Open websites in the Page view and let agents drive them

This commit is contained in:
Ian Gulliver
2026-10-21 21:34:26 -07:00
parent d6e1895591
commit 57d660c57f
8 changed files with 786 additions and 390 deletions
+248 -24
View File
@@ -8,10 +8,11 @@ import WebKit
enum Page: Hashable {
case file(taskId: String, path: String)
case preview(taskId: String, port: Int)
case web(taskId: String, id: Int)
var taskId: String {
switch self {
case .file(let taskId, _), .preview(let taskId, _): taskId
case .file(let taskId, _), .preview(let taskId, _), .web(let taskId, _): taskId
}
}
@@ -27,6 +28,7 @@ enum Page: Hashable {
switch self {
case .file(_, let path): (path as NSString).lastPathComponent
case .preview(let taskId, let port): store.forwards.name(taskId, port) ?? "port \(port)"
case .web: store.pages.title(self)
}
}
@@ -35,6 +37,7 @@ enum Page: Hashable {
switch self {
case .file(let taskId, let path): store.pages.url(taskId, path)
case .preview(let taskId, let port): store.forwards.url(taskId, port)
case .web: store.pages.existing(self)?.url
}
}
}
@@ -61,10 +64,18 @@ extension Store {
sheet = Sheet(owner: .task(page.taskId), view: .tab(.page))
}
func openWeb(_ taskId: String, _ url: URL) -> Int {
route("#/task/\(taskId)")
let page = pages.web(taskId, url)
openPage(page)
return pages.tab(page)
}
func closePage(_ page: Page) {
guard var shown = shownPages[page.taskId], let i = shown.pages.firstIndex(of: page) else {
return
}
pages.drop(page)
shown.pages.remove(at: i)
if shown.pages.isEmpty {
shownPages[page.taskId] = nil
@@ -82,16 +93,67 @@ extension Store {
func currentPage() -> Page? {
selectedTaskId.flatMap { shownPages[$0]?.current }
}
func pageView(_ page: Page) -> WKWebView? {
if case .web = page {
return pages.existing(page)
}
return page.url(self).map { pages.view(page, $0) }
}
func pageTabs(_ taskId: String) -> String {
guard let shown = shownPages[taskId] else {
return "This task's page view has no tabs; open one with show_page."
}
return shown.pages.map { page in
var line = "tab \(pages.tab(page)): \(page.title(self))"
if let url = pages.existing(page)?.url ?? page.url(self) {
line += " — \(url.absoluteString)"
}
if page == shown.current {
line += " (showing)"
}
return line
}.joined(separator: "\n")
}
func drivenPage(_ taskId: String, _ tab: Int?) throws -> WKWebView {
guard let shown = shownPages[taskId] else {
throw ApiError(status: 0, message: "this task's page view has no tabs; open one with show_page")
}
var page = shown.current
if let tab {
guard let p = shown.pages.first(where: { pages.tab($0) == tab }) else {
throw ApiError(status: 0, message: "this task's page view has no tab \(tab); drive_page tabs lists them")
}
page = p
}
guard let view = pageView(page) else {
throw ApiError(status: 0, message: "tab \(pages.tab(page)) isn't ready yet")
}
return view
}
}
@MainActor
@Observable
final class Pages {
private(set) var port: Int?
private(set) var titles: [Page: String] = [:]
@ObservationIgnored var served: [String: [String: Data]] = [:]
@ObservationIgnored var popup: ((Page) -> Void)?
@ObservationIgnored private var listener: PageListener?
@ObservationIgnored private var views: [Page: WKWebView] = [:]
@ObservationIgnored private var sources: [Page: URL] = [:]
@ObservationIgnored private var titleWatches: [Page: NSKeyValueObservation] = [:]
@ObservationIgnored private var tabs: [Page: Int] = [:]
@ObservationIgnored private var lastTab = 0
@ObservationIgnored private let delegate = PageDelegate()
init(api: API) {
delegate.popup = { [weak self] opener, config in
self?.adopt(opener, config)
}
listener = PageListener(api: api, ready: { [weak self] port in
Task { @MainActor in
self?.port = port
@@ -115,6 +177,108 @@ final class Pages {
return c.url
}
func existing(_ page: Page) -> WKWebView? {
views[page]
}
func view(_ page: Page, _ url: URL) -> WKWebView {
let view = views[page] ?? make(page, WKWebViewConfiguration())
if sources[page] != url {
sources[page] = url
view.load(URLRequest(url: url))
}
return view
}
func web(_ taskId: String, _ url: URL) -> Page {
lastTab += 1
let page = Page.web(taskId: taskId, id: lastTab)
make(page, WKWebViewConfiguration()).load(URLRequest(url: url))
return page
}
private func adopt(_ opener: WKWebView, _ config: WKWebViewConfiguration) -> WKWebView? {
guard let from = views.first(where: { $0.value === opener })?.key else {
return nil
}
lastTab += 1
let page = Page.web(taskId: from.taskId, id: lastTab)
let view = make(page, config)
popup?(page)
return view
}
private func make(_ page: Page, _ config: WKWebViewConfiguration) -> WKWebView {
config.websiteDataStore = .default()
let view = WKWebView(frame: .zero, configuration: config)
view.uiDelegate = delegate
view.navigationDelegate = delegate
views[page] = view
if case .web(_, let id) = page {
tabs[page] = id
titleWatches[page] = view.observe(\.title) { [weak self] view, _ in
Task { @MainActor in
self?.titles[page] = view.title
}
}
}
return view
}
func drop(_ page: Page) {
views[page] = nil
sources[page] = nil
titleWatches[page] = nil
titles[page] = nil
tabs[page] = nil
}
func tab(_ page: Page) -> Int {
if let tab = tabs[page] {
return tab
}
lastTab += 1
tabs[page] = lastTab
return lastTab
}
func title(_ page: Page) -> String {
if let title = titles[page], !title.isEmpty {
return title
}
return views[page]?.url?.host() ?? "page"
}
func navigate(_ view: WKWebView, _ url: URL) async throws -> String {
try await delegate.wait(view.load(URLRequest(url: url)))
return "Loaded \(view.url?.absoluteString ?? url.absoluteString): \(view.title ?? "")"
}
func js(_ view: WKWebView, _ code: String) async throws -> String {
let value = try await view.callAsyncJavaScript(code, contentWorld: .page)
guard let value, !(value is NSNull) else {
return value == nil ? "undefined" : "null"
}
if let s = value as? String {
return s
}
if let data = try? JSONSerialization.data(withJSONObject: value, options: [.fragmentsAllowed, .sortedKeys]) {
return String(decoding: data, as: UTF8.self)
}
return String(describing: value)
}
func capture(_ view: WKWebView) async throws -> Data {
guard unsafe view.window != nil, !view.isHiddenOrHasHiddenAncestor else {
throw ApiError(status: 0, message: "the tab isn't on screen: the human has the page view closed or another tab showing")
}
let image = try await view.takeSnapshot(configuration: nil)
guard let tiff = image.tiffRepresentation, let png = NSBitmapImageRep(data: tiff)?.representation(using: .png, properties: [:]) else {
throw ApiError(status: 0, message: "couldn't encode the capture as PNG")
}
return png
}
func close() {
listener?.close()
listener = nil
@@ -299,8 +463,8 @@ struct PageTabs: View {
}
ZStack {
ForEach(shown.pages, id: \.self) { page in
if let url = page.url(store) {
WebPage(url: url, reloads: page.isFile ? watch.reloads : 0, shown: page == shown.current)
if let view = store.pageView(page) {
WebPage(view: view, reloads: page.isFile ? watch.reloads : 0, shown: page == shown.current, refresh: page.isFile)
} else if page == shown.current {
EmptyText(text: "Loading…")
}
@@ -352,21 +516,87 @@ private struct PageTabButton: View {
}
}
private struct WebPage: NSViewRepresentable {
let url: URL
let reloads: Int
let shown: Bool
@MainActor
private final class PageDelegate: NSObject, WKUIDelegate, WKNavigationDelegate {
var popup: ((WKWebView, WKWebViewConfiguration) -> WKWebView?)?
private var waits: [ObjectIdentifier: CheckedContinuation<Void, any Error>] = [:]
final class Coordinator: NSObject, WKUIDelegate {
var url: URL?
var reloads = 0
func wait(_ navigation: WKNavigation?) async throws {
guard let navigation else {
return
}
try await withCheckedThrowingContinuation { c in
waits[ObjectIdentifier(navigation)] = c
}
}
func webView(_ webView: WKWebView, createWebViewWith configuration: WKWebViewConfiguration, for action: WKNavigationAction, windowFeatures: WKWindowFeatures) -> WKWebView? {
if let url = action.request.url {
NSWorkspace.shared.open(url)
}
private func finish(_ navigation: WKNavigation?, _ error: (any Error)?) {
guard let navigation, let c = waits.removeValue(forKey: ObjectIdentifier(navigation)) else {
return
}
if let error {
c.resume(throwing: error)
} else {
c.resume()
}
}
func webView(_ webView: WKWebView, createWebViewWith configuration: WKWebViewConfiguration, for action: WKNavigationAction, windowFeatures: WKWindowFeatures) -> WKWebView? {
popup?(webView, configuration)
}
func webView(_ webView: WKWebView, runJavaScriptAlertPanelWithMessage message: String, initiatedByFrame frame: WKFrameInfo) async {
_ = await ask(webView, frame, message, buttons: ["OK"], field: nil)
}
func webView(_ webView: WKWebView, runJavaScriptConfirmPanelWithMessage message: String, initiatedByFrame frame: WKFrameInfo) async -> Bool {
await ask(webView, frame, message, buttons: ["OK", "Cancel"], field: nil) == .alertFirstButtonReturn
}
func webView(_ webView: WKWebView, runJavaScriptTextInputPanelWithPrompt prompt: String, defaultText: String?, initiatedByFrame frame: WKFrameInfo) async -> String? {
let field = NSTextField(frame: NSRect(x: 0, y: 0, width: 300, height: 24))
field.stringValue = defaultText ?? ""
guard await ask(webView, frame, prompt, buttons: ["OK", "Cancel"], field: field) == .alertFirstButtonReturn else {
return nil
}
return field.stringValue
}
private func ask(_ webView: WKWebView, _ frame: WKFrameInfo, _ message: String, buttons: [String], field: NSTextField?) async -> NSApplication.ModalResponse {
guard let window = unsafe webView.window, !webView.isHiddenOrHasHiddenAncestor else {
return .cancel
}
let alert = NSAlert()
alert.messageText = frame.securityOrigin.host.isEmpty ? "This page says" : "\(frame.securityOrigin.host) says"
alert.informativeText = message
for title in buttons {
alert.addButton(withTitle: title)
}
alert.accessoryView = field
return await alert.beginSheetModal(for: window)
}
func webView(_ webView: WKWebView, didFinish navigation: WKNavigation!) {
finish(navigation, nil)
}
func webView(_ webView: WKWebView, didFail navigation: WKNavigation!, withError error: any Error) {
finish(navigation, error)
}
func webView(_ webView: WKWebView, didFailProvisionalNavigation navigation: WKNavigation!, withError error: any Error) {
finish(navigation, error)
}
}
private struct WebPage: NSViewRepresentable {
let view: WKWebView
let reloads: Int
let shown: Bool
let refresh: Bool
final class Coordinator {
var reloads = 0
}
func makeCoordinator() -> Coordinator {
@@ -374,22 +604,16 @@ private struct WebPage: NSViewRepresentable {
}
func makeNSView(context: Context) -> WKWebView {
let config = WKWebViewConfiguration()
config.websiteDataStore = .nonPersistent()
let view = WKWebView(frame: .zero, configuration: config)
view.uiDelegate = context.coordinator
context.coordinator.reloads = reloads
if refresh, !view.isLoading {
view.reload()
}
return view
}
func updateNSView(_ view: WKWebView, context: Context) {
view.isHidden = !shown
let c = context.coordinator
if c.url != url {
c.url = url
c.reloads = reloads
view.load(URLRequest(url: url))
return
}
if c.reloads != reloads {
c.reloads = reloads
view.reload()
+20
View File
@@ -190,11 +190,31 @@ final class Remote {
let (code, status, signature) = try await sign(Data(req.data?.data ?? []), as: req.ident ?? "", timeoutMs: req.timeoutMs ?? 120_000)
return ClientReply(id: req.id, exitCode: code, output: status, data: Base64EncodedData(signature))
case .openPage:
if let address = req.url {
guard let url = URL(string: address) else {
throw ApiError(status: 0, message: "\(address) isn't a URL")
}
return ClientReply(id: req.id, output: "tab \(store!.openWeb(req.task ?? "", url))")
}
store!.showPage(req.task ?? "", req.path ?? "")
return ClientReply(id: req.id)
case .openFile:
store!.showInFiles(req.task ?? "", req.path ?? "")
return ClientReply(id: req.id)
case .pageTabs:
return ClientReply(id: req.id, output: store!.pageTabs(req.task ?? ""))
case .pageNavigate:
let view = try store!.drivenPage(req.task ?? "", req.tab)
guard let url = URL(string: req.url ?? "") else {
throw ApiError(status: 0, message: "\(req.url ?? "") isn't a URL")
}
return ClientReply(id: req.id, output: try await store!.pages.navigate(view, url))
case .pageJs:
let view = try store!.drivenPage(req.task ?? "", req.tab)
return ClientReply(id: req.id, output: try await store!.pages.js(view, req.code ?? ""))
case .pageCapture:
let view = try store!.drivenPage(req.task ?? "", req.tab)
return ClientReply(id: req.id, data: Base64EncodedData(try await store!.pages.capture(view)))
}
}
+3
View File
@@ -114,6 +114,9 @@ final class Store {
paneFocus = PaneFocus(api: api)
remote = Remote(api: api)
remote.store = self
pages.popup = { [weak self] page in
self?.openPage(page)
}
}
func sends(_ key: String) -> LocalSends {