Add BrowserSessionRow grouping

src/core/Cline.ts

@@ -1356,6 +1356,7 @@ export class Cline {
 								// if the block is complete and we don't have a valid action this is a mistake
 								this.consecutiveMistakeCount++
 								pushToolResult(await this.sayAndCreateMissingParamError("browser_action", "action"))
+								await this.browserSession.closeBrowser()
 							}
 							break
 						}
@@ -1363,14 +1364,18 @@ export class Cline {
 						try {
 							if (block.partial) {
 								if (action === "launch") {
-									await this.ask("browser_action_launch", url, block.partial).catch(() => {})
+									await this.ask(
+										"browser_action_launch",
+										removeClosingTag("url", url),
+										block.partial
+									).catch(() => {})
 								} else {
 									await this.say(
 										"browser_action",
 										JSON.stringify({
 											action: action as BrowserAction,
-											coordinate,
-											text,
+											coordinate: removeClosingTag("coordinate", coordinate),
+											text: removeClosingTag("text", text),
 										} satisfies ClineSayBrowserAction),
 										undefined,
 										block.partial
@@ -1385,6 +1390,7 @@ export class Cline {
 										pushToolResult(
 											await this.sayAndCreateMissingParamError("browser_action", "url")
 										)
+										await this.browserSession.closeBrowser()
 										break
 									}
 									this.consecutiveMistakeCount = 0
@@ -1401,6 +1407,7 @@ export class Cline {
 											pushToolResult(
 												await this.sayAndCreateMissingParamError("browser_action", "coordinate")
 											)
+											await this.browserSession.closeBrowser()
 											break // can't be within an inner switch
 										}
 									}
@@ -1410,6 +1417,7 @@ export class Cline {
 											pushToolResult(
 												await this.sayAndCreateMissingParamError("browser_action", "text")
 											)
+											await this.browserSession.closeBrowser()
 											break
 										}
 									}
@@ -1446,13 +1454,13 @@ export class Cline {
 								// NOTE: it's okay that we call this message since the partial inspect_site is finished streaming. The only scenario we have to avoid is sending messages WHILE a partial message exists at the end of the messages array. For example the api_req_finished message would interfere with the partial message, so we needed to remove that.
 								// await this.say("inspect_site_result", "") // no result, starts the loading spinner waiting for result
 
-								await this.say("browser_action_result", JSON.stringify(browserActionResult))
 								switch (action) {
 									case "launch":
 									case "click":
 									case "type":
 									case "scroll_down":
 									case "scroll_up":
+										await this.say("browser_action_result", JSON.stringify(browserActionResult))
 										pushToolResult(
 											formatResponse.toolResult(
 												`The browser action has been executed. The console logs and screenshot have been captured for your analysis.\n\nConsole logs:\n${
@@ -1473,6 +1481,7 @@ export class Cline {
 								break
 							}
 						} catch (error) {
+							await this.browserSession.closeBrowser() // if any error occurs, the browser session is terminated
 							await handleError("executing browser action", error)
 							break
 						}

webview-ui/src/components/chat/BrowserSessionRow.tsx

@@ -0,0 +1,159 @@
+import deepEqual from "fast-deep-equal"
+import React, { memo, useEffect, useRef } from "react"
+import { useSize } from "react-use"
+import { BrowserActionResult, ClineMessage, ClineSayBrowserAction } from "../../../../src/shared/ExtensionMessage"
+import { vscode } from "../../utils/vscode"
+import CodeAccordian from "../common/CodeAccordian"
+import CodeBlock, { CODE_BLOCK_BG_COLOR } from "../common/CodeBlock"
+import { ChatRowContent } from "./ChatRow"
+
+interface BrowserSessionRowProps {
+	messages: ClineMessage[]
+	isExpanded: boolean
+	onToggleExpand: () => void
+	lastModifiedMessage?: ClineMessage
+	isLast: boolean
+	onHeightChange: (isTaller: boolean) => void
+}
+
+const BrowserSessionRow = memo((props: BrowserSessionRowProps) => {
+	const { messages, isLast, onHeightChange } = props
+	const prevHeightRef = useRef(0)
+
+	const [browserSessionRow, { height }] = useSize(
+		<div style={{ padding: "10px 6px 10px 15px" }}>
+			<h3>Browser Session Group</h3>
+			{messages.map((message, index) => (
+				<BrowserSessionRowContent key={message.ts} {...props} message={message} />
+			))}
+			<h3>END Browser Session Group</h3>
+		</div>
+	)
+
+	useEffect(() => {
+		const isInitialRender = prevHeightRef.current === 0
+		if (isLast && height !== 0 && height !== Infinity && height !== prevHeightRef.current) {
+			if (!isInitialRender) {
+				onHeightChange(height > prevHeightRef.current)
+			}
+			prevHeightRef.current = height
+		}
+	}, [height, isLast, onHeightChange])
+
+	return browserSessionRow
+}, deepEqual)
+
+interface BrowserSessionRowContentProps extends Omit<BrowserSessionRowProps, "messages"> {
+	message: ClineMessage
+}
+
+const BrowserSessionRowContent = ({
+	message,
+	isExpanded,
+	onToggleExpand,
+	lastModifiedMessage,
+	isLast,
+}: BrowserSessionRowContentProps) => {
+	const headerStyle: React.CSSProperties = {
+		display: "flex",
+		alignItems: "center",
+		gap: "10px",
+		marginBottom: "10px",
+	}
+
+	// Copy all the rendering logic from ChatRowContent
+	// This includes handling all message types: api_req_started, browser_action, text, etc.
+	// The implementation would be identical to ChatRowContent
+
+	switch (message.type) {
+		case "say":
+			switch (message.say) {
+				case "api_req_started":
+				case "text":
+					return (
+						<ChatRowContent
+							message={message}
+							isExpanded={isExpanded}
+							onToggleExpand={onToggleExpand}
+							lastModifiedMessage={lastModifiedMessage}
+							isLast={isLast}
+						/>
+					)
+
+				case "browser_action":
+					const browserAction = JSON.parse(message.text || "{}") as ClineSayBrowserAction
+					return (
+						<div style={{ marginBottom: 10 }}>
+							<div style={{ fontWeight: "bold" }}>{browserAction.action}</div>
+							{browserAction.coordinate && <div>{browserAction.coordinate}</div>}
+							{browserAction.text && <div>{browserAction.text}</div>}
+						</div>
+					)
+
+				case "browser_action_result":
+					const { screenshot, logs, currentMousePosition, currentUrl } = JSON.parse(
+						message.text || "{}"
+					) as BrowserActionResult
+					return (
+						<div style={{ marginBottom: 10 }}>
+							{currentMousePosition && <div>{currentMousePosition}</div>}
+							{currentUrl && <div>{currentUrl}</div>}
+							{screenshot && (
+								<img
+									src={screenshot}
+									alt="Browser action screenshot"
+									style={{
+										width: "calc(100% - 2px)",
+										height: "auto",
+										objectFit: "contain",
+										marginBottom: logs ? 7 : 0,
+										borderRadius: 3,
+										cursor: "pointer",
+										marginLeft: "1px",
+									}}
+									onClick={() => vscode.postMessage({ type: "openImage", text: screenshot })}
+								/>
+							)}
+							{logs && (
+								<CodeAccordian
+									code={logs}
+									language="shell"
+									isConsoleLogs={true}
+									isExpanded={isExpanded}
+									onToggleExpand={onToggleExpand}
+								/>
+							)}
+						</div>
+					)
+
+				default:
+					return null
+			}
+
+		case "ask":
+			switch (message.ask) {
+				case "browser_action_launch":
+					return (
+						<>
+							<div style={headerStyle}>
+								<span style={{ fontWeight: "bold" }}>Browser Session Started</span>
+							</div>
+							<div
+								style={{
+									borderRadius: 3,
+									border: "1px solid var(--vscode-editorGroup-border)",
+									overflow: "hidden",
+									backgroundColor: CODE_BLOCK_BG_COLOR,
+								}}>
+								<CodeBlock source={`${"```"}shell\n${message.text}\n${"```"}`} forceWrap={true} />
+							</div>
+						</>
+					)
+
+				default:
+					return null
+			}
+	}
+}
+
+export default BrowserSessionRow

webview-ui/src/components/chat/ChatRow.tsx

@@ -1,13 +1,8 @@
 import { VSCodeBadge, VSCodeProgressRing } from "@vscode/webview-ui-toolkit/react"
 import deepEqual from "fast-deep-equal"
 import React, { memo, useEffect, useMemo, useRef } from "react"
-import {
-	BrowserActionResult,
-	ClineApiReqInfo,
-	ClineMessage,
-	ClineSayBrowserAction,
-	ClineSayTool,
-} from "../../../../src/shared/ExtensionMessage"
+import { useSize } from "react-use"
+import { ClineApiReqInfo, ClineMessage, ClineSayTool } from "../../../../src/shared/ExtensionMessage"
 import { COMMAND_OUTPUT_STRING } from "../../../../src/shared/combineCommandSequences"
 import { vscode } from "../../utils/vscode"
 import CodeAccordian, { removeLeadingNonAlphanumeric } from "../common/CodeAccordian"
@@ -15,7 +10,6 @@ import CodeBlock, { CODE_BLOCK_BG_COLOR } from "../common/CodeBlock"
 import MarkdownBlock from "../common/MarkdownBlock"
 import Thumbnails from "../common/Thumbnails"
 import { highlightMentions } from "./TaskHeader"
-import { useSize } from "react-use"
 
 interface ChatRowProps {
 	message: ClineMessage
@@ -66,7 +60,13 @@ const ChatRow = memo(
 
 export default ChatRow
 
-const ChatRowContent = ({ message, isExpanded, onToggleExpand, lastModifiedMessage, isLast }: ChatRowContentProps) => {
+export const ChatRowContent = ({
+	message,
+	isExpanded,
+	onToggleExpand,
+	lastModifiedMessage,
+	isLast,
+}: ChatRowContentProps) => {
 	const [cost, apiReqCancelReason, apiReqStreamingFailedMessage] = useMemo(() => {
 		if (message.text != null && message.say === "api_req_started") {
 			const info: ClineApiReqInfo = JSON.parse(message.text)
@@ -617,58 +617,6 @@ const ChatRowContent = ({ message, isExpanded, onToggleExpand, lastModifiedMessa
 							</div>
 						</>
 					)
-				case "browser_action":
-					const browserAction = JSON.parse(message.text || "{}") as ClineSayBrowserAction
-					return (
-						<div
-							style={{
-								marginTop: -10,
-								width: "100%",
-							}}>
-							<div style={{ fontWeight: "bold" }}>{browserAction.action}</div>
-							{browserAction.coordinate && <div>{browserAction.coordinate}</div>}
-							{browserAction.text && <div>{browserAction.text}</div>}
-						</div>
-					)
-				case "browser_action_result":
-					const { screenshot, logs, currentMousePosition, currentUrl } = JSON.parse(
-						message.text || "{}"
-					) as BrowserActionResult
-					return (
-						<div
-							style={{
-								marginTop: -10,
-								width: "100%",
-							}}>
-							{currentMousePosition && <div>{currentMousePosition}</div>}
-							{currentUrl && <div>{currentUrl}</div>}
-							{screenshot && (
-								<img
-									src={screenshot}
-									alt="Inspect screenshot"
-									style={{
-										width: "calc(100% - 2px)",
-										height: "auto",
-										objectFit: "contain",
-										marginBottom: logs ? 7 : 0,
-										borderRadius: 3,
-										cursor: "pointer",
-										marginLeft: "1px",
-									}}
-									onClick={() => vscode.postMessage({ type: "openImage", text: screenshot })}
-								/>
-							)}
-							{logs && (
-								<CodeAccordian
-									code={logs}
-									language="shell"
-									isConsoleLogs={true}
-									isExpanded={isExpanded}
-									onToggleExpand={onToggleExpand}
-								/>
-							)}
-						</div>
-					)
 				default:
 					return (
 						<>
@@ -800,29 +748,6 @@ const ChatRowContent = ({ message, isExpanded, onToggleExpand, lastModifiedMessa
 							</div>
 						</>
 					)
-				case "browser_action_launch":
-					// const isInspecting =
-					// isLast && lastModifiedMessage?.say === "inspect_site_result" && !lastModifiedMessage?.images
-
-					return (
-						<>
-							<div style={headerStyle}>
-								{/* {isInspecting ? <ProgressIndicator /> : toolIcon("inspect")} */}
-								<span style={{ fontWeight: "bold" }}>
-									<>Cline wants to use the browser:</>
-								</span>
-							</div>
-							<div
-								style={{
-									borderRadius: 3,
-									border: "1px solid var(--vscode-editorGroup-border)",
-									overflow: "hidden",
-									backgroundColor: CODE_BLOCK_BG_COLOR,
-								}}>
-								<CodeBlock source={`${"```"}shell\n${message.text}\n${"```"}`} forceWrap={true} />
-							</div>
-						</>
-					)
 				default:
 					return null
 			}

webview-ui/src/components/chat/ChatView.tsx

@@ -4,7 +4,13 @@ import { useCallback, useEffect, useMemo, useRef, useState } from "react"
 import { useDeepCompareEffect, useEvent, useMount } from "react-use"
 import { Virtuoso, type VirtuosoHandle } from "react-virtuoso"
 import styled from "styled-components"
-import { ClineAsk, ClineSayTool, ExtensionMessage } from "../../../../src/shared/ExtensionMessage"
+import {
+	ClineAsk,
+	ClineMessage,
+	ClineSayBrowserAction,
+	ClineSayTool,
+	ExtensionMessage,
+} from "../../../../src/shared/ExtensionMessage"
 import { findLast } from "../../../../src/shared/array"
 import { combineApiRequests } from "../../../../src/shared/combineApiRequests"
 import { combineCommandSequences } from "../../../../src/shared/combineCommandSequences"
@@ -14,6 +20,7 @@ import { vscode } from "../../utils/vscode"
 import HistoryPreview from "../history/HistoryPreview"
 import { normalizeApiConfiguration } from "../settings/ApiOptions"
 import Announcement from "./Announcement"
+import BrowserSessionRow from "./BrowserSessionRow"
 import ChatRow from "./ChatRow"
 import ChatTextArea from "./ChatTextArea"
 import TaskHeader from "./TaskHeader"
@@ -437,6 +444,66 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 		})
 	}, [modifiedMessages])
 
+	const isBrowserSessionMessage = (message: ClineMessage): boolean => {
+		// which of visible messages are browser session messages, see above
+		if (message.type === "ask") {
+			return ["browser_action_launch"].includes(message.ask!)
+		}
+		if (message.type === "say") {
+			return ["api_req_started", "text", "browser_action", "browser_action_result"].includes(message.say!)
+		}
+		return false
+	}
+
+	const groupedMessages = useMemo(() => {
+		const result: (ClineMessage | ClineMessage[])[] = []
+		let currentGroup: ClineMessage[] = []
+		let isInBrowserSession = false
+
+		const endBrowserSession = () => {
+			if (currentGroup.length > 0) {
+				result.push([...currentGroup])
+				currentGroup = []
+				isInBrowserSession = false
+			}
+		}
+
+		visibleMessages.forEach((message) => {
+			if (message.ask === "browser_action_launch") {
+				// complete existing browser session if any
+				endBrowserSession()
+				// start new
+				isInBrowserSession = true
+				currentGroup.push(message)
+			} else if (isInBrowserSession) {
+				if (isBrowserSessionMessage(message)) {
+					currentGroup.push(message)
+
+					// Check if this is a close action
+					if (message.say === "browser_action") {
+						const browserAction = JSON.parse(message.text || "{}") as ClineSayBrowserAction
+						if (browserAction.action === "close") {
+							endBrowserSession()
+						}
+					}
+				} else {
+					// complete existing browser session if any
+					endBrowserSession()
+					result.push(message)
+				}
+			} else {
+				result.push(message)
+			}
+		})
+
+		// Handle case where browser session is the last group
+		if (currentGroup.length > 0) {
+			result.push([...currentGroup])
+		}
+
+		return result
+	}, [visibleMessages])
+
 	// scrolling
 
 	const scrollToBottomSmooth = useMemo(
@@ -465,10 +532,19 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 	const toggleRowExpansion = useCallback(
 		(ts: number) => {
 			const isCollapsing = expandedRows[ts] ?? false
-			const isLast = visibleMessages.at(-1)?.ts === ts
-			const isSecondToLast = visibleMessages.at(-2)?.ts === ts
+			const lastGroup = groupedMessages.at(-1)
+			const isLast = Array.isArray(lastGroup) ? lastGroup[0].ts === ts : lastGroup?.ts === ts
+			const secondToLastGroup = groupedMessages.at(-2)
+			const isSecondToLast = Array.isArray(secondToLastGroup)
+				? secondToLastGroup[0].ts === ts
+				: secondToLastGroup?.ts === ts
+
 			const isLastCollapsedApiReq =
-				visibleMessages.at(-1)?.say === "api_req_started" && !expandedRows[visibleMessages.at(-1)?.ts ?? 0]
+				isLast &&
+				!Array.isArray(lastGroup) && // Make sure it's not a browser session group
+				lastGroup?.say === "api_req_started" &&
+				!expandedRows[lastGroup.ts]
+
 			setExpandedRows((prev) => ({
 				...prev,
 				[ts]: !prev[ts],
@@ -496,7 +572,7 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 				} else {
 					const timer = setTimeout(() => {
 						virtuosoRef.current?.scrollToIndex({
-							index: visibleMessages.length - (isLast ? 1 : 2),
+							index: groupedMessages.length - (isLast ? 1 : 2),
 							align: "start",
 						})
 					}, 0)
@@ -504,7 +580,7 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 				}
 			}
 		},
-		[visibleMessages, expandedRows, scrollToBottomAuto, isAtBottom]
+		[groupedMessages, expandedRows, scrollToBottomAuto, isAtBottom]
 	)
 
 	const handleRowHeightChange = useCallback(
@@ -548,18 +624,37 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 	}, [task])
 
 	const itemContent = useCallback(
-		(index: number, message: any) => (
-			<ChatRow
-				key={message.ts}
-				message={message}
-				isExpanded={expandedRows[message.ts] || false}
-				onToggleExpand={() => toggleRowExpansion(message.ts)}
+		(index: number, messageOrGroup: ClineMessage | ClineMessage[]) => {
+			// browser session group
+			if (Array.isArray(messageOrGroup)) {
+				const firstMessage = messageOrGroup[0]
+				return (
+					<BrowserSessionRow
+						key={firstMessage.ts}
+						messages={messageOrGroup}
+						isExpanded={expandedRows[firstMessage.ts] || false}
+						onToggleExpand={() => toggleRowExpansion(firstMessage.ts)}
 						lastModifiedMessage={modifiedMessages.at(-1)}
-				isLast={index === visibleMessages.length - 1}
+						isLast={index === groupedMessages.length - 1}
 						onHeightChange={handleRowHeightChange}
 					/>
-		),
-		[expandedRows, modifiedMessages, visibleMessages.length, toggleRowExpansion, handleRowHeightChange]
+				)
+			}
+
+			// regular message
+			return (
+				<ChatRow
+					key={messageOrGroup.ts}
+					message={messageOrGroup}
+					isExpanded={expandedRows[messageOrGroup.ts] || false}
+					onToggleExpand={() => toggleRowExpansion(messageOrGroup.ts)}
+					lastModifiedMessage={modifiedMessages.at(-1)}
+					isLast={index === groupedMessages.length - 1}
+					onHeightChange={handleRowHeightChange}
+				/>
+			)
+		},
+		[expandedRows, modifiedMessages, groupedMessages.length, toggleRowExpansion, handleRowHeightChange]
 	)
 
 	return (
@@ -627,7 +722,7 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 							}}
 							// increasing top by 3_000 to prevent jumping around when user collapses a row
 							increaseViewportBy={{ top: 3_000, bottom: Number.MAX_SAFE_INTEGER }} // hack to make sure the last message is always rendered to get truly perfect scroll to bottom animation when new messages are added (Number.MAX_SAFE_INTEGER is safe for arithmetic operations, which is all virtuoso uses this value for in src/sizeRangeSystem.ts)
-							data={visibleMessages} // messages is the raw format returned by extension, modifiedMessages is the manipulated structure that combines certain messages of related type, and visibleMessages is the filtered structure that removes messages that should not be rendered
+							data={groupedMessages} // messages is the raw format returned by extension, modifiedMessages is the manipulated structure that combines certain messages of related type, and visibleMessages is the filtered structure that removes messages that should not be rendered
 							itemContent={itemContent}
 							atBottomStateChange={(isAtBottom) => {
 								setIsAtBottom(isAtBottom)
@@ -637,7 +732,7 @@ const ChatView = ({ isHidden, showAnnouncement, hideAnnouncement, showHistoryVie
 								setShowScrollToBottom(disableAutoScrollRef.current && !isAtBottom)
 							}}
 							atBottomThreshold={10} // anything lower causes issues with followOutput
-							initialTopMostItemIndex={visibleMessages.length - 1}
+							initialTopMostItemIndex={groupedMessages.length - 1}
 						/>
 					</div>
 					{showScrollToBottom ? (