Justin Chen

This commit introduces an **enhancement** to the **Chat UI** to provide clearer status updates for subagents. It modifies the **subagent content part** to automatically update the displayed description of a subagent from a generic "running" state to "Ran subagent" once its execution is complete, but only if no custom description was initially provided. This logic, implemented in `chatSubagentContentPart.ts` via methods like `markAsInactive` and `update`, ensures that users receive more explicit feedback on subagent completion. New tests in `chatSubagentContentPart.test.ts` verify this behavior, improving the overall clarity of the **chat experience** when interacting with subagents.

This commit introduces a **bug fix** within the **Chat editing** functionality, specifically addressing the behavior when **restoring chat snapshots**. It adds crucial logic to `src/vs/workbench/contrib/chat/browser/chatEditing/chatEditingActions.ts` to **cancel any currently active chat request** via the `restoreSnapshotWithConfirmationByRequestId` function. This prevents potential conflicts or stale states that could arise from an ongoing request interfering with the restored chat context, thereby ensuring a more consistent and reliable **user experience** within the **Chat feature**.

This commit provides a **bug fix** for the **terminal chat agent tools**, specifically addressing an issue where markdown was not rendering correctly in terminal invocation messages. It **removes the redundant `escapeMarkdownSyntaxTokens` function call** from `src/vs/workbench/contrib/terminalContrib/chatAgentTools/browser/tools/runInTerminalTool.ts`, which was inadvertently causing improper markdown rendering. This ensures that commands generated by the chat agent and displayed in the terminal are presented with their intended markdown formatting, improving the clarity and readability of these interactions. The change directly impacts the **user experience when interacting with AI-generated terminal commands**.

This commit introduces a **new capability** to enhance the **chat user experience** by refining how the first request in a chat session is handled. It defines a new context key, `isFirstRequest`, to accurately identify the initial interaction, which is then used to adjust the availability of **chat actions**. Specifically, **`chatForkActions`** now exclude the first request, and the **`Restore Checkpoint` action**'s availability is modified for better usability within the **`chatEditing`** module. Furthermore, a **new `Start Over` action** is added to `chatEditingActions.ts`, allowing users to easily clear the chat and begin anew, making the initial chat state more "friendly."

This commit implements a **bug fix** within the **Copilot extension**, specifically addressing how **tool calls** handle image content. It introduces a crucial check in the `onImage` method within `extensions/copilot/src/extension/prompts/node/panel/toolCalling.tsx`. This **defensive enhancement** prevents the system from attempting to process image data if the current model or organization does not support vision capabilities. The change ensures **graceful handling** of unsupported visual inputs, improving the overall **stability and robustness** of the tool calling feature.

This commit **fixes** an issue in the **Chat UI** by preventing the **Autopilot handoff** suggestion from being displayed when its corresponding policy is disabled. Specifically, it modifies the `_updateHandoffDisplay` function in `src/vs/workbench/contrib/chat/browser/widget/chatContentParts/chatSuggestNextWidget.ts` to include a policy restriction check. This **UI improvement** ensures that the handoff option only appears when genuinely available, leading to a more accurate and less cluttered user experience within the **workbench's chat feature**.

This commit **fixes** a serialization issue within the **Chat** feature's session operation logging. It **adjusts the serialization logic** in `src/vs/workbench/contrib/chat/common/model/chatSessionOperationLog.ts` to ensure that "thinking" responses are handled **dynamically**. This **refinement** prevents incorrect or static serialization of intermediate chat states, thereby improving the accuracy and robustness of chat session logging and state management.

This commit introduces a **new capability** to the **chat attachment system**, providing a **user-friendly warning** when more than `MAX_IMAGES_PER_REQUEST` (currently 20) images are attached to a single chat request. It enhances the **chat input and display UI** by implementing logic within `chatInputPart.ts` to identify and mark excessive image attachments. The system now defines an `ImageLimitExceeded` state and uses `chatAttachmentWidgets.ts` to display a clear warning, improving the overall **chat user experience** by informing users about attachment limits. This prevents users from inadvertently sending an excessive number of images in a single request.

This commit significantly **improves chat session checkpointing and restoration**, particularly for user-attached context and files. It introduces the ability to **restore attachments to the chat input**, including re-fetching image data, when navigating chat history or restoring a session snapshot. The underlying **chat model** and operation logs are refactored to correctly serialize and deserialize these attachments, ensuring a more robust and complete restoration experience. This **grow** work also lays the groundwork for a **checkpoint hover on first request**, enhancing the user experience by providing better feedback on session state.

This **maintenance** commit significantly improves the user experience within the **Chat UI** by expanding the visible content area for tool call confirmations. It **increases the height threshold** in `chatToolConfirmationSubPart.ts` before content is truncated, allowing more information to be displayed by default. Concurrently, styling adjustments in `chat.css` increase the maximum height of the message wrapper and refine the 'see more' link's positioning. This ensures users can view a larger portion of tool invocation details directly, enhancing readability and reducing unnecessary interactions.

This commit provides a **bug fix** for rendering issues within the **Chat UI**, specifically addressing how "thinking" content and lazy-loaded markdown are displayed. It refines the `chatThinkingContentPart` to ensure diff statistics are only shown when actual changes exist and guarantees that dynamically generated markdown content is correctly appended to the DOM. Furthermore, the `chatListRenderer` is updated to prevent redundant DOM node re-appending, improving the overall stability and visual accuracy of the **chat experience** during content generation and edits.

This commit **enhances the visual presentation** of the **sessions chat widget** by introducing padding to its input box. Specifically, it applies a **styling adjustment** within `chatWidget.css` to improve the spacing and layout of the input field. This is a **UI/UX improvement** that makes the chat interface appear cleaner and more user-friendly, directly affecting the user experience when interacting with the sessions input area.

This commit **enhances the `runInTerminalTool`** within the **terminal chat agent tools** by explicitly defining the user as the audience for certain language model parts. It introduces `audience: [LanguageModelPartAudience.User]` to a configuration object in `src/vs/workbench/contrib/terminalContrib/chatAgentTools/browser/tools/runInTerminalTool.ts`. This **new capability** ensures that image-related content or configurations are correctly scoped and presented to the end-user when commands are executed through the terminal's chat integration. The change improves the precision of audience targeting for features leveraging language models in the terminal.

This commit introduces **image resizing functionality** within the **Copilot extension's prompt panel** to enhance how visual data is handled. It specifically ensures that images received as data parts from **tool calls** are properly resized before further processing. This **new functionality** improves the display and upload readiness of images, guaranteeing they are correctly processed within the prompt panel's UI, particularly through the `imageDataPartToTSX` function.

This commit introduces a **performance enhancement** to the **Chat UI** by implementing a caching mechanism for **LLM-generated thinking headers**. It modifies the `chatThinkingContentPart.ts` module to **store and retrieve these titles** from local storage, preventing redundant LLM calls when re-rendering chat sessions. This **new capability** ensures that previously generated thinking headers are quickly available, improving the responsiveness and consistency of the chat experience. The caching is specifically applied to **non-local sessions**, optimizing performance for remote or persistent chat interactions.

This commit **fixes and improves the persistence** of **chat permission levels** (autopilot modes) across different chat sessions and input changes. It ensures that the selected permission level for the **chat input** is correctly stored within the **chat model** and **session metadata**, and then reliably restored when a session is reloaded or switched. Additionally, a **"Learn More about Permissions" option** has been integrated into the **chat permission picker** for enhanced user guidance. This is a **bug fix and feature enhancement** for the **chat experience**, providing users with reliably preserved autopilot mode settings and easier access to permission information.

This commit introduces a **new capability** to the **Chat UI**, enabling the display of aggregated diff statistics directly within the **thinking header** of chat responses. It enhances the `chatMarkdownContentPart` to emit diff events from collapsed code blocks, which are then collected and passed by the `chatListRenderer` to the `chatThinkingContentPart` for aggregation and presentation. This allows users to quickly see a summary of added and removed lines, providing immediate context on the scope of proposed code changes without needing to expand full diffs. New CSS styles are added for presentation, and comprehensive tests ensure correct aggregation and display.

This commit introduces **new features** and **enhancements** to the **Copilot Autopilot** functionality, improving both user experience and operational reliability. It adds **progress indicators** through `getAutopilotContinuationMessage`, `showAutopilotProgress`, and `resolveAutopilotProgress` to provide better feedback during ongoing operations within the `toolCallingLoop`. Crucially, the commit also implements logic in `ensureAutopilotTools` to guarantee the **`task_complete` tool** is always available, preventing potential operational failures within the **Autopilot** system. This **feature addition** is validated by new unit tests in `toolCallingLoopAutopilot.spec.ts`, ensuring the robustness and transparency of the **Autopilot** system.

This commit introduces a **debounce mechanism** to the **Chat UI's "working" shimmer effect**, specifically within the `chatListRenderer.ts` module. This **enhancement** prevents the shimmer from displaying too briefly or flickering unnecessarily, improving the perceived responsiveness and user experience of the chat interface. By integrating a new helper method, `hasBeenCaughtUpLongEnough`, the system now intelligently controls when the "working" indicator is shown, leading to a smoother visual feedback loop for users.

This commit **refines the Copilot CLI chat experience** by preventing internal sub-agent markdown content from being promoted to top-level messages. It **fixes** the chat history building logic within the `copilotCLITools.ts` module, specifically in the `buildChatHistoryFromEvents` function, to exclude this content. Additionally, message event handlers in `copilotcliSession.ts` are updated to stop streaming sub-agent markdown to the chat UI. This ensures that users only see relevant, high-level responses in the **Copilot CLI chat**, significantly improving clarity and reducing visual clutter by filtering out internal process details. A new test case was also added to validate this exclusion behavior.