diff --git a/src/main/browser/cdp-auxiliary-commands.ts b/src/main/browser/cdp-auxiliary-commands.ts new file mode 100644 index 00000000000..88a02d3f43c --- /dev/null +++ b/src/main/browser/cdp-auxiliary-commands.ts @@ -0,0 +1,213 @@ +import type { WebContents } from 'electron' +import type { + BrowserCaptureStartResult, + BrowserCaptureStopResult, + BrowserConsoleEntry, + BrowserConsoleResult, + BrowserCookie, + BrowserCookieDeleteResult, + BrowserCookieGetResult, + BrowserCookieSetResult, + BrowserGeolocationResult, + BrowserInterceptDisableResult, + BrowserInterceptEnableResult, + BrowserInterceptedRequest, + BrowserNetworkEntry, + BrowserNetworkLogResult, + BrowserViewportResult +} from '../../shared/runtime-types' +import type { CdpCommandSender, SnapshotResult } from './snapshot-engine' + +export type CdpTabState = { + navigationId: string | null + snapshotResult: SnapshotResult | null + debuggerAttached: boolean + debuggerDetachListener: (() => void) | null + debuggerMessageListener: ((_event: unknown, method: string, params: unknown) => void) | null + iframeSessions: Map + capturing: boolean + consoleLog: BrowserConsoleEntry[] + networkLog: BrowserNetworkEntry[] + intercepting: boolean + interceptPatterns: string[] + pausedRequests: Map + networkRequestMap: Map +} + +type ActiveCdpContext = { + guest: WebContents + sender: CdpCommandSender + state: CdpTabState +} + +export type CdpAuxiliaryCommandHost = { + run: (operation: (context: ActiveCdpContext) => Promise) => Promise + // Why: captureStop must not attach the debugger — stopping a capture on a tab whose + // debugger detached (DevTools opened) would otherwise throw before clearing state. + runOnState: ( + operation: (context: Pick) => Promise + ) => Promise + current: () => Pick +} + +export class CdpAuxiliaryCommands { + constructor(private readonly host: CdpAuxiliaryCommandHost) {} + + cookieGet(url?: string): Promise { + return this.host.run(async ({ sender }) => { + const params: Record = {} + if (url) { + params.urls = [url] + } + const { cookies } = (await sender('Network.getCookies', params)) as { + cookies: BrowserCookie[] + } + return { cookies } + }) + } + + cookieSet(cookie: { + name: string + value: string + domain?: string + path?: string + secure?: boolean + httpOnly?: boolean + sameSite?: string + expires?: number + }): Promise { + return this.host.run(async ({ sender }) => { + let domain = cookie.domain + if (!domain) { + const { result } = (await sender('Runtime.evaluate', { + expression: 'location.hostname', + returnByValue: true + })) as { result: { value: string } } + domain = result.value + } + const params: Record = { + name: cookie.name, + value: cookie.value, + domain, + path: cookie.path ?? '/', + secure: cookie.secure ?? false, + httpOnly: cookie.httpOnly ?? false, + sameSite: cookie.sameSite ?? 'Lax' + } + if (cookie.expires !== undefined) { + params.expires = cookie.expires + } + const { success } = (await sender('Network.setCookie', params)) as { success: boolean } + return { success } + }) + } + + cookieDelete(name: string, domain?: string, url?: string): Promise { + return this.host.run(async ({ sender }) => { + const params: Record = { name } + if (domain) { + params.domain = domain + } + if (url) { + params.url = url + } + if (!domain && !url) { + const { result } = (await sender('Runtime.evaluate', { + expression: 'location.href', + returnByValue: true + })) as { result: { value: string } } + params.url = result.value + } + await sender('Network.deleteCookies', params) + return { deleted: true } + }) + } + + setViewport( + width: number, + height: number, + deviceScaleFactor = 1, + mobile = false + ): Promise { + return this.host.run(async ({ sender }) => { + await sender('Emulation.setDeviceMetricsOverride', { + width, + height, + deviceScaleFactor, + mobile + }) + await Promise.resolve(sender('Emulation.setVisibleSize', { width, height })).catch(() => {}) + return { width, height, deviceScaleFactor, mobile } + }) + } + + setGeolocation( + latitude: number, + longitude: number, + accuracy = 1 + ): Promise { + return this.host.run(async ({ sender }) => { + await sender('Emulation.setGeolocationOverride', { latitude, longitude, accuracy }) + return { latitude, longitude, accuracy } + }) + } + + interceptEnable(patterns: string[] = ['*']): Promise { + return this.host.run(async ({ sender, state }) => { + await sender('Fetch.enable', { patterns: patterns.map((urlPattern) => ({ urlPattern })) }) + state.intercepting = true + state.interceptPatterns = patterns + return { enabled: true, patterns } + }) + } + + interceptDisable(): Promise { + return this.host.run(async ({ sender, state }) => { + await sender('Fetch.disable') + state.intercepting = false + state.interceptPatterns = [] + state.pausedRequests.clear() + return { disabled: true } + }) + } + + interceptList(): { requests: BrowserInterceptedRequest[] } { + const { state } = this.host.current() + return { requests: [...state.pausedRequests.values()] } + } + + captureStart(): Promise { + return this.host.run(async ({ sender, state }) => { + await sender('Runtime.enable') + state.capturing = true + state.consoleLog = [] + state.networkLog = [] + state.networkRequestMap.clear() + return { capturing: true } + }) + } + + captureStop(): Promise { + return this.host.runOnState(async ({ state }) => { + state.capturing = false + state.networkRequestMap.clear() + return { stopped: true } + }) + } + + consoleLog(limit = 100): BrowserConsoleResult { + const { state } = this.host.current() + return { + entries: state.consoleLog.slice(-limit), + truncated: state.consoleLog.length > limit + } + } + + networkLog(limit = 100): BrowserNetworkLogResult { + const { state } = this.host.current() + return { + entries: state.networkLog.slice(-limit), + truncated: state.networkLog.length > limit + } + } +} diff --git a/src/main/browser/cdp-bridge.ts b/src/main/browser/cdp-bridge.ts index fc5991924a9..73dec07f2d2 100644 --- a/src/main/browser/cdp-bridge.ts +++ b/src/main/browser/cdp-bridge.ts @@ -6,9 +6,7 @@ import type { BrowserCheckResult, BrowserClearResult, BrowserClickResult, - BrowserConsoleEntry, BrowserConsoleResult, - BrowserCookie, BrowserCookieDeleteResult, BrowserCookieGetResult, BrowserCookieSetResult, @@ -39,16 +37,12 @@ import type { BrowserViewportResult, BrowserWaitResult } from '../../shared/runtime-types' -import { - buildSnapshot, - type CdpCommandSender, - type RefEntry, - type SnapshotResult -} from './snapshot-engine' +import { buildSnapshot, type CdpCommandSender, type RefEntry } from './snapshot-engine' import { insertTextThroughCdp } from './browser-text-insertion' import type { BrowserManager } from './browser-manager' import { ANTI_DETECTION_SCRIPT } from './anti-detection' import { BrowserError } from './browser-error' +import { CdpAuxiliaryCommands, type CdpTabState } from './cdp-auxiliary-commands' const CAPTURE_LOG_LIMIT = 1000 @@ -56,25 +50,6 @@ const CAPTURE_LOG_LIMIT = 1000 // pulling Chromium in. Existing importers of this path keep working. export { BrowserError } from './browser-error' -type TabState = { - navigationId: string | null - snapshotResult: SnapshotResult | null - debuggerAttached: boolean - debuggerDetachListener: (() => void) | null - debuggerMessageListener: ((_event: unknown, method: string, params: unknown) => void) | null - iframeSessions: Map - // Why: capture state is per-tab so one tab's console/network events don't pollute another's buffer. - capturing: boolean - consoleLog: BrowserConsoleEntry[] - networkLog: BrowserNetworkEntry[] - // Why: interception state lets the agent selectively continue or block individual requests. - intercepting: boolean - interceptPatterns: string[] - pausedRequests: Map - // Why: maps CDP requestId → networkLog entry so loadingFinished attributes size to the right overlapping response. - networkRequestMap: Map -} - type QueuedCommand = { execute: () => Promise resolve: (value: unknown) => void @@ -83,13 +58,45 @@ type QueuedCommand = { export class CdpBridge { private activeWebContentsId: number | null = null - private readonly tabState = new Map() + private readonly tabState = new Map() private readonly commandQueues = new Map() private readonly processingQueues = new Set() private readonly browserManager: BrowserManager + private readonly auxiliaryCommands: CdpAuxiliaryCommands constructor(browserManager: BrowserManager) { this.browserManager = browserManager + this.auxiliaryCommands = new CdpAuxiliaryCommands({ + run: ( + operation: (context: { + guest: Electron.WebContents + sender: CdpCommandSender + state: CdpTabState + }) => Promise + ): Promise => + this.enqueueCommand(async () => { + const guest = this.getActiveGuest() + const sender = this.makeCdpSender(guest) + await this.ensureDebuggerAttached(guest) + const state = this.getOrCreateTabState(this.resolveTabId(guest.id)) + return operation({ guest, sender, state }) + }), + runOnState: ( + operation: (context: { guest: Electron.WebContents; state: CdpTabState }) => Promise + ): Promise => + this.enqueueCommand(async () => { + const guest = this.getActiveGuest() + const state = this.getOrCreateTabState(this.resolveTabId(guest.id)) + return operation({ guest, state }) + }), + current: () => { + const guest = this.getActiveGuest() + return { + guest, + state: this.getOrCreateTabState(this.resolveTabId(guest.id)) + } + } + }) } setActiveTab(webContentsId: number): void { @@ -636,25 +643,11 @@ export class CdpBridge { // ── Cookie management ── - async cookieGet(url?: string): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - const params: Record = {} - if (url) { - params.urls = [url] - } - const { cookies } = (await sender('Network.getCookies', params)) as { - cookies: BrowserCookie[] - } - - return { cookies } - }) + cookieGet(url?: string): Promise { + return this.auxiliaryCommands.cookieGet(url) } - async cookieSet(cookie: { + cookieSet(cookie: { name: string value: string domain?: string @@ -664,211 +657,66 @@ export class CdpBridge { sameSite?: string expires?: number }): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - // Why: Network.setCookie needs a domain or url to scope the cookie; infer the domain from the current page when omitted. - let domain = cookie.domain - if (!domain) { - const { result: urlResult } = (await sender('Runtime.evaluate', { - expression: 'location.hostname', - returnByValue: true - })) as { result: { value: string } } - domain = urlResult.value - } - - const params: Record = { - name: cookie.name, - value: cookie.value, - domain, - path: cookie.path ?? '/', - secure: cookie.secure ?? false, - httpOnly: cookie.httpOnly ?? false, - sameSite: cookie.sameSite ?? 'Lax' - } - if (cookie.expires !== undefined) { - params.expires = cookie.expires - } - - const { success } = (await sender('Network.setCookie', params)) as { success: boolean } - return { success } - }) + return this.auxiliaryCommands.cookieSet(cookie) } - async cookieDelete( - name: string, - domain?: string, - url?: string - ): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - const params: Record = { name } - if (domain) { - params.domain = domain - } - if (url) { - params.url = url - } - // Why: Network.deleteCookies needs a domain or url; infer from the current page if neither was given. - if (!domain && !url) { - const { result: urlResult } = (await sender('Runtime.evaluate', { - expression: 'location.href', - returnByValue: true - })) as { result: { value: string } } - params.url = urlResult.value - } - - await sender('Network.deleteCookies', params) - return { deleted: true } - }) + cookieDelete(name: string, domain?: string, url?: string): Promise { + return this.auxiliaryCommands.cookieDelete(name, domain, url) } // ── Viewport emulation ── - async setViewport( + setViewport( width: number, height: number, deviceScaleFactor = 1, mobile = false ): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - await sender('Emulation.setDeviceMetricsOverride', { - width, - height, - deviceScaleFactor, - mobile - }) - // Why: metrics-only resize can leave the compositor surface at the old size, cropping remote screencast clients. - await Promise.resolve(sender('Emulation.setVisibleSize', { width, height })).catch(() => {}) - - return { width, height, deviceScaleFactor, mobile } - }) + return this.auxiliaryCommands.setViewport(width, height, deviceScaleFactor, mobile) } // ── Geolocation ── - async setGeolocation( + setGeolocation( latitude: number, longitude: number, accuracy = 1 ): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - await sender('Emulation.setGeolocationOverride', { latitude, longitude, accuracy }) - return { latitude, longitude, accuracy } - }) + return this.auxiliaryCommands.setGeolocation(latitude, longitude, accuracy) } // ── Request interception ── - async interceptEnable(patterns: string[] = ['*']): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - const requestPatterns = patterns.map((p) => ({ urlPattern: p })) - await sender('Fetch.enable', { patterns: requestPatterns }) - - state.intercepting = true - state.interceptPatterns = patterns - - return { enabled: true, patterns } - }) + interceptEnable(patterns: string[] = ['*']): Promise { + return this.auxiliaryCommands.interceptEnable(patterns) } - async interceptDisable(): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - await sender('Fetch.disable') - state.intercepting = false - state.interceptPatterns = [] - state.pausedRequests.clear() - - return { disabled: true } - }) + interceptDisable(): Promise { + return this.auxiliaryCommands.interceptDisable() } interceptList(): { requests: BrowserInterceptedRequest[] } { - const guest = this.getActiveGuest() - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - return { requests: [...state.pausedRequests.values()] } + return this.auxiliaryCommands.interceptList() } // TODO: Add interceptContinue/interceptBlock once agent-browser supports per-request decisions (CLI is URL-pattern-only). // ── Console/network capture ── - async captureStart(): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const sender = this.makeCdpSender(guest) - await this.ensureDebuggerAttached(guest) - - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - await sender('Runtime.enable') - state.capturing = true - state.consoleLog = [] - state.networkLog = [] - state.networkRequestMap.clear() - - return { capturing: true } - }) + captureStart(): Promise { + return this.auxiliaryCommands.captureStart() } - async captureStop(): Promise { - return this.enqueueCommand(async () => { - const guest = this.getActiveGuest() - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - state.capturing = false - state.networkRequestMap.clear() - - return { stopped: true } - }) + captureStop(): Promise { + return this.auxiliaryCommands.captureStop() } consoleLog(limit = 100): BrowserConsoleResult { - const guest = this.getActiveGuest() - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - const entries = state.consoleLog.slice(-limit) - return { entries, truncated: state.consoleLog.length > limit } + return this.auxiliaryCommands.consoleLog(limit) } networkLog(limit = 100): BrowserNetworkLogResult { - const guest = this.getActiveGuest() - const tabId = this.resolveTabId(guest.id) - const state = this.getOrCreateTabState(tabId) - - const entries = state.networkLog.slice(-limit) - return { entries, truncated: state.networkLog.length > limit } + return this.auxiliaryCommands.networkLog(limit) } async back(): Promise<{ url: string; title: string }> { @@ -1077,7 +925,7 @@ export class CdpBridge { return null } - private getOrCreateTabState(tabId: string): TabState { + private getOrCreateTabState(tabId: string): CdpTabState { let state = this.tabState.get(tabId) if (!state) { state = { @@ -1100,7 +948,7 @@ export class CdpBridge { return state } - private removeDebuggerListeners(guest: Electron.WebContents, state: TabState): void { + private removeDebuggerListeners(guest: Electron.WebContents, state: CdpTabState): void { const detachListener = state.debuggerDetachListener const messageListener = state.debuggerMessageListener state.debuggerDetachListener = null