feat-internal: Export Chat as Zip with Repository Context

[zhichli]

Adds "Export Chat as Zip" action to package chat sessions with repository context for benchmark creation.

What it does

• Creates zip containing chat.json (session data) and repos/*.json (repo state per workspace folder)
• Captures git state: remote URL, branch, commits, working tree diffs
• Diffs stored as unified patches compatible with git apply

Workspace Scenarios

Scenario	workspaceType	syncStatus	Reproducibility
Synced with remote	remote-git	synced	✅ Clone + checkout
Unpushed commits	remote-git	unpushed	TBD: git bundle
Unpublished branch	remote-git	unpublished	TBD: git bundle
Local-only git	local-git	local-only	TBD: folder bundle
No git	plain-folder	no-git	TBD: folder bundle

This PR implements the synced scenario. Other scenarios require git bundle support (future work).

Sample repo.json

{
  "workspaceType": "remote-git",
  "syncStatus": "synced",
  "remoteUrl": "https://github.com/example-org/my-project.git",
  "remoteVendor": "github",
  "remoteTrackingBranch": "origin/feature/chat-export",
  "remoteBaseBranch": "origin/main",
  "remoteHeadCommit": "abc123def456",
  "localBranch": "feature/chat-export",
  "localHeadCommit": "abc123def456",
  "diffs": [
    {
      "relativePath": "src/utils/helper.ts",
      "changeType": "modified",
      "status": "M",
      "unifiedDiff": "--- a/src/utils/helper.ts\n+++ b/src/utils/helper.ts\n@@ -10,6 +10,7 @@\n function helper() {\n+  console.log('debug');\n   return true;\n }"
    }
  ]
}

[Copilot]

Pull request overview

This PR introduces a new feature for internal users to export chat sessions as zip files, including repository information when available. The export captures the chat conversation data along with contextual information about the Git repository state at the time the chat session was started.

Key changes:

• Adds a new ChatRepoInfoContribution that captures repository context (remote URL, commit hash, changed files) when a chat request is first submitted
• Implements an "Export Chat as Zip" action that creates a zip file containing chat data and optional repository information
• Extends the chat model interfaces to store and serialize repository data alongside chat sessions
• Adds native service support for creating zip files from in-memory file content

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
src/vs/workbench/contrib/chat/common/model/chatModel.ts	Adds IExportableRepoData and IExportableRepoDiff interfaces; extends ChatModel to store and serialize repository data
src/vs/workbench/contrib/chat/browser/chatRepoInfo.ts	New contribution that captures git repository information on first chat request
src/vs/workbench/contrib/chat/browser/chat.contribution.ts	Registers the ChatRepoInfoContribution
src/vs/workbench/contrib/chat/electron-browser/actions/chatExportZip.ts	Implements the export chat as zip action with repository data
src/vs/workbench/contrib/chat/electron-browser/chat.contribution.ts	Registers the export zip action
src/vs/platform/native/common/native.ts	Adds createZipFile method to native service interface
src/vs/platform/native/electron-main/nativeHostMainService.ts	Implements createZipFile using existing zip utility
src/vs/workbench/test/electron-browser/workbenchTestServices.ts	Updates test service mock with createZipFile stub
src/vs/workbench/contrib/chat/test/common/model/mockChatModel.ts	Updates MockChatModel with repoData property and setRepoData method

Comments suppressed due to low confidence (2)

src/vs/workbench/contrib/chat/common/model/chatModel.ts:1339

• The repoData field was added to ISerializableChatData3 without a comment explaining what it is or noting backward compatibility concerns, while the inputState field above had such a comment (which was removed in this change). Consider adding a comment like "/** Repository context captured when the session started (optional, fully backwards compatible) */" to maintain consistency and help future maintainers understand the field's purpose and compatibility guarantees.

	repoData?: IExportableRepoData;

src/vs/workbench/contrib/chat/browser/chatRepoInfo.ts:123

• The ChatRepoInfoContribution lacks test coverage. Since this contribution performs important operations like reading git configuration, capturing repository state, and handling various edge cases (missing git history, worktrees, etc.), consider adding tests to verify the repo data capture logic and error handling behavior.

	static readonly ID = 'workbench.contrib.chatRepoInfo';

	constructor(
		@IChatService private readonly chatService: IChatService,
		@ISCMService private readonly scmService: ISCMService,
		@IFileService private readonly fileService: IFileService,
		@ILogService private readonly logService: ILogService,
	) {
		super();

		this._register(this.chatService.onDidSubmitRequest(async ({ chatSessionResource }) => {
			const model = this.chatService.getSession(chatSessionResource);
			if (!model || model.repoData) {
				return;
			}
			await this.captureAndSetRepoData(model);
		}));
	}

	private async captureAndSetRepoData(model: IChatModel): Promise<void> {
		try {
			const repoData = await this.captureRepoInfo();
			if (repoData) {
				model.setRepoData(repoData);
				if (!repoData.headCommitHash) {
					this.logService.warn('[ChatRepoInfo] Captured repo data without commit hash - git history may not be ready');
				}
			} else {
				this.logService.debug('[ChatRepoInfo] No SCM repository available for chat session');
			}
		} catch (error) {
			this.logService.warn('[ChatRepoInfo] Failed to capture repo info:', error);
		}
	}

	private async captureRepoInfo(): Promise<IExportableRepoData | undefined> {
		// Get the first SCM repository
		const repositories = [...this.scmService.repositories];
		if (repositories.length === 0) {
			return undefined;
		}

		const repository = repositories[0];
		const rootUri = repository.provider.rootUri;
		if (!rootUri) {
			return undefined;
		}

		let remoteUrl: string | undefined;
		try {
			const gitConfigUri = rootUri.with({ path: `${rootUri.path}/.git/config` });
			const exists = await this.fileService.exists(gitConfigUri);
			if (exists) {
				const content = await this.fileService.readFile(gitConfigUri);
				const remotes = getRemotes(content.value.toString());
				remoteUrl = remotes[0];
			}
		} catch (error) {
			this.logService.warn('[ChatRepoInfo] Failed to read git remote URL:', error);
		}

		let branchName: string | undefined;
		let headCommitHash: string | undefined;
		const historyProvider = repository.provider.historyProvider?.get();
		if (historyProvider) {
			const historyItemRef = historyProvider.historyItemRef.get();
			branchName = historyItemRef?.name;
			headCommitHash = historyItemRef?.revision;
		}

		let repoType: 'github' | 'ado' = 'github';
		if (remoteUrl?.includes('dev.azure.com') || remoteUrl?.includes('visualstudio.com')) {
			repoType = 'ado';
		}

		const diffs: IExportableRepoDiff[] = [];
		let changedFileCount = 0;

		for (const group of repository.provider.groups) {
			for (const resource of group.resources) {
				changedFileCount++;
				diffs.push({
					uri: resource.sourceUri.toString(),
					originalUri: resource.multiDiffEditorOriginalUri?.toString() ?? resource.sourceUri.toString(),
					renameUri: undefined,
					status: group.label || group.id,
					diff: undefined
				});
			}
		}

		return {
			remoteUrl,
			repoType,
			branchName,
			headCommitHash,
			changedFileCount,
			diffs
		};
	}
}

[vs-code-engineering]

📬 CODENOTIFY

The following users are being notified based on files changed in this PR:

@bpasero

Matched files:

• src/vs/platform/native/common/native.ts
• src/vs/platform/native/electron-main/nativeHostMainService.ts

[roblourens]

Let's pass around URIs, not string paths

[roblourens]

It seems like this always captures the repo data for every chat model, every user? But this seems like it coudl be really expensive? Should it only happen when the user requests it?

[roblourens]

It seems like we do this multiple times per ChatModel, and just keep replacing the previous repo data? I don't understand that part.

[roblourens]

You dispose disposables but it's still in _pendingSessions so it would leak the memory

[roblourens]

I think you can expose this._sessionModels.onDidCreateModel rather than create a new event emitter. eg onDidCreateModel = this._sessionModels.onDidCreateModel

[roblourens]

Maybe it has to be returned from a getter

[bpasero]

Left a minor comment, no objections having a zip method in native host service, thats easier than having it anywhere else.

[bpasero]

No need to lazy import this.