AI Coding Assistant Lingma:Qoder CN changelogs

Code Review Agent: Supports the /code-review command with a built-in Code Review Agent that triggers code review within a specified scope in Agent mode.

Browser Agent: Supports the /browser command with a built-in Browser Agent that accesses web pages and reads online documentation in Agent mode.

Hook Mechanism: Introduces a new Hook mechanism into the Agent execution flow, allowing custom logic to be inserted at key execution points. Supports UserPromptSubmit, PreToolUse, PostToolUse, PostToolUseFailure, and Stop events.

Clearer conversation revert feedback: Improved the Revert interaction to help users better understand the revert scope and current state.

Fixed known issues.

DevContainer Support: Added DevContainer capability, enabling standardized development environments via .devcontainer/devcontainer.json configuration for consistent team development experiences.

Windows Terminal Sandbox for Agent Mode: Agent-executed commands now run in a secure sandbox, effectively isolating potentially risky operations.

Extension Host Process Isolation: Improved process isolation for built-in AI coding extensions by running them in a separate Extension Host process, enhancing runtime stability and overall user experience.

Agent Sessions in Separate Windows: Supports opening Agent sessions in a new window for more efficient multi-session task handling.

Quest Mode Re-editing Sent Messages: In Quest mode, you can now re-edit previously sent messages for greater conversational flexibility.

AKSK Authentication: Supports login via Access Key/Secret Key.

File Search in Spec/Plan Preview: The Spec/plan preview panel now includes file search for quickly locating target files.

Skill UI Dynamic Interface Generation: When the Agent executes a Skill, interactive HTML components can be rendered in real-time within the conversation, supporting forms, charts, configuration panels, and other rich interaction scenarios. (First-time use requires the Agent to create an interface for the corresponding Skill.)

Quest Mode Artifact Navigation Upgraded: Improved the navigation bar and file tree hierarchy in the artifact area.

Improved Quest mode experience when no project is open, making it easier to find historical tasks.

Repo Wiki improved error code hints.

Fixed issues related to problem reporting.

Quest Mode Artifact Files — Add to Chat: Files in the Quest mode artifact area now support fine-grained modification via "Add to chat". Select content and add it to the conversation; works for both code and spec files.

Quest mode task list now sorted by most recent conversation in descending order.

Lingma Enterprise Edition improved IDE marketplace network connectivity.

Added BYOK configuration entry for third-party models in the model selector.

Custom Model Support: Lingma Pro supports connecting to Alibaba Cloud Bailian, Zhipu, Kimi, and MiniMax coding models via API keys.

Agent Mode Cross-Workspace Search: Agent mode now supports code search and file editing across different projects in the workspace.

VS Code kernel upgraded to 1.106.3.

Repo Wiki Upgrade: Visual generation progress; improved operation guidance with separate generate/update flows; bidirectional sync with import/export descriptions and auto-export capability; Git-linked projects auto-display Commit IDs after generation; reference sources now include line numbers; new user activity display showing remaining free generation count. (Note: Repo Wiki is currently in Beta for Enterprise Edition.)

Input box now supports adding up to 20 images and 20 files.

Quest mode file tree supports right-click operations: add to context, copy relative path, open in folder.

Improved Agent mode interaction for empty projects.

Visual Management for Skills and Custom Agents: Provides a visual management interface for skills, custom agents, and commands, supporting auto-creation and import operations.

Extended custom agent configuration options — now supports assigning skills to custom agents.

Improved Quest mode code file opening experience in the artifact area.

Built-in create-skill and create-agent: Built-in skills to guide and simplify skill and custom agent creation. Also supports invoking skills and subagents via the / command.

Quest Mode File Tree: Quest mode now includes a file tree for browsing project files and adding them to context.

Vercel Skill in Quest Mode: Quest mode supports invoking the Vercel Skill via conversation for quick application deployment.

One-Click Prompt Optimization: Improved handling of prompts that contain @context references.

.agents/skills Directory Support: Now supports reading skills from the .agents/skills directory, following industry standards.

Custom subagent execution details now support expandable floating windows.

NEXT now supports reading clipboard history.

Enhanced Agent skill invocation capability in Quest tasks.

Improved Quest mode new task and toolbar collapse experience.

Updated Quest mode homepage examples.

Login stability improvements. (If you encounter login or permission issues after upgrading, try logging out and back in.)

Fixed occasional system exception.

Fixed occasional file editing hang.

Fixed IDE crash when context includes images.

Improved handling logic when tools are not enabled in Ask mode.

Qwen-Coder-Qoder Model Upgrade: A deeply customized model built for end-to-end programming experiences. Based on Qwen-Coder with large-scale reinforcement learning for the Qoder Agent framework, tools, and scenarios. In real-world software engineering benchmarks, its task resolution rate surpasses Cursor Composer-1, with a 50% improvement in terminal command accuracy on Windows.

Quest Autonomous Programming (Beta): An autonomous programming agent that delivers high-quality, end-to-end, deliverable results without continuous human intervention. Includes: intent recognition and requirement clarification; spec co-creation before execution; long-running task capability with agent supervision; self-verification and quality assurance; continuous self-evolution with style learning; intent-based routing to the most suitable capability; spec-driven development for traceable results; and 0-to-1 website and prototype creation integrated with top-tier visual models.

Agentic Chat — Multi-Agent Parallel Execution: Built-in planning agent for complex tasks with human-in-the-loop planning; upgraded custom extensions supporting subagents, skills, and commands; multi-session parallel conversation windows for handling multiple tasks simultaneously.

Qwen-Coder-Qoder Model: A deeply customized model built for the Qoder end-to-end programming experience. Based on Qwen-Coder with large-scale reinforcement learning for the Qoder Agent framework, tools, and scenarios. In real-world software engineering benchmarks, its task resolution rate surpasses Cursor Composer-1, with a 50% improvement in terminal command accuracy on Windows.

Completions and NES Upgraded to NEXT: From passive completion to proactive prediction, supporting features like automatic import insertion — a comprehensive upgrade in both capability and interaction experience.

Supports manual triggering of NEXT code completions.

Full Database Support (@database): Supports referencing databases as context in Qoder Ask/Agent mode.

Supports generating and executing SQL in Query Console.

AGENTS.md Compatibility: When an AGENTS.md file exists in the current project root directory, the Agent automatically recognizes and loads its contents.

Supports .aiignore.md extension for file restriction configuration.

Allows editing files outside the project via tools.

Various product experience improvements.

Supports IntelliJ 2026.1 EAP+ versions.

Fixed various known issues.

Fixed some known defects.

Fixed occasional abnormal CPU and memory consumption.

@openFiles Annotation for Context: Add all open files at once, including project files, SDK files, and scratch files.

Added tool-enhanced real-time awareness of editor file content, improving Agent effectiveness.

Fixed some known defects.

Fixed defects.

Fixed login state loss issue.

Fixed system paste not working in CLion and Rider.

Fixed occasional failure of the one-click fix feature.

Fixed occasional issue where the default AI response language switched to English.

Supports enterprise model selection.

Fixed various defects.

Fixed defects.

Fixed defects.

Fixed custom instruction execution exceptions in specific scenarios.

Other defect fixes.

Improved file editing permission control for Enterprise Edition.

Fixed defects.

Upgraded Coding Capabilities: Powered by the latest Qwen model, supports parallel tool calls and deeply optimized coding — complex coding tasks are now executed more efficiently and accurately.

To-Dos Tool: The coding Agent can now perform structured to-do breakdown before executing coding tasks, enabling more effective handling of complex, long-running tasks.

Message Queuing: While the Agent is replying or executing, you can edit and send follow-up messages directly — no need to wait, enabling rapid continuous iteration.

Full Auto-Execution Mode: The coding Agent supports configurable auto-execution mode. When enabled, a terminal command deny list can be configured for safety while significantly improving coding task efficiency.

Enhanced Project Awareness: Optimized Grep, Search Codebase, and other retrieval capabilities for more effective project searches.

Enhanced Memory Awareness: Optimized memory and memory consolidation capabilities — Lingma understands you better over time.

Improved Tool Presentation: De-emphasized the UI presentation of file reading, web search, project search, and directory search tools, while highlighting confirmation actions — helping developers quickly access key information.

Agent mode is now the default for smart conversations.

Adding context in smart conversations now requires typing @ to invoke the selection window.

Ask mode now enables retrieval tool calls by default, with the option to disable in settings.

Supports adding selected code lines from the editor to the conversation as context via right-click.

Supports configuring multiple AI rules for a project with configurable activation methods.

Improved inline chat editing experience, enhancing the file editing workflow.

Removed the model selection entry and file editing mode for a simpler user experience.

Issue reporting now supports adding images.

Fixed defects.

Improved the fallback experience for inline code suggestions.

Fixed defects.

Improved inline suggestion prediction generation quality.

Improved error messages when file modification fails.

Improved error messages for path anomalies.

Fixed occasional plugin installation failure and Kotlin file function shortcut entry not working in K2 mode.

Fixed IDE lag caused by Gradle clean/build in Android Studio.

Fixed occasional input box unresponsiveness.

MCP service now supports Streamable HTTP.

Improved memory generation quality.

Improved MCP service UI experience.

Fixed defects.

Code snippets in smart conversation responses now support one-click apply to code files.

Problem fix now supports Agent mode for repair.

Debug error fix now uses Agent mode for responses.

Fixed various defects.

Improved MCP tool parameter configuration experience.

Improved input box focus indicator style.

Fixed occasional extraction failure for Windows 7 users during installation.

Other defect fixes.

MCP Marketplace now supports English language descriptions.

Fixed Ask mode not respecting language preference when tools are enabled.

Ask mode now supports project and code search, web search, and other tools.

Agent mode now supports the web search tool.

Agent mode now supports selecting file directories (#folder).

Adding context now supports mixed search of project files and folders with search highlighting.

Improved the experience of pasting code snippets from the project into the conversation window.

Compatible with the latest IDEs' Reworked Terminal — no longer shows unsupported prompts.

Supports selecting instructions during conversations.

Improved knowledge base loading speed when selecting #teamDocs.

Next Edit Suggestion (NES): Added inline suggestion prediction feature, supporting next modification prediction. Can be enabled in plugin settings.

MCP tool execution in Agent mode now supports configurable auto-execution.

Improved terminal command auto-execution whitelist experience — supports one-click addition during conversations.

Supports dragging folders from the file list to the conversation input as context.

Improved file modification accuracy — automatically retries when file modification fails.

Improved the experience of pasting code from the project into the conversation input.

Improved code change generation quality in inline chat.

Fixed inline chat not visible when extending beyond the editor area.

Fixed missing completion count in plugin settings.

Added conversation mode selection suggestion tips.

Supports one-click fix when MCP service connection is abnormal.

Improved output quality when using the qwen3-thinking model in Agent mode.

Reduced flowchart generation probability in code explanation commands.

Fixed occasional failure to detect currently open files in Ask mode.

Fixed occasional output compatibility issue when switching between Agent and file editing modes during project search.

Fixed occasional compatibility issue with PyCharm 2025.1 in file editing mode.

Inline Chat: Supports Q&A and code modification directly in the editor area.

Improved large file modification quality in smart conversations.

Tool Usage: Can autonomously use over ten built-in programming tools, such as file read/write, code search, and error investigation. Also supports MCP tool configuration and integrates the ModelScope MCP Marketplace for developer customization.

Terminal Command Execution: When executing coding tasks, the Agent can autonomously decide which commands to run, automatically write and execute terminal commands — significantly improving coding task efficiency.

Project-Level Changes: Based on task descriptions, the Agent can autonomously break down tasks and modify multiple code files within the project. Iterative refinement and snapshot rollbacks are supported through multi-turn conversations.

Automatic Project Awareness: Based on task descriptions, the Agent automatically detects project frameworks, tech stacks, required code files, and error information — no need to manually add project context.

Memory Awareness: Supports LLM-based autonomous memory. During conversations, Lingma gradually builds rich memories about the developer, project, and issues — understanding you better with each interaction.

Supports navigation to files, methods, and definitions within the project from smart conversations.

Adding context in smart conversations now supports adding project directories.

Adding context in smart conversations now supports specifying knowledge bases (Enterprise Edition).

Adding context in smart conversations now supports multi-select interaction for files, directories, and gitCommits.

Fixed defects.

Version isolation between Lingma VS Code and JetBrains plugins. Going forward, the two plugins will maintain independent login states, session history, and data.

Fixed defects.

Fixed defects.

Supports configuring multiple AI rules for a project.

Improved MCP service header configuration experience.

Supports clicking to view modified files in conversation responses.

Improved code completion speed.

Improved error messages for path anomalies.

Improved context search experience — now supports simultaneous search of files and directories.

Supports adding open files as context.

MCP service now supports Streamable HTTP.

Improved memory generation quality.

Improved MCP service UI experience.

Fixed defects.

Problem fix now supports Agent mode for responses.

Fixed various defects.

Supports dragging files and directories to the conversation input as context by holding Shift (VS Code 1.100 or later).

Improved MCP tool parameter configuration experience.

Improved the experience of pasting code from the project into the conversation input.

Fixed various defects.

MCP Marketplace now supports English language descriptions.

Fixed Ask mode not respecting language preference when tools are enabled.

Ask mode now supports project and code search, web search, and other tools.

Agent mode now supports the web search tool.

Agent mode now supports selecting file directories (#folder).

Improved personal settings page loading performance.

Fixed occasional account error in SSH Remote scenarios.

Fixed occasional failure to recognize pasted images.

MCP tool execution in Agent mode now supports configurable auto-execution.

Improved terminal command auto-execution whitelist experience — supports one-click addition during conversations.

Improved file modification accuracy — automatically retries when file modification fails.

Improved code change generation quality in inline chat.

Added conversation mode selection suggestion tips.

Supports one-click fix when MCP service connection is abnormal.

Improved output quality when using the qwen3-thinking model in Agent mode.

Reduced flowchart generation probability in code explanation commands.

Fixed occasional output compatibility issue when switching between Agent and file editing modes during project search.

Improved large file modification quality in smart conversations.

Inline Edit: Added inline edit functionality, invocable via ⌘⇧I (Ctrl+Shift+I).

Fixed delayed list update when MCP service JSON file changes.

Fixed occasional inability to see action buttons during MCP tool execution.

Fixed personal settings page display issue on Windows.

Tool Usage: Can autonomously use over ten built-in programming tools, such as file read/write, code search, and error investigation. Also supports MCP tool configuration and integrates the ModelScope MCP Marketplace for developer customization.

Terminal Command Execution: When executing coding tasks, the Agent can autonomously decide which commands to run, automatically write and execute terminal commands — significantly improving coding task efficiency.

Project-Level Changes: Based on task descriptions, the Agent can autonomously break down tasks and modify multiple code files within the project. Iterative refinement and snapshot rollbacks are supported through multi-turn conversations.

Automatic Project Awareness: Based on task descriptions, the Agent automatically detects project frameworks, tech stacks, required code files, and error information — no need to manually add project context.

Memory Awareness: Supports LLM-based autonomous memory. During conversations, Lingma gradually builds rich memories about the developer, project, and issues — understanding you better with each interaction.

Supports navigation to files, methods, and definitions within the project from smart conversations.

Adding context in smart conversations now supports adding project directories.

Adding context in smart conversations now supports specifying knowledge bases (Enterprise Edition).

Adding context in smart conversations now supports multi-select interaction for files, directories, and gitCommits.

Supports Alibaba Cloud International Site account login.

Q&A input box now supports Shift + Enter for line breaks.

Improved search box focus experience when selecting context.

Removed the offline completion mode quick toggle from the status bar.

Supports International Site Enterprise Edition developer login.

When typing in the Q&A input box, recommends related code files based on selected files.

Supports Project Rules configuration.

Supports one-click issue feedback on startup errors.

Using Lingma on the ModelScope platform now supports web login.

Enhanced Go language inline code completion quality.

Fixed various defects.

Plugin now supports switching between Personal and Enterprise Edition identities.

Improved inline completion for Java inner class calls.

Improved inline completion similarity awareness.

Improved auto-import of context when selecting input history in the input box.

Improved full-file awareness when selecting code snippets for Q&A.

Improved model response timeout issues.

Fixed model selection not clickable in certain scenarios.

Smart Q&A and AI Developer now support model selection.

Enhanced AI Developer memory of ongoing conversation history.

Mermaid diagrams now render on demand by default, with one-click collapse support.

Code explanation responses are now more concise.

Improved application speed for AI Developer fragment modifications in large files.

Improved file path when AI Developer creates new files.

Fixed image Q&A failing when the image was not fully processed.

Fixed occasional multiple triggers when the cursor is between tags during completion.

Supports one-click submission of plugin error logs.

Improved startup issue resolution experience — supports one-click fix.

Improved code file generation from images in AI Developer.

Enhanced SQL generation in code completion.

Improved code generation from images in Smart Q&A.

AI Developer now supports generating code file changes from images.

Enterprise Edition accounts now support web login in SSH Remote scenarios.

Fixed occasional language detection failure when generating files in empty projects.

AI Developer: Added multi-file code modification and tool usage capabilities. The AI Developer can help you end-to-end complete coding tasks such as feature implementation, bug fixing, unit test generation, and batch code changes. Also introduced an upgraded interaction interface with a unified file change management view and code change review view — simply describe your requirements in natural language, and the AI Developer will modify multiple files for you to review and adopt.

Continued enhancement of Java and Go cross-file awareness.

Improved Java and Go cross-file completion quality.

Enterprise code generation enhancement now supports C/C++.

Fixed inability to select .ipynb files when choosing context files.

Smart Q&A now supports codeChanges and gitCommit as context.

Multimodal image Q&A now supports copying and pasting images.

Simplified smart Q&A context selection interaction.

Version isolation between Qoder CN VS Code and JetBrains plugins. Going forward, the two plugins will maintain independent login states, session history, and data.

Supports configuring Project Rules.

Added model selection history.

Fixed Q&A window line breaks overlapping the model selection area.

Fixed model occasionally resetting.

Fixed overlapping smart completions in some C++ projects.

Q&A window input box now supports Shift + Enter for line breaks.

Q&A window now supports model selection.

Fixed Q&A window sending reminder text as messages.

Fixed incorrect Q&A state when right-click context menu is triggered without code selection.

Fixed potential crash scenarios.

Fixed known issues with smart completion, improving the overall completion experience.

Supports triggering optimize code and generate unit tests via Q&A input / command or editor right-click menu.

Improved Enterprise Edition domain address input logic.

Improved code block interaction in the Q&A window — supports select-all highlighting and copying.

Supports triggering code explanation and comment generation via Q&A input / command or editor right-click menu.

Fixed incorrect C# syntax highlighting in the Q&A window.

Fixed settings panel loading error.

Q&A window input now supports multi-line text with line breaks.

Fixed unbalanced brackets in if statements during code completion in certain scenarios.

Fixed login prompt bar not appearing in certain situations.

Improved dynamic response logic in code completion scenarios.

Q&A window now supports output link display.

Fixed inability to stop conversations in the Q&A window.

Supports syntax highlighting in dark theme.

Q&A window now supports syntax highlighting for Markdown, SQL, C++, and more languages.

Intermediate versions primarily focused on user-reported issue fixes and quality improvements.

Lingma Visual Studio 2019 Plugin Released: Supports line-level/function-level real-time code continuation, natural language to code generation, and development smart Q&A.

Added settings window with login/logout support.

Fixed manual trigger shortcut not working in certain scenarios.

Fixed Markdown rendering issues.

Supports customizing keyboard shortcuts for accepting inline code suggestions.

Fixed occasional misalignment in inline code generation and editor event preemption by other plugins.

Lingma Visual Studio 2022 Plugin Released: Supports line-level/function-level real-time code continuation, natural language to code generation, and development smart Q&A.