agent-cli 0.95.3__tar.gz → 0.95.5__tar.gz

{agent_cli-0.95.3 → agent_cli-0.95.5}/.github/workflows/release.yml

@@ -56,6 +56,8 @@ jobs:
           APPLE_ID: ${{ secrets.APPLE_ID }}
           APPLE_APP_SPECIFIC_PASSWORD: ${{ secrets.APPLE_APP_SPECIFIC_PASSWORD }}
           APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
+          APP_VERSION: ${{ github.event.release.tag_name }}
+          BUILD_VERSION: ${{ github.run_number }}
         run: NOTARIZE=1 macos/build-macos-app.sh --dmg
       - name: Upload macOS app to GitHub Release
         env:

{agent_cli-0.95.3 → agent_cli-0.95.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agent-cli
-Version: 0.95.3
+Version: 0.95.5
 Summary: A suite of AI-powered command-line tools for text correction, audio transcription, and voice assistance.
 Project-URL: Homepage, https://github.com/basnijholt/agent-cli
 Author-email: Bas Nijholt <bas@nijho.lt>

{agent_cli-0.95.3 → agent_cli-0.95.5}/agent_cli/agents/transcribe.py

@@ -252,6 +252,7 @@ def log_transcription(
     }
 
     # Append to log file
+    log_file.parent.mkdir(parents=True, exist_ok=True)
     with log_file.open("a", encoding="utf-8") as f:
         f.write(json.dumps(log_entry) + "\n")
 

{agent_cli-0.95.3 → agent_cli-0.95.5}/macos/AgentCLI/README.md

@@ -1,9 +1,12 @@
 # Agent CLI macOS app
 
 This directory contains a native SwiftUI menu bar wrapper for `agent-cli`.
-The app does not reimplement any agent behavior. It bundles `uv`, installs a
-private `agent-cli[audio,llm]` tool into the user's Application Support
-directory on first use, and shells out to that private executable.
+The app does not reimplement any agent behavior. By default, it bundles `uv`,
+installs a private `agent-cli[audio,llm]` tool into the user's Application
+Support directory on first use, and shells out to that private executable.
+Users who already manage their own `agent-cli` install can enable
+**Use User-Installed agent-cli** in Settings to run the `agent-cli` found on
+PATH with their normal config instead.
 
 Transcription is the default zero-config path. The first transcription action
 installs and starts the local Whisper launchd daemon with
@@ -18,9 +21,14 @@ The packaged app registers native macOS global hotkeys itself:
 - `Cmd+Shift+A` autocorrects clipboard text
 - `Cmd+Shift+V` starts voice edit
 
-Choose **Keyboard Shortcuts...** from the menu bar app to change these
-shortcuts. The settings UI uses the `KeyboardShortcuts` Swift package for
-shortcut parsing, `UserDefaults` storage, and global key-up handlers.
+Choose **Settings...** from the menu bar app to change these shortcuts, enable
+**Start at Login**, or switch between the bundled runtime and a user-installed
+`agent-cli`. The settings UI uses the `KeyboardShortcuts` Swift package for
+shortcut parsing and `UserDefaults` storage. Transcription shortcuts are handled
+by a small CGEvent tap so `Fn`, `Fn+Space`, and plain Space remain distinct; the
+clipboard utility shortcuts still use `KeyboardShortcuts` global handlers. The
+login option uses Apple's login item API for the main app bundle, so macOS may
+require approval in System Settings → General → Login Items.
 
 ## Build
 

agent_cli-0.95.5/macos/AgentCLI/Resources/dmg-background.svg

@@ -0,0 +1,8 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="600" height="360" viewBox="0 0 600 360">
+  <rect width="600" height="360" fill="#f6f7f9"/>
+  <rect x="28" y="28" width="544" height="304" rx="24" fill="#ffffff" stroke="#d7dbe2"/>
+  <path d="M255 180h110" fill="none" stroke="#7b8494" stroke-width="10" stroke-linecap="round"/>
+  <path d="M335 140l40 40-40 40" fill="none" stroke="#7b8494" stroke-width="10" stroke-linecap="round" stroke-linejoin="round"/>
+  <text x="300" y="78" text-anchor="middle" font-family="-apple-system, BlinkMacSystemFont, Helvetica, Arial, sans-serif" font-size="24" font-weight="600" fill="#242933">Install Agent CLI</text>
+  <text x="300" y="112" text-anchor="middle" font-family="-apple-system, BlinkMacSystemFont, Helvetica, Arial, sans-serif" font-size="15" fill="#596171">Drag AgentCLI.app to Applications</text>
+</svg>

{agent_cli-0.95.3 → agent_cli-0.95.5}/macos/AgentCLI/Sources/AgentCLI/AgentCLIApp.swift

@@ -6,7 +6,10 @@ struct AgentCLIApp: App {
     @NSApplicationDelegateAdaptor(AppDelegate.self) private var appDelegate
     @Environment(\.openWindow) private var openWindow
     @StateObject private var runner = AgentCommandRunner.shared
+    @StateObject private var loginItemController = LoginItemController.shared
     @StateObject private var shortcutSummary = ShortcutSummaryState.shared
+    @AppStorage(RuntimeSettings.useUserInstalledAgentCLIKey)
+    private var useUserInstalledAgentCLI = false
 
     var body: some Scene {
         MenuBarExtra {
@@ -38,11 +41,40 @@ struct AgentCLIApp: App {
 
             Divider()
 
+            Menu {
+                let recentTranscriptions = RecentTranscriptionReader.recentTranscriptions()
+                if recentTranscriptions.isEmpty {
+                    Text("No recent transcriptions")
+                } else {
+                    ForEach(recentTranscriptions) { transcription in
+                        Button {
+                            runner.copyRecentTranscription(transcription)
+                        } label: {
+                            Label(transcription.menuTitle, systemImage: "doc.on.clipboard")
+                        }
+                    }
+                }
+            } label: {
+                Label("Recent Recordings", systemImage: "clock.arrow.circlepath")
+            }
+
+            Divider()
+
+            Button {
+                loginItemController.toggle()
+            } label: {
+                Label(
+                    loginItemController.presentation.menuTitle,
+                    systemImage: loginItemController.presentation.isEnabled ? "checkmark.circle" : "circle"
+                )
+            }
+            .disabled(!loginItemController.presentation.canToggle)
+
             Button {
                 openWindow(id: "settings")
                 NSApp.activate(ignoringOtherApps: true)
             } label: {
-                Label("Keyboard Shortcuts...", systemImage: "command")
+                Label("Settings...", systemImage: "gearshape")
             }
 
             Menu {
@@ -55,7 +87,10 @@ struct AgentCLIApp: App {
                 Button {
                     runner.run(.installOrUpdateCLI)
                 } label: {
-                    Label("Update CLI Runtime", systemImage: "arrow.down.circle")
+                    Label(
+                        useUserInstalledAgentCLI ? "Check User CLI" : "Update CLI Runtime",
+                        systemImage: useUserInstalledAgentCLI ? "checkmark.circle" : "arrow.down.circle"
+                    )
                 }
 
                 Button {
@@ -103,15 +138,15 @@ struct AgentCLIApp: App {
                 Divider()
 
                 Button {
-                    runner.openNotificationSettings()
+                    runner.repairNotificationPermission()
                 } label: {
-                    Label("Open Notification Settings", systemImage: "bell.badge")
+                    Label("Fix Notification Permission...", systemImage: "bell.badge")
                 }
 
                 Button {
-                    runner.openAccessibilitySettings()
+                    runner.resetAccessibilityPermission()
                 } label: {
-                    Label("Open Accessibility Settings", systemImage: "figure.wave")
+                    Label("Reset Accessibility Permission...", systemImage: "figure.wave")
                 }
 
                 Button {

{agent_cli-0.95.3 → agent_cli-0.95.5}/macos/AgentCLI/Sources/AgentCLI/AgentCommand.swift

@@ -3,12 +3,14 @@ import Foundation
 enum AgentBootstrapRequirement: Equatable {
     case cliRuntime
     case transcription
+    case transcriptionModel
 }
 
 struct AgentCommand {
     let identifier: String
     let title: String
     let arguments: [String]
+    let appliesTranscriptionExtraInstructions: Bool
     let forceBootstrap: Bool
     let bootstrapRequirement: AgentBootstrapRequirement
     let showsRecordingIndicator: Bool
@@ -20,6 +22,7 @@ struct AgentCommand {
         identifier: String,
         title: String,
         arguments: [String],
+        appliesTranscriptionExtraInstructions: Bool = false,
         forceBootstrap: Bool = false,
         bootstrapRequirement: AgentBootstrapRequirement = .cliRuntime,
         showsRecordingIndicator: Bool = false,
@@ -30,6 +33,7 @@ struct AgentCommand {
         self.identifier = identifier
         self.title = title
         self.arguments = arguments
+        self.appliesTranscriptionExtraInstructions = appliesTranscriptionExtraInstructions
         self.forceBootstrap = forceBootstrap
         self.bootstrapRequirement = bootstrapRequirement
         self.showsRecordingIndicator = showsRecordingIndicator
@@ -38,10 +42,26 @@ struct AgentCommand {
         self.finishNotificationTitle = finishNotificationTitle
     }
 
+    func resolvedArguments(extraInstructions: String?) -> [String] {
+        guard appliesTranscriptionExtraInstructions else { return arguments }
+
+        let trimmedInstructions = extraInstructions?.trimmingCharacters(in: .whitespacesAndNewlines) ?? ""
+        guard !trimmedInstructions.isEmpty else { return arguments }
+
+        return arguments + ["--extra-instructions", trimmedInstructions]
+    }
+
     static let toggleTranscription = AgentCommand(
         identifier: "transcribe",
         title: "Toggle Transcription",
-        arguments: ["transcribe", "--toggle", "--quiet"],
+        arguments: [
+            "transcribe",
+            "--toggle",
+            "--quiet",
+            "--transcription-log",
+            RecentTranscriptionReader.defaultLogPath,
+        ],
+        appliesTranscriptionExtraInstructions: true,
         bootstrapRequirement: .transcription,
         showsRecordingIndicator: true,
         startNotificationTitle: "Transcription Started",

{agent_cli-0.95.3 → agent_cli-0.95.5}/macos/AgentCLI/Sources/AgentCLI/AgentCommandRunner.swift

@@ -1,4 +1,5 @@
 import AppKit
+import ApplicationServices
 import Foundation
 import SwiftUI
 import UserNotifications
@@ -18,6 +19,8 @@ private enum HoldTranscriptionState {
     }
 }
 
+typealias AgentBootstrap = @Sendable (AgentBootstrapRequirement, Bool) -> CommandResult
+
 @MainActor
 final class AgentCommandRunner: ObservableObject {
     static let shared = AgentCommandRunner()
@@ -29,10 +32,12 @@ final class AgentCommandRunner: ObservableObject {
     @Published private var activeCommandCount = 0
     private var recordingIndicator = RecordingIndicatorController()
     private let pasteController: TranscriptPasteController
+    private let bootstrap: AgentBootstrap
     private var pendingStopRecordingCommands: Set<String> = []
     private var holdTranscriptionState: HoldTranscriptionState = .idle
     private var holdToTranscribePasteTarget: FocusedTextTarget?
     private var pasteAfterRecordingCommands: Set<String> = []
+    private var hasStartedTranscriptionWarmUp = false
 
     var isRunning: Bool {
         activeCommandCount > 0
@@ -51,10 +56,14 @@ final class AgentCommandRunner: ObservableObject {
         return Self.compactMenuStatus(statusMessage)
     }
 
-    private init(
-        pasteController: TranscriptPasteController = TranscriptPasteController()
+    init(
+        pasteController: TranscriptPasteController = TranscriptPasteController(),
+        bootstrap: @escaping AgentBootstrap = { requirement, force in
+            AgentRuntime.shared.ensureReady(for: requirement, force: force)
+        }
     ) {
         self.pasteController = pasteController
+        self.bootstrap = bootstrap
         hasLastError = FileManager.default.fileExists(atPath: AgentRuntime.shared.lastErrorURL.path)
     }
 
@@ -68,22 +77,58 @@ final class AgentCommandRunner: ObservableObject {
         URL(string: "x-apple.systempreferences:com.apple.preference.security?Privacy_Accessibility")!
     ]
 
-    func beginHoldToTranscribe() {
+    func warmUpTranscription() {
+        guard !hasStartedTranscriptionWarmUp else { return }
+        hasStartedTranscriptionWarmUp = true
+
+        activeCommandCount += 1
+        if statusMessage == "Ready" {
+            statusMessage = "Preparing voice service..."
+        }
+
+        let bootstrap = self.bootstrap
+        DispatchQueue.global(qos: .utility).async {
+            let result = bootstrap(.transcriptionModel, false)
+
+            Task { @MainActor in
+                self.activeCommandCount = max(0, self.activeCommandCount - 1)
+                if !result.output.isEmpty {
+                    self.lastOutput = result.output
+                }
+
+                if result.exitCode == 0 {
+                    if self.statusMessage == "Preparing voice service..." {
+                        self.statusMessage = "Ready"
+                    }
+                    return
+                }
+
+                self.recordFailure(title: "Startup Voice Service Warm-Up", result: result)
+                self.statusMessage = result.output.isEmpty
+                    ? "Voice service warm-up failed with exit code \(result.exitCode)"
+                    : "Voice service warm-up failed: \(Self.summarize(result.output))"
+            }
+        }
+    }
+
+    @discardableResult
+    func beginHoldToTranscribe() -> Bool {
         guard holdTranscriptionState == .idle else {
             if holdTranscriptionState.isFinishing {
                 statusMessage = "Finishing previous hold-to-transcribe request"
             }
-            return
+            return false
         }
         guard !recordingIndicator.isRecordingCommand(.toggleTranscription), !isStopPending(for: .toggleTranscription) else {
             statusMessage = "Transcription is already recording"
-            return
+            return false
         }
 
         holdTranscriptionState = .recording
         holdToTranscribePasteTarget = FocusedTextTarget.capture()
         pasteAfterRecordingCommands.insert(AgentCommand.toggleTranscription.identifier)
         run(.toggleTranscription)
+        return true
     }
 
     func endHoldToTranscribe() {
@@ -100,6 +145,18 @@ final class AgentCommandRunner: ObservableObject {
         stopHeldTranscriptionWhenReady()
     }
 
+    @discardableResult
+    func stopTranscriptionFromFunctionKeyIfNeeded() -> Bool {
+        guard holdTranscriptionState == .idle,
+              recordingIndicator.isRecordingCommand(.toggleTranscription),
+              !isStopPending(for: .toggleTranscription) else {
+            return false
+        }
+
+        run(.toggleTranscription)
+        return true
+    }
+
     func run(_ command: AgentCommand) {
         let isStopRequest = command.showsRecordingIndicator && recordingIndicator.isRecordingCommand(command)
         let shouldStartRecording = command.showsRecordingIndicator && !isStopRequest
@@ -118,12 +175,14 @@ final class AgentCommandRunner: ObservableObject {
             ? "Stopping \(command.title)..."
             : "Running \(command.title)..."
 
+        let bootstrap = self.bootstrap
+        let commandArguments = command.resolvedArguments(extraInstructions: TranscriptionSettings.extraInstructions)
         DispatchQueue.global(qos: .userInitiated).async {
-            let bootstrap = AgentRuntime.shared.ensureReady(for: command.bootstrapRequirement, force: command.forceBootstrap)
-            guard bootstrap.exitCode == 0 else {
-                let message = Self.statusMessage(for: command, result: bootstrap)
-                let notificationTitle = Self.notificationTitle(for: command, result: bootstrap)
-                let notificationBody = Self.notificationBody(for: command, result: bootstrap, statusMessage: message)
+            let bootstrapResult = bootstrap(command.bootstrapRequirement, command.forceBootstrap)
+            guard bootstrapResult.exitCode == 0 else {
+                let message = Self.statusMessage(for: command, result: bootstrapResult)
+                let notificationTitle = Self.notificationTitle(for: command, result: bootstrapResult)
+                let notificationBody = Self.notificationBody(for: command, result: bootstrapResult, statusMessage: message)
                 Task { @MainActor in
                     if isStopRequest {
                         self.clearStopRequested(for: command)
@@ -133,8 +192,8 @@ final class AgentCommandRunner: ObservableObject {
                     }
                     self.clearHoldTranscriptionState(for: command)
                     self.activeCommandCount = max(0, self.activeCommandCount - 1)
-                    self.lastOutput = bootstrap.output
-                    self.recordFailure(command: command, result: bootstrap)
+                    self.lastOutput = bootstrapResult.output
+                    self.recordFailure(command: command, result: bootstrapResult)
                     self.statusMessage = message
                     self.notify(title: notificationTitle, body: notificationBody)
                 }
@@ -149,7 +208,7 @@ final class AgentCommandRunner: ObservableObject {
                 }
             }
 
-            let result = AgentRuntime.shared.runAgentCLI(arguments: command.arguments)
+            let result = AgentRuntime.shared.runAgentCLI(arguments: commandArguments)
             let message = Self.statusMessage(for: command, result: result)
             let notificationTitle = Self.notificationTitle(for: command, result: result)
             let notificationBody = Self.notificationBody(for: command, result: result, statusMessage: message)
@@ -201,17 +260,18 @@ final class AgentCommandRunner: ObservableObject {
 
         statusMessage = "Stopping Toggle Transcription..."
 
+        let bootstrap = self.bootstrap
         DispatchQueue.global(qos: .userInitiated).async {
-            let bootstrap = AgentRuntime.shared.ensureReady(
-                for: AgentCommand.stopTranscription.bootstrapRequirement,
-                force: AgentCommand.stopTranscription.forceBootstrap
+            let bootstrapResult = bootstrap(
+                AgentCommand.stopTranscription.bootstrapRequirement,
+                AgentCommand.stopTranscription.forceBootstrap
             )
-            guard bootstrap.exitCode == 0 else {
-                let message = Self.statusMessage(for: AgentCommand.stopTranscription, result: bootstrap)
+            guard bootstrapResult.exitCode == 0 else {
+                let message = Self.statusMessage(for: AgentCommand.stopTranscription, result: bootstrapResult)
                 Task { @MainActor in
                     self.holdTranscriptionState = .idle
-                    self.lastOutput = bootstrap.output
-                    self.recordFailure(command: AgentCommand.stopTranscription, result: bootstrap)
+                    self.lastOutput = bootstrapResult.output
+                    self.recordFailure(command: AgentCommand.stopTranscription, result: bootstrapResult)
                     self.statusMessage = message
                     self.notify(
                         title: "Toggle Transcription Failed",
@@ -293,6 +353,12 @@ final class AgentCommandRunner: ObservableObject {
         statusMessage = "Copied last output"
     }
 
+    func copyRecentTranscription(_ transcription: RecentTranscription) {
+        NSPasteboard.general.clearContents()
+        NSPasteboard.general.setString(transcription.text, forType: .string)
+        statusMessage = "Copied recent transcription"
+    }
+
     func openLastError() {
         guard FileManager.default.fileExists(atPath: AgentRuntime.shared.lastErrorURL.path) else {
             hasLastError = false
@@ -348,23 +414,96 @@ final class AgentCommandRunner: ObservableObject {
     }
 
     func notificationsDisabled() {
-        statusMessage = "Notifications are disabled. Use Open Notification Settings to enable Agent CLI notifications."
+        statusMessage = "Notifications are disabled. Use Fix Notification Permission to enable Agent CLI notifications."
+    }
+
+    func repairNotificationPermission() {
+        let center = UNUserNotificationCenter.current()
+        center.getNotificationSettings { settings in
+            switch settings.authorizationStatus {
+            case .notDetermined:
+                UNUserNotificationCenter.current().requestAuthorization(options: [.alert]) { granted, _ in
+                    Task { @MainActor in
+                        self.statusMessage = granted
+                            ? "Notification permission enabled"
+                            : "Notifications are disabled. Enable Agent CLI in Notification Settings."
+                    }
+                }
+            case .denied:
+                Task { @MainActor in
+                    _ = self.openNotificationSettings()
+                    self.statusMessage = "Notifications are disabled. Enable Agent CLI in Notification Settings."
+                }
+            default:
+                Task { @MainActor in
+                    self.statusMessage = "Notification permission is already enabled"
+                }
+            }
+        }
     }
 
-    func openNotificationSettings() {
+    @discardableResult
+    func openNotificationSettings() -> Bool {
         for url in Self.notificationSettingsURLs where NSWorkspace.shared.open(url) {
             statusMessage = "Opened Notification Settings"
-            return
+            return true
         }
         statusMessage = "Could not open Notification Settings"
+        return false
     }
 
-    func openAccessibilitySettings() {
+    func resetAccessibilityPermission() {
+        let result = runTCCReset(service: "Accessibility")
+        try? FileManager.default.removeItem(at: AgentRuntime.shared.accessibilityPromptMarkerURL)
+        requestAccessibilityPermissionPrompt()
+        _ = openAccessibilitySettings()
+
+        guard result.exitCode == 0 else {
+            let output = result.output.trimmingCharacters(in: .whitespacesAndNewlines)
+            statusMessage = output.isEmpty
+                ? "Could not reset Accessibility permission"
+                : "Could not reset Accessibility permission: \(output)"
+            return
+        }
+
+        statusMessage = "Accessibility permission reset. Enable Agent CLI in Accessibility, then reopen AgentCLI if auto-insert still fails."
+    }
+
+    @discardableResult
+    func openAccessibilitySettings() -> Bool {
         for url in Self.accessibilitySettingsURLs where NSWorkspace.shared.open(url) {
             statusMessage = "Opened Accessibility Settings. Accessibility permission controls auto-inserting transcripts."
-            return
+            return true
         }
         statusMessage = "Could not open Accessibility Settings"
+        return false
+    }
+
+    private func requestAccessibilityPermissionPrompt() {
+        let promptOption = kAXTrustedCheckOptionPrompt.takeUnretainedValue() as String
+        let options = [promptOption: true] as CFDictionary
+        _ = AXIsProcessTrustedWithOptions(options)
+    }
+
+    private func runTCCReset(service: String) -> CommandResult {
+        let bundleIdentifier = Bundle.main.bundleIdentifier ?? "lt.nijho.agent-cli.menubar"
+        let process = Process()
+        let pipe = Pipe()
+
+        process.executableURL = URL(fileURLWithPath: "/usr/bin/tccutil")
+        process.arguments = ["reset", service, bundleIdentifier]
+        process.standardOutput = pipe
+        process.standardError = pipe
+
+        do {
+            try process.run()
+            process.waitUntilExit()
+            let data = pipe.fileHandleForReading.readDataToEndOfFile()
+            let output = String(data: data, encoding: .utf8) ?? ""
+            return CommandResult(exitCode: process.terminationStatus, output: output)
+        } catch {
+            return CommandResult(exitCode: 1, output: error.localizedDescription)
+        }
     }
 
     @discardableResult