Merge branch 'Zoo-Code-Org:main' into fix/shell-defaultProfileName-type-guard

Author: daewoongoh

.changeset/improve-apply-diff-prompt.md

@@ -0,0 +1,5 @@
+---
+"zoo-code": patch
+---
+
+Enhance the `apply_diff` tool description and parameter instructions to recommend `:start_line:` with exact syntax and emphasize copy-paste exact matching requirements, improving success rates for Gemini Flash and other smaller/faster models.

.github/workflows/e2e.yml

@@ -34,6 +34,22 @@ jobs:
             - name: Install xvfb
               if: github.event_name != 'pull_request' || steps.e2e-marker.outputs.cache-hit != 'true'
               run: sudo apt-get install -y xvfb
+
+            - name: Get VS Code version from package.json
+              if: github.event_name != 'pull_request' || steps.e2e-marker.outputs.cache-hit != 'true'
+              id: vscode-ver
+              run: |
+                  VERSION=$(node -p 'require("./apps/vscode-e2e/package.json").devDependencies["@types/vscode"]')
+                  echo "version=$VERSION" >> $GITHUB_OUTPUT
+
+            - name: Cache VS Code test binary
+              if: github.event_name != 'pull_request' || steps.e2e-marker.outputs.cache-hit != 'true'
+              uses: actions/cache@0057852bfaa89a56745cba8c7296529d2fc39830 # v4
+              with:
+                  path: |
+                      apps/vscode-e2e/.vscode-test/
+                  key: vscode-test-${{ runner.os }}-${{ steps.vscode-ver.outputs.version }}-v1
+
             - name: Run mocked E2E tests
               id: run-e2e
               # merge_group and workflow_dispatch always run; cache skip is pull_request only

README.md

@@ -87,7 +87,7 @@ for this exact support, so if you are having problems or if you have question, j
 - [简体中文](locales/zh-CN/README.md)
 - [繁體中文](locales/zh-TW/README.md)
 - ...
-    </details>
+      </details>
 
 ---
 

apps/vscode-e2e/src/runTest.ts

@@ -1,6 +1,7 @@
 import * as path from "path"
 import * as os from "os"
 import * as fs from "fs/promises"
+import { readFileSync } from "fs"
 
 import { runTests } from "@vscode/test-electron"
 import { LLMock } from "@copilotkit/aimock"
@@ -156,12 +157,16 @@ async function main() {
 		}
 
 		// Download VS Code, unzip it and run the integration test
+		// Read VS Code version from package.json to keep in sync with @types/vscode
+		const pkg = JSON.parse(readFileSync(path.resolve(__dirname, "../package.json"), "utf-8"))
+		const vscodeVersion = process.env.VSCODE_VERSION || pkg.devDependencies["@types/vscode"]
+
 		await runTests({
 			extensionDevelopmentPath,
 			extensionTestsPath,
 			launchArgs: [testWorkspace],
 			extensionTestsEnv,
-			version: process.env.VSCODE_VERSION || "1.100.0",
+			version: vscodeVersion,
 		})
 	} catch (error) {
 		console.error("Failed to run tests", error)

apps/vscode-e2e/src/suite/providers/zai.test.ts

@@ -220,6 +220,10 @@ suite("Z.ai GLM provider", function () {
 
 		await waitUntilCompleted({ api, taskId })
 		const capturedMaxTokens = requestCapture.maxTokens
+		assert.ok(
+			capturedMaxTokens !== undefined,
+			"max_tokens should have been captured by the fetch interceptor before task completion",
+		)
 
 		const completionMessage = messages.find(
 			({ say, text }) => (say === "completion_result" || say === "text") && text?.trim() === "4",
@@ -229,8 +233,6 @@ suite("Z.ai GLM provider", function () {
 
 		// Verify max_tokens uses the restored default clamp (20% of context window)
 		// unless the user explicitly overrides it via modelMaxTokens.
-		// Snapshot immediately after waitUntilCompleted to avoid straggling async calls
-		// from this task overwriting requestCapture before the assertion runs.
 		assert.strictEqual(
 			capturedMaxTokens,
 			40_000,
@@ -264,6 +266,10 @@ suite("Z.ai GLM provider", function () {
 
 		await waitUntilCompleted({ api, taskId })
 		const capturedMaxTokens = requestCapture.maxTokens
+		assert.ok(
+			capturedMaxTokens !== undefined,
+			"max_tokens should have been captured by the fetch interceptor before task completion",
+		)
 
 		const completionMessage = messages.find(
 			({ say, text }) => (say === "completion_result" || say === "text") && text?.trim() === "4",
@@ -273,11 +279,10 @@ suite("Z.ai GLM provider", function () {
 
 		// Verify max_tokens uses the restored default clamp (20% of context window)
 		// unless the user explicitly overrides it via modelMaxTokens.
-		// Snapshot immediately after waitUntilCompleted to avoid straggling async calls
-		// from the prior test overwriting requestCapture before this assertion runs.
+		const expectedMaxTokens = 40_551 // Math.ceil(202_752 * 0.2) for glm-5-turbo
 		assert.strictEqual(
 			capturedMaxTokens,
-			40_551,
+			expectedMaxTokens,
 			`max_tokens should default to the glm-5-turbo clamp (40_551) but was ${capturedMaxTokens}`,
 		)
 	})

packages/types/src/__tests__/message.test.ts

@@ -1,6 +1,14 @@
 // pnpm --filter @roo-code/types test src/__tests__/message.test.ts
 
-import { clineAsks, isIdleAsk, isInteractiveAsk, isResumableAsk, isNonBlockingAsk } from "../message.js"
+import {
+	clineAsks,
+	getCompletionCheckpoint,
+	isIdleAsk,
+	isInteractiveAsk,
+	isResumableAsk,
+	isNonBlockingAsk,
+	type ClineMessage,
+} from "../message.js"
 
 describe("ask messages", () => {
 	test("all ask messages are classified", () => {
@@ -12,3 +20,46 @@ describe("ask messages", () => {
 		}
 	})
 })
+
+describe("getCompletionCheckpoint", () => {
+	it("returns the first checkpoint after the latest user prompt before completion", () => {
+		const messages: ClineMessage[] = [
+			{ type: "say", say: "text", ts: 1, text: "Initial task" },
+			{ type: "say", say: "checkpoint_saved", ts: 2, text: "initial-checkpoint" },
+			{ type: "say", say: "completion_result", ts: 3, text: "First completion" },
+			{ type: "say", say: "user_feedback", ts: 4, text: "Change it" },
+			{ type: "say", say: "checkpoint_saved", ts: 5, text: "latest-prompt-checkpoint" },
+			{ type: "say", say: "checkpoint_saved", ts: 6, text: "later-edit-checkpoint" },
+			{ type: "ask", ask: "completion_result", ts: 7, text: "", partial: false },
+		]
+
+		expect(getCompletionCheckpoint(messages)).toEqual({
+			ts: 5,
+			commitHash: "latest-prompt-checkpoint",
+		})
+	})
+
+	it("returns the first checkpoint after an initial task row before completion", () => {
+		const messages: ClineMessage[] = [
+			{ type: "say", say: "task", ts: 1, text: "Initial task" },
+			{ type: "say", say: "checkpoint_saved", ts: 2, text: "checkpoint-after-initial-task" },
+			{ type: "ask", ask: "completion_result", ts: 3, text: "Task complete", partial: false },
+		]
+
+		expect(getCompletionCheckpoint(messages)).toEqual({
+			ts: 2,
+			commitHash: "checkpoint-after-initial-task",
+		})
+	})
+
+	it("returns undefined when completion has no checkpoint after the latest user prompt", () => {
+		const messages: ClineMessage[] = [
+			{ type: "say", say: "text", ts: 1, text: "Initial task" },
+			{ type: "say", say: "checkpoint_saved", ts: 2, text: "initial-checkpoint" },
+			{ type: "say", say: "user_feedback", ts: 3, text: "Change it" },
+			{ type: "ask", ask: "completion_result", ts: 4, text: "", partial: false },
+		]
+
+		expect(getCompletionCheckpoint(messages)).toBeUndefined()
+	})
+})

packages/types/src/global-settings.ts

@@ -1,19 +1,18 @@
 import { z } from "zod"
 
-import { type Keys } from "./type-fu.js"
+import { codebaseIndexConfigSchema, codebaseIndexModelsSchema } from "./codebase-index.js"
+import { experimentsSchema } from "./experiment.js"
+import { historyItemSchema } from "./history.js"
+import { customModePromptsSchema, customSupportPromptsSchema, modeConfigSchema } from "./mode.js"
 import {
 	type ProviderSettings,
 	PROVIDER_SETTINGS_KEYS,
 	providerSettingsEntrySchema,
 	providerSettingsSchema,
 } from "./provider-settings.js"
-import { historyItemSchema } from "./history.js"
-import { codebaseIndexModelsSchema, codebaseIndexConfigSchema } from "./codebase-index.js"
-import { experimentsSchema } from "./experiment.js"
 import { telemetrySettingsSchema } from "./telemetry.js"
-import { modeConfigSchema } from "./mode.js"
-import { customModePromptsSchema, customSupportPromptsSchema } from "./mode.js"
 import { toolNamesSchema } from "./tool.js"
+import { type Keys } from "./type-fu.js"
 import { languagesSchema } from "./vscode.js"
 
 /**
@@ -23,6 +22,16 @@ import { languagesSchema } from "./vscode.js"
  */
 export const DEFAULT_WRITE_DELAY_MS = 1000
 
+/**
+ * Default fuzzy matching threshold for the multi-search-replace diff strategy.
+ * A value of 1.0 (exact match) is used by default for safety, especially when
+ * auto-approval for writes is enabled. This prevents unintended changes from
+ * being applied due to minor mismatches. Users can lower this threshold manually
+ * in settings to reduce "Edit Unsuccessful" errors caused by minor whitespace
+ * or formatting differences, accepting a higher risk of unintended edits.
+ */
+export const DEFAULT_DIFF_FUZZY_THRESHOLD = 1.0
+
 /**
  * Terminal output preview size options for persisted command output.
  *
@@ -102,6 +111,12 @@ export const globalSettingsSchema = z.object({
 	alwaysAllowWriteOutsideWorkspace: z.boolean().optional(),
 	alwaysAllowWriteProtected: z.boolean().optional(),
 	writeDelayMs: z.number().min(0).optional(),
+	/**
+	 * Fuzzy matching threshold for the multi-search-replace diff strategy.
+	 * Range: 0.5 (50% minimum similarity) to 1.0 (exact match only).
+	 * `@default` 1.0
+	 */
+	diffFuzzyThreshold: z.number().min(0.5).max(1).optional(),
 	requestDelaySeconds: z.number().optional(),
 	alwaysAllowMcp: z.boolean().optional(),
 	alwaysAllowModeSwitch: z.boolean().optional(),

packages/types/src/message.ts

@@ -150,6 +150,7 @@ export const clineSays = [
 	"api_req_rate_limit_wait",
 	"api_req_deleted",
 	"text",
+	"task",
 	"image",
 	"reasoning",
 	"completion_result",
@@ -275,6 +276,76 @@ export const clineMessageSchema = z.object({
 
 export type ClineMessage = z.infer<typeof clineMessageSchema>
 
+export interface CompletionCheckpoint {
+	ts: number
+	commitHash: string
+}
+
+const isInitialTaskMessage = (message: ClineMessage | undefined): boolean => {
+	return message?.type === "say" && (message.say === "text" || message.say === "task")
+}
+
+const isUserFeedbackMessage = (message: ClineMessage): boolean => {
+	return message.type === "say" && message.say === "user_feedback"
+}
+
+const isCompletionMessage = (message: ClineMessage): boolean => {
+	return (
+		(message.type === "ask" && message.ask === "completion_result") ||
+		(message.type === "say" && message.say === "completion_result")
+	)
+}
+
+const isCheckpointMessage = (message: ClineMessage): boolean => {
+	return message.type === "say" && message.say === "checkpoint_saved" && typeof message.text === "string"
+}
+
+function findLastIndexBefore(
+	messages: ClineMessage[],
+	beforeIndex: number,
+	predicate: (message: ClineMessage) => boolean,
+): number {
+	for (let i = beforeIndex - 1; i >= 0; i--) {
+		const message = messages[i]
+
+		if (message && predicate(message)) {
+			return i
+		}
+	}
+
+	return -1
+}
+
+/**
+ * Finds the checkpoint that should anchor completion-result actions.
+ *
+ * The baseline is the first checkpoint created after the latest user prompt in
+ * the turn that produced the completion. Restoring to that checkpoint reverts
+ * changes made for the latest prompt, and diffing from it shows the same scoped
+ * changes.
+ */
+export function getCompletionCheckpoint(messages: ClineMessage[]): CompletionCheckpoint | undefined {
+	const completionIndex = findLastIndexBefore(messages, messages.length, isCompletionMessage)
+	const searchEnd = completionIndex === -1 ? messages.length : completionIndex
+	const latestUserFeedbackIndex = findLastIndexBefore(messages, searchEnd, isUserFeedbackMessage)
+	const latestUserPromptIndex =
+		latestUserFeedbackIndex !== -1 ? latestUserFeedbackIndex : isInitialTaskMessage(messages[0]) ? 0 : -1
+
+	if (latestUserPromptIndex === -1) {
+		return undefined
+	}
+
+	for (let i = latestUserPromptIndex + 1; i < searchEnd; i++) {
+		const message = messages[i]
+
+		if (message && isCheckpointMessage(message)) {
+			return { ts: message.ts, commitHash: message.text! }
+		}
+	}
+
+	return undefined
+}
+
 /**
  * TokenUsage
  */

packages/types/src/vscode-extension-host.ts

@@ -332,6 +332,7 @@ export type ExtensionState = Pick<
 	taskHistory: HistoryItem[]
 
 	writeDelayMs: number
+	diffFuzzyThreshold: number
 
 	enableCheckpoints: boolean
 	checkpointTimeout: number // Timeout for checkpoint initialization in seconds (default: 15)
@@ -517,6 +518,8 @@ export interface WebviewMessage {
 		| "openCustomModesSettings"
 		| "checkpointDiff"
 		| "checkpointRestore"
+		| "completionCheckpointDiff"
+		| "completionCheckpointRestore"
 		| "deleteMcpServer"
 		| "codebaseIndexEnabled"
 		| "telemetrySetting"

src/__tests__/single-open-invariant.spec.ts

@@ -73,6 +73,46 @@ describe("Single-open-task invariant", () => {
 		expect(addClineToStack).toHaveBeenCalledTimes(1)
 	})
 
+	it("Subtask create: keeps existing task open when parentTask is provided", async () => {
+		vi.spyOn(ProfileValidatorMod.ProfileValidator, "isProfileAllowed").mockReturnValue(true)
+
+		const removeClineFromStack = vi.fn().mockResolvedValue(undefined)
+		const addClineToStack = vi.fn().mockResolvedValue(undefined)
+		const parentTask = { taskId: "parent-1" }
+
+		const provider = {
+			clineStack: [parentTask],
+			setValues: vi.fn(),
+			getState: vi.fn().mockResolvedValue({
+				apiConfiguration: { apiProvider: "anthropic", consecutiveMistakeLimit: 0 },
+				organizationAllowList: "*",
+				enableCheckpoints: true,
+				checkpointTimeout: 60,
+				cloudUserInfo: null,
+			}),
+			removeClineFromStack,
+			addClineToStack,
+			setProviderProfile: vi.fn(),
+			log: vi.fn(),
+			getStateToPostToWebview: vi.fn(),
+			providerSettingsManager: { getModeConfigId: vi.fn(), listConfig: vi.fn() },
+			customModesManager: { getCustomModes: vi.fn().mockResolvedValue([]) },
+			taskCreationCallback: vi.fn(),
+			contextProxy: {
+				extensionUri: {},
+				setValue: vi.fn(),
+				getValue: vi.fn(),
+				setProviderSettings: vi.fn(),
+				getProviderSettings: vi.fn(() => ({})),
+			},
+		} as unknown as ClineProvider
+
+		await (ClineProvider.prototype as any).createTask.call(provider, "Subtask", undefined, parentTask as any)
+
+		expect(removeClineFromStack).not.toHaveBeenCalled()
+		expect(addClineToStack).toHaveBeenCalledTimes(1)
+	})
+
 	it("History resume path always closes current before rehydration (non-rehydrating case)", async () => {
 		const removeClineFromStack = vi.fn().mockResolvedValue(undefined)
 		const addClineToStack = vi.fn().mockResolvedValue(undefined)