feat: Integrate SelectedTextKit for robust text retrieval with multi-strategy support

This commit is contained in:
Beingpax 2025-10-28 18:30:25 +05:45
parent 5bdbb39ac8
commit 357804c03d
5 changed files with 55 additions and 181 deletions

View File

@ -97,6 +97,7 @@ If you encounter any issues or have questions, please:
- [LaunchAtLogin](https://github.com/sindresorhus/LaunchAtLogin) - Launch at login functionality
- [MediaRemoteAdapter](https://github.com/ejbills/mediaremote-adapter) - Media playback control during recording
- [Zip](https://github.com/marmelroy/Zip) - File compression and decompression utilities
- [SelectedTextKit](https://github.com/tisfeng/SelectedTextKit) - A modern macOS library for getting selected text
---

View File

@ -13,6 +13,7 @@
E1A261122CC143AC00B233D1 /* KeyboardShortcuts in Frameworks */ = {isa = PBXBuildFile; productRef = E1A261112CC143AC00B233D1 /* KeyboardShortcuts */; };
E1ADD45A2CC5352A00303ECB /* LaunchAtLogin in Frameworks */ = {isa = PBXBuildFile; productRef = E1ADD4592CC5352A00303ECB /* LaunchAtLogin */; };
E1ADD45F2CC544F100303ECB /* Sparkle in Frameworks */ = {isa = PBXBuildFile; productRef = E1ADD45E2CC544F100303ECB /* Sparkle */; };
E1BBBDCC2EB0DF6700C3ABFE /* SelectedTextKit in Frameworks */ = {isa = PBXBuildFile; productRef = E1BBBDCB2EB0DF6700C3ABFE /* SelectedTextKit */; };
E1D7EF992E35E16C00640029 /* MediaRemoteAdapter in Frameworks */ = {isa = PBXBuildFile; productRef = E1D7EF982E35E16C00640029 /* MediaRemoteAdapter */; };
E1D7EF9A2E35E19B00640029 /* MediaRemoteAdapter in Embed Frameworks */ = {isa = PBXBuildFile; productRef = E1D7EF982E35E16C00640029 /* MediaRemoteAdapter */; settings = {ATTRIBUTES = (CodeSignOnCopy, ); }; };
E1ECEC162E44591300DFFBA8 /* Zip in Frameworks */ = {isa = PBXBuildFile; productRef = E1ECEC152E44591300DFFBA8 /* Zip */; };
@ -86,6 +87,7 @@
E11BBA4E2E5DC555000AB839 /* FluidAudio in Frameworks */,
E1D7EF992E35E16C00640029 /* MediaRemoteAdapter in Frameworks */,
E17382402E4C7D0E001BAEBE /* whisper.xcframework in Frameworks */,
E1BBBDCC2EB0DF6700C3ABFE /* SelectedTextKit in Frameworks */,
E1ADD45F2CC544F100303ECB /* Sparkle in Frameworks */,
E1A261122CC143AC00B233D1 /* KeyboardShortcuts in Frameworks */,
);
@ -165,6 +167,7 @@
E1D7EF982E35E16C00640029 /* MediaRemoteAdapter */,
E1ECEC152E44591300DFFBA8 /* Zip */,
E11BBA4D2E5DC555000AB839 /* FluidAudio */,
E1BBBDCB2EB0DF6700C3ABFE /* SelectedTextKit */,
);
productName = VoiceInk;
productReference = E11473B02CBE0F0A00318EE4 /* VoiceInk.app */;
@ -255,6 +258,7 @@
E1D7EF972E35E16C00640029 /* XCRemoteSwiftPackageReference "mediaremote-adapter" */,
E1ECEC142E44590200DFFBA8 /* XCRemoteSwiftPackageReference "Zip" */,
E11BBA4C2E5DC555000AB839 /* XCRemoteSwiftPackageReference "FluidAudio" */,
E1BBBDCA2EB0DF6700C3ABFE /* XCRemoteSwiftPackageReference "SelectedTextKit" */,
);
preferredProjectObjectVersion = 77;
productRefGroup = E11473B12CBE0F0A00318EE4 /* Products */;
@ -659,6 +663,14 @@
minimumVersion = 2.6.4;
};
};
E1BBBDCA2EB0DF6700C3ABFE /* XCRemoteSwiftPackageReference "SelectedTextKit" */ = {
isa = XCRemoteSwiftPackageReference;
repositoryURL = "https://github.com/tisfeng/SelectedTextKit";
requirement = {
kind = revision;
revision = f6c1574ad6f747e1110b544c35e6ec64ed6f52f1;
};
};
E1D7EF972E35E16C00640029 /* XCRemoteSwiftPackageReference "mediaremote-adapter" */ = {
isa = XCRemoteSwiftPackageReference;
repositoryURL = "https://github.com/ejbills/mediaremote-adapter";
@ -698,6 +710,11 @@
package = E1ADD45D2CC544F100303ECB /* XCRemoteSwiftPackageReference "Sparkle" */;
productName = Sparkle;
};
E1BBBDCB2EB0DF6700C3ABFE /* SelectedTextKit */ = {
isa = XCSwiftPackageProductDependency;
package = E1BBBDCA2EB0DF6700C3ABFE /* XCRemoteSwiftPackageReference "SelectedTextKit" */;
productName = SelectedTextKit;
};
E1D7EF982E35E16C00640029 /* MediaRemoteAdapter */ = {
isa = XCSwiftPackageProductDependency;
package = E1D7EF972E35E16C00640029 /* XCRemoteSwiftPackageReference "mediaremote-adapter" */;

View File

@ -1,6 +1,15 @@
{
"originHash" : "0b9379abd19d2f53581c233273d09235e935a8d2b1180cf253dd69baa2784b39",
"originHash" : "53977eb7bc5f27d05d8be33f99b4c824929a33b2f189fcc8c4c2a6662bcd8075",
"pins" : [
{
"identity" : "axswift",
"kind" : "remoteSourceControl",
"location" : "https://github.com/clavierorg/AXSwift.git",
"state" : {
"revision" : "2efa201fa18184de98674165592fd1ecb0c00a6d",
"version" : "0.3.5"
}
},
{
"identity" : "fluidaudio",
"kind" : "remoteSourceControl",
@ -19,6 +28,15 @@
"version" : "2.4.0"
}
},
{
"identity" : "keysender",
"kind" : "remoteSourceControl",
"location" : "https://github.com/jordanbaird/KeySender",
"state" : {
"revision" : "99584bf1a03ab600cc44be89a2dcd98b2dbeb9de",
"version" : "0.0.5"
}
},
{
"identity" : "launchatlogin-modern",
"kind" : "remoteSourceControl",
@ -37,6 +55,14 @@
"revision" : "3529aa25023082a2ceadebcd2c9c4a9430ee96b9"
}
},
{
"identity" : "selectedtextkit",
"kind" : "remoteSourceControl",
"location" : "https://github.com/tisfeng/SelectedTextKit",
"state" : {
"revision" : "f6c1574ad6f747e1110b544c35e6ec64ed6f52f1"
}
},
{
"identity" : "sparkle",
"kind" : "remoteSourceControl",

View File

@ -139,8 +139,8 @@ class AIEnhancementService: ObservableObject {
lastRequestTime = Date()
}
private func getSystemMessage(for mode: EnhancementPrompt) -> String {
let selectedText = SelectedTextService.fetchSelectedText()
private func getSystemMessage(for mode: EnhancementPrompt) async -> String {
let selectedText = await SelectedTextService.fetchSelectedText()
let selectedTextContext = if let selectedText = selectedText, !selectedText.isEmpty {
"\n\n<CURRENTLY_SELECTED_TEXT>\n\(selectedText)\n</CURRENTLY_SELECTED_TEXT>"
@ -198,7 +198,7 @@ class AIEnhancementService: ObservableObject {
}
let formattedText = "\n<TRANSCRIPT>\n\(text)\n</TRANSCRIPT>"
let systemMessage = getSystemMessage(for: mode)
let systemMessage = await getSystemMessage(for: mode)
// Persist the exact payload being sent (also used for UI)
await MainActor.run {

View File

@ -1,186 +1,16 @@
import Foundation
import AppKit
import ApplicationServices
import SelectedTextKit
class SelectedTextService {
static func fetchSelectedText() -> String? {
guard ensureAccessibilityPermission() else {
return nil
}
let strategies: [() -> String?] = [
getSelectedTextViaAccessibility,
getSelectedTextViaKeyboardCopy
]
for fetch in strategies {
if let text = fetch(), !text.isEmpty {
return text
}
}
return nil
}
private static func getSelectedTextViaAccessibility() -> String? {
let systemWideElement = AXUIElementCreateSystemWide()
var focusedElement: AnyObject?
let focusedResult = AXUIElementCopyAttributeValue(
systemWideElement,
kAXFocusedUIElementAttribute as CFString,
&focusedElement
)
guard focusedResult == .success, let focusedElement else {
return nil
}
let element = focusedElement as! AXUIElement
if let selectedText = stringAttribute(kAXSelectedTextAttribute as CFString, of: element) {
static func fetchSelectedText() async -> String? {
let strategies: [TextStrategy] = [.accessibility, .menuAction, .shortcut]
do {
let selectedText = try await SelectedTextManager.shared.getSelectedText(strategies: strategies)
return selectedText
}
guard
let selectedRange = rangeAttribute(kAXSelectedTextRangeAttribute as CFString, of: element),
let value = stringAttribute(kAXValueAttribute as CFString, of: element),
let textRange = Range(selectedRange, in: value)
else {
} catch {
print("Failed to get selected text: \(error)")
return nil
}
let substring = String(value[textRange])
return substring.isEmpty ? nil : substring
}
private static func getSelectedTextViaKeyboardCopy() -> String? {
return readSelectedTextUsingCopyAction {
let source = CGEventSource(stateID: .hidSystemState)
let cmdDown = CGEvent(keyboardEventSource: source, virtualKey: 0x37, keyDown: true)
cmdDown?.flags = .maskCommand
let cDown = CGEvent(keyboardEventSource: source, virtualKey: 0x08, keyDown: true)
cDown?.flags = .maskCommand
let cUp = CGEvent(keyboardEventSource: source, virtualKey: 0x08, keyDown: false)
cUp?.flags = .maskCommand
let cmdUp = CGEvent(keyboardEventSource: source, virtualKey: 0x37, keyDown: false)
cmdDown?.post(tap: .cghidEventTap)
cDown?.post(tap: .cghidEventTap)
cUp?.post(tap: .cghidEventTap)
cmdUp?.post(tap: .cghidEventTap)
return true
}
}
private static func readSelectedTextUsingCopyAction(_ copyAction: () -> Bool) -> String? {
let pasteboard = NSPasteboard.general
let originalClipboardText = pasteboard.string(forType: .string)
let originalPasteboardItems = pasteboard.pasteboardItems?.map { item in
item.types.reduce(into: [NSPasteboard.PasteboardType: Data]()) { acc, type in
if let data = item.data(forType: type) {
acc[type] = data
}
}
}
defer {
restorePasteboard(
pasteboard,
items: originalPasteboardItems,
string: originalClipboardText
)
}
// Clear clipboard to prepare for selection detection
pasteboard.clearContents()
guard copyAction() else { return nil }
// Wait for copy operation to complete
Thread.sleep(forTimeInterval: 0.1)
// Read the copied text
return pasteboard.string(forType: .string)
}
private static func restorePasteboard(
_ pasteboard: NSPasteboard,
items: [[NSPasteboard.PasteboardType: Data]]?,
string: String?
) {
pasteboard.clearContents()
if let items, !items.isEmpty {
let restoredItems: [NSPasteboardItem] = items.compactMap { dataMap in
guard !dataMap.isEmpty else { return nil }
let item = NSPasteboardItem()
for (type, data) in dataMap {
item.setData(data, forType: type)
}
return item
}
if !restoredItems.isEmpty {
pasteboard.writeObjects(restoredItems)
return
}
}
if let string {
_ = pasteboard.setString(string, forType: .string)
}
}
private static func stringAttribute(_ attribute: CFString, of element: AXUIElement) -> String? {
var value: AnyObject?
let result = AXUIElementCopyAttributeValue(element, attribute, &value)
guard result == .success, let value else {
return nil
}
if let text = value as? String {
return text
}
if let attributed = value as? NSAttributedString {
return attributed.string
}
return nil
}
private static func rangeAttribute(_ attribute: CFString, of element: AXUIElement) -> NSRange? {
var value: AnyObject?
let result = AXUIElementCopyAttributeValue(element, attribute, &value)
guard result == .success, let value else {
return nil
}
guard CFGetTypeID(value) == AXValueGetTypeID() else {
return nil
}
let rangeValue = value as! AXValue
guard AXValueGetType(rangeValue) == .cfRange else {
return nil
}
var range = CFRange()
let success = AXValueGetValue(rangeValue, .cfRange, &range)
guard success else {
return nil
}
return NSRange(location: range.location, length: range.length)
}
private static func ensureAccessibilityPermission() -> Bool {
AXIsProcessTrustedWithOptions(nil)
}
}