Everywhere v0.7.4

🐛 Bug Fixes

• 🔥 Fixed an issue where multi-turn conversations would fail when both "Tool Calling" and "Deep Thinking" were enabled simultaneously. This previously caused models like DeepSeek and Kimi to fail functioning properly.
• Fixed an issue where the automatic selection of the "System Assistant" might fail to work in certain situations.
• Fixed an issue where the items parameter in the "Manage Todo List" tool was incorrectly set as required, which previously caused the model's initial tool call to fail and led to confusion.

Full Changelog: v0.7.3...v0.7.4

Everywhere v0.7.3

✨ Features

• 🔥Added DeepSeek V4 Pro and DeepSeek V4 Flash to the preset mode, and removed the outdated DeepSeek Chat and DeepSeek Reasoner models.
• Added an "Disabled" level to the "Reasoning Effort" option in the chat window. For supported models (such as DeepSeek V4, certain Claude models, Xiaomi Mimo, etc.), you can now use this level to completely disable the model's deep thinking process.

🐛 Bug Fixes

• 🔥 Fixed an issue where the chat window could not be brought up via shortcuts when the "Automatically Add Element" option was disabled. Special thanks to all the users who helped us troubleshoot and isolate this issue! (#345, #350, #354)
• Fixed an issue where the send button remains disabled after selecting a strategy, forcing users to input text before being able to send.
• Fixed an issue where generic strategies failed to display in the list.
• Fixed an issue where the input box was not correctly populated with the original text when editing a user message.
• Fixed an error that occurred when Gemini models attempted to execute parallel tool calls.
• Fixed an issue where auto-generated chat topics would occasionally appear as garbled text under specific circumstances.

Everywhere v0.7.2

✨ Features & 🚀 Improvements

• 🔥 Refactored the global hotkey listener logic, which is expected to resolve the issue of occasional hotkey listener failures.
• Optimized chat attachment handling logic: When the selected assistant does not support the current file format, the app now allows adding it with a warning prompt. Upon sending, the assistant will receive the attachment's metadata and be prompted to process it using appropriate tools.
• Optimized the settings loading logic upon application startup: If the configuration file encounters an error, it will now be automatically backed up and preserved. Also added an automatic error fallback mechanism for invalid enum values.
• Added Kimi K2.6 to the list of preset models.
• Improved the error handling logic and error messages during software updates and Everywhere Cloud Services login processes.
• Added handling mechanisms for network request timeouts to prevent potential errors.
• (Windows) Added digital signatures to .dll files to reduce the risk of false positives from antivirus software.

🐛 Bug Fixes

• Fixed an issue where the Xiaomi Mimo model could not correctly enable deep thinking (#282).
• Fixed an issue where the chat would interrupt if the model still attempted to call tools after tool execution permissions were disabled.
• Fixed an issue where the chat would interrupt if the assistant attempted to call a non-existent tool.
• Fixed an issue where the chat would interrupt due to thread contention issues.
• Fixed an issue that caused errors when adding a visual element attachment if the target program unexpectedly exited.
• (Windows) Fixed an issue where chat notifications failed to display normally under certain conditions, subsequently causing the chat to interrupt.
• (macOS) Fixed an issue where the global hotkey listener failed to work.
• (macOS) Fixed an issue where the file selection dialog could not be opened in some situations.

Full Changelog: v0.7.1...v0.7.2

Everywhere v0.7.1

⚠️ This update fixes many issues from the 0.7.0 release, including critical crashes. We highly recommend all users to update immediately.
⚠️ Due to toast display bug, in-app upgrade is not working. Please download here or from official website.

✨ Features & 🚀 Improvements

• Added Claude Opus 4.7 and more latest models in the preset mode, and corrected some previously misconfigured parameters.
• Optimized the local file cache by adding proper file extensions for easier local management and inspection.
• Polished text descriptions within the UI and added more quick navigation links.

🐛 Bug Fixes

• 🔥 (Windows): Fixed a crash occurring on some systems when bringing up the chat window via shortcuts, which was caused by the "Element Fly-in Effect" (#344).
• 🔥 (Windows) Refactored the visual element selector, completely resolving the issue where using global shortcuts to select elements could cause modifier keys like Ctrl or Alt to get stuck.
• 🔥 (Windows): Fixed an issue where the chat window would disappear immediately when using shortcuts to pick visual elements while the window was unpinned and the "Element Fly-in Effect" was enabled.
• Fixed an issue where images and file attachments could not be pasted directly into the chat input area.
• Fixed an issue where pasted images could not be correctly read by models (#349).
• Fixed an issue where the "Topmost When Typing" and "Pin When Typing" options were not taking effect.
• Fixed an issue where in-app toast notifications would sometimes disappear instantly.
• (Windows) Fixed an issue where toggling the chat window's pinned/unpinned state would cause screen flickering, which previously interrupted IME input.
• (macOS) Fixed an issue where file attachments could not be properly selected and added within the chat window.

Full Changelog: v0.7.0...v0.7.1

Everywhere v0.7.0

📢 Overview

This release brings you 2 core features: official Cloud Services and Strategy Engine, fully enhancing the out-of-the-box AI experience and convenient automated task processing capabilities. In addition, this update covers up to 19 new features, 27 improvements, and 15 bug fixes, dedicated to creating a smoother, smarter, and more stable user experience.

☁️ Cloud Services

In 0.7.0, we have integrated official Cloud Services into Everywhere, providing you with a more out-of-the-box AI experience:

• 🧠 Out-of-the-box AI Models: The initial batch of services has integrated large language models from OpenAI, Gemini, Anthropic, and DeepSeek. No API configuration is required; log in to call them directly.
• 💬 Cross-platform Chat Cloud Sync: Added cloud sync for chat history, letting your inspiration flow seamlessly across different devices. (Note: Currently, only text message sync is supported; image and file attachment sync will be updated in future versions.)
• 🔑 BYOK (Bring Your Own Key): If you prefer to bring your own key, the addition of cloud services will not affect your local configuration. You can still use it freely as before.

🎉 Click the avatar in the bottom left corner of Everywhere to register/log in. Complete registration now to receive 500,000 credits. For more service and billing details, please visit the website.

🧩 Strategy Engine

Strategy Engine is now live! With the Strategy Engine, Everywhere can intelligently perceive your context and needs, dispatching assistants with one click to resolve complex tasks.

It currently covers various common scenarios such as browsers, text selection, images, and documents. More practical strategies will continue to be expanded in the future.

In future releases, the Strategy Engine will serve as one of Everywhere's core features, gradually rolling out the following capabilities:

• Custom Strategies: Users can create and configure personalized strategies according to their needs, defining trigger conditions, prompts, and tool lists via STRATEGY.md file to achieve no-code automation.
• Preprocessors: Allows preprocessing inputs before strategy execution, such as context parameter extraction and attachment format conversion, to enhance the applicability and effectiveness of strategies.
• Strategy Workshop: Allows developers and users to share various custom strategies, aiming to build a rich and diverse strategy ecosystem.

🖥️ UI

Revamped main page navigation design and optimized UX

---

Refactored changelog page, allowing you to view the latest release notes before updating

---

Introduced ask_user_question tool

When handling complex tasks, if the assistant encounters uncertainties, it can now actively pause and ask you questions, waiting for you to confirm details before proceeding.

---

✨ Features

🎉 Core Experience

• 🔥 Concurrent Chats Support: Now supports running multiple chats simultaneously. Each session processes inputs and outputs independently (#284).
• 🔥 Custom System Assistant: Chat title generation is now handled by an independent system assistant, automatically selecting the most suitable model in native and preset modes, with fully customizable configuration.
• Editing a user message during a chat will now create a new branch.
• Added MiniMax (minimaxi.com) to the list of preset model providers.

🛠️ Interaction & Workflow

• 🔥 When rejecting a tool call from the assistant, you can now provide a "reason for rejection". This helps the assistant understand your intention and correct its subsequent behavior (#304).
• Added "Always Topmost" and "Topmost When Typing" options to chat window (#295).
• Added global shortcuts for "Pick Visual Element" and "Take Screenshot" (#269).
• Added a "Maximum Context Rounds" option to control context length (#156).

💻 UI & Experience

• 🔥 Added a search function to the assistant's avatar editor and expanded more Emoji options.
• Added support for Mermaid diagram rendering (#218).
• When chat window is in the background or minimized, system will show a notification if a conversation is completed, authorization is needed, or an error occurs.
• Introduced brand new dynamic visual feedback when selecting visual elements and building context (can be disabled in settings).
• Added tokens/s statistics (#323).
• Supported using middle mouse button click to quickly delete attachments in the chat input area.
• Added a toggle option for automatic chat title generation.
• When manually checking for updates, error reasons will be prompted if network or server anomalies occur.
• Added a caching mechanism for the MCP plugin tool list.

🚀 Improvements

• 🔥 Tool execution and confirmation UI are now directly embedded in chat flow rather than interrupting with dialogs. You can switch chats or abort operations at any time.
• 🔥 The assistant's deep thinking and tool execution bubbles now collapse automatically, and unnecessary "Time Elapsed" displays have been removed, making the chat interface cleaner.
• 🔥 The app now filters out unsupported attachments based on input modalities supported by the current model, preventing model errors caused by unsupported attachments like images (#297, #330).
• Improved the performance and stability of Markdown rendering.
• Added supplementary hint texts and images to some pages.
• Moved the "Temporary Chat" button to the title bar of chat window.
• Improved the editing experience when setting up shortcut keys.
• Each conversation now supports sending up to 50 attachments.
• Optimized the visual context construction algorithm, improving performance when handling long content (#300).
• Added more parameter settings for assistants, such as modalities and context input limits; some parameters have been renamed.
• Temperature and Top-p parameters are now precise to two decimal places (#318).
• Optimized the description text and error messages for Temperature and Top-p parameters.
• Improved the stability of inter-process communication, reducing occasional communication errors.
• When encountering issues, in-chat error reporting will display more detailed error messages to help understand and troubleshoot problems.
• Improved the readability and instructiveness of numerous error messages.
• Improved the stability of Tavily search engine.
• Added an automatic reconnection mechanism for MCP plugins to avoid 4xx errors caused by long-time connections (#232).
• Optimized format compatibility when importing MCP tools from JSON.
• Improved the compatibility of OpenAI Chat Completions.
• Optimized the loading process of system tray icon upon application startup.
• Optimized time parameters in system prompts to increase model cache hit rate.
• (Windows) Added an official digital signature to the installer to enhance security and reduce false positives from antivirus software.
• (Windows) Optimized the textual description for "Startup as Administrator" (#326).
• (macOS) Strengthened the security of system tools to prevent potential command template injection risks.

🐛 Bug Fixes

• Fixed an issue where files could not be dragged into the chat window while the assistant was generating a reply.
• Fixed an issue where chat titles sometimes failed to generate automatically.
• Fixed an issue where the chat history list would sometimes freeze, preventing past records from loading.
• Fixed an issue where the chat window could not be brought up by shortcuts in specific scenarios.
• Fixed an occasional crash issue that could occur when deleting chat history.
• Fixed an issue where assistant configurations did not take effect correctly upon startup.
• Fixed an issue where the app would sometimes incorrectly prevent the computer from shutting down normally.
• Fixed an issue where Ollama models did not display deep thinking content.
• Fixed an issue where, in rare cases, tool calls from Gemini models might have parameter errors.
• Fixed an issue where the "Copy Logs to Clipboard" button on the MCP plugin page was broken.
• Fixed an issue where the reasoning_content field was not correctly processed for some models.
• (Windows) Fixed an issue where the element selector displayed an offset when multiple monitors were connected.
• (Windows) Fixed an issue where the selected range did not match the actual position when picking visual elements.
• (Windows) Fixed an issue where capturing screen elements could cause the software to crash in certain situations.
• (Windows) Fixed an issue that could trigger an ERROR_NO_ASSOCIATION (No associated program found) error when opening or editing settings files (#333).
• Other various minor fixes and stability improvements.

Full Changelog: v0.6.7...v0.7.0

Everywhere v0.6.7

✨ Features

• Visual Context Supercharged: Introducing get_visual_tree tool. Your assistant can now call this tool to fetch on-screen elements freely, massively boosting its contextual comprehension and self-awareness capabilities. Note: This is an experimental feature and may not work perfectly with all secenarios. Feedback is highly appreciated!
• Model Update: Removed the deprecated Gemini 3 Pro Preview model and added the newly released Gemini 3.1 Pro Preview.

🚀 Improvements

• Framework Upgrade: Updated the underlying MCP (Model Context Protocol) framework version, which may resolve previously reported edge-case issues (#232).

🐛 Bug Fixes

• Critical Context Fix (Windows): Fixed a fatal typo in the visual context extraction algorithm on Windows. This bug previously caused premature truncation of complex contexts, hiding crucial information below the selected element from the assistant.
• Resolved a rendering crash that could occur after enabling the text selection feature (#313).
• Fixed a random crash bug related to Grid layout calculations, bringing further stability to the application.

Full Changelog: v0.6.6...v0.6.7

Everywhere v0.6.6

✨ Features

• Enhanced the OpenAI schema and deprecated the standalone DeepSeek schema to improve compatibility with dynamic reasoning models.

🚀 Improvements

• Optimized the installation process for the built-in Puppeteer.
• Removed "OpenAI" from the UserAgent string to prevent requests from being blocked by Cloudflare WAFs.
• Refreshed the loading animation style for better experience.

🐛 Bug Fixes

• Fixed an issue preventing tool calls from functioning correctly with models like Kimi 2.5 under the OpenAI schema.
• Fixed permission validation issues for chat plugins.
• Fixed a bug where sub-agents could recursively call themselves.
• Fixed Anthropic URL parsing; appending # to the URL now forces the use of the raw address.
• (macOS) Fixed an issue where the built-in Puppeteer failed to install.

Full Changelog: v0.6.5...v0.6.6

Everywhere v0.6.5

✨ Features

• Model Updates: Added Claude Opus 4.6 and Haiku 3, setting Sonnet 4.5 as default; Google now defaults to Gemini 3 Flash Preview; DeepSeek now defaults to Deepseek Reasoner; OpenRouter added Kimi 2.5 and Gemini 3 Flash Preview, while upgrading Grok to v4.1.
• Auto-Approval for Tools: Added management for auto-agreeing tool execution; terminal-based tools remain manual for security reasons (#292).
• Text Selection Enhancements (Windows): Improved selection stability, prevented selection results from polluting clipboard history, and resolved the issue where Ctrl+C was sent when used with terminal applications (#281).
• Model URL Preview: Added the ability to preview model URLs to improve transparency and user experience.

🚀 Improvements

• Refined error messages for unsupported image inputs and context length limit violations.
• Optimized the loading animations within the chat window.

🐛 Bug Fixes

• Fixed the OpenAI.ChangeTrackingList type mismatch error (#171).

Full Changelog: v0.6.4...v0.6.5

Everywhere v0.6.4

✨ Features

• Added reasoning effort settings to the chat window
• Replaced the visual context length limit with predefined tiers for a more user-friendly experience

🚀 Improvements

• Improved the logic for scenarios where no custom assistant is selected by adding hints and preventing input loss
• Enhanced telemetry with the inclusion of metrics data

Full Changelog: v0.6.3...v0.6.4

Everywhere v0.6.3

✨ Features

• Added confirmation and undo functionality when deleting chat history
• Added MCP options to the tool menu in the chat window

🚀 Improvements

• Optimized the layout of the chat tools page
• Other general UI/UX updates and enhancements

🐛 Bug Fixes

• Fixed an issue where DeepSeek reasoning models caused a 400 error when calling tools after outputting body content (#287)

Full Changelog: v0.6.2...v0.6.3