditions, causing the same message to be sent to the AI repeatedly.
• Fixed Scraper Module Crash: Addressed a critical issue where improper handling by an event listener in the scraper module could cause the application to panic, enhancing the stability of the web content fetching feature.

🔧 Technical Optimizations

• Optimized Database Initialization Logic: Removed the hardcoded creation time for the initial default conversation. The timestamp is now dynamically generated when the database is first created, making the logic more robust.

---

v1.1.12 版本专注于提升核心稳定性和优化关键用户体验。我们修复了可能导致程序崩溃的底层问题，并改进了助手和应用启动的工作流程，使其更加流畅和直观。

✨ 功能增强

• 优化助手快捷键工作流: 现在，使用助手快捷键（默认为 Alt+Z）会直接发送粘贴板中的文本进行处理，无需再次按回车键确认，显著提升了操作效率。
• 改善应用首次启动体验: 为了提供更友好的初始界面，应用在首次启动时会自动加载默认会话，而不是显示空白页。

🐞 Bug 修复

• 修复助手快捷键重复触发问题: 解决了在某些情况下，助手快捷键会连续触发两次，导致同一条消息被重复发送给 AI 的问题。
• 修复采集器（Scraper）模块崩溃问题: 修复了因 scraper 模块中的事件监听器处理不当而可能引发程序崩溃（panic）的严重问题，提升了网页内容抓取功能的稳定性。

🔧 技术优化

• 优化数据库初始化逻辑: 移除了为初始会话硬编码的创建时间，改为在数据库首次创建时动态生成，使逻辑更加严谨。