agent-cli 0.105.2__tar.gz → 0.107.0__tar.gz

{agent_cli-0.105.2 → agent_cli-0.107.0}/.github/workflows/docker.yml

@@ -49,7 +49,7 @@ jobs:
           fetch-depth: 0  # Full history for version detection
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v4.3.0
+        uses: docker/setup-buildx-action@v4.4.1
 
       - name: Log in to Container Registry
         uses: docker/login-action@v4.5.1
@@ -79,7 +79,7 @@ jobs:
             type=sha,suffix=-${{ matrix.suffix }}
 
       - name: Build and push
-        uses: docker/build-push-action@v7.3.0
+        uses: docker/build-push-action@v7.4.0
         with:
           context: .
           file: ${{ matrix.dockerfile }}
@@ -122,7 +122,7 @@ jobs:
           fetch-depth: 0  # Full history for version detection
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v4.3.0
+        uses: docker/setup-buildx-action@v4.4.1
 
       - name: Log in to Container Registry
         uses: docker/login-action@v4.5.1
@@ -152,7 +152,7 @@ jobs:
             type=sha
 
       - name: Build and push
-        uses: docker/build-push-action@v7.3.0
+        uses: docker/build-push-action@v7.4.0
         with:
           context: .
           file: ${{ matrix.dockerfile }}

{agent_cli-0.105.2 → agent_cli-0.107.0}/.github/workflows/pytest.yml

@@ -43,6 +43,6 @@ jobs:
         run: uv run --extra audio --extra diarization --extra llm --extra rag --extra memory --extra vad --extra faster-whisper --extra piper --extra kokoro --extra server --extra speed --extra test pytest -vvv
       - name: Upload coverage reports to Codecov
         if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.13'
-        uses: codecov/codecov-action@v7.1.0
+        uses: codecov/codecov-action@v7.1.1
         env:
           CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

{agent_cli-0.105.2 → agent_cli-0.107.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: agent-cli
-Version: 0.105.2
+Version: 0.107.0
 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-0.105.2 → agent_cli-0.107.0}/agent_cli/agents/transcribe.py

@@ -477,6 +477,9 @@ async def _async_main(  # noqa: PLR0912, PLR0915, C901
                     live_preview_config=live_preview_config,
                 )
 
+        if transcript and not transcript.strip():
+            transcript = None
+
         elapsed = time.monotonic() - start_time
 
         # Apply diarization if enabled
@@ -554,6 +557,9 @@ async def _async_main(  # noqa: PLR0912, PLR0915, C901
                 live=live,
                 context=combined_context,
             )
+            if general_cfg.clipboard and not (processed_transcript or "").strip():
+                # Keep the raw speech recoverable if cleanup overwrote it with blank text.
+                pyperclip.copy(transcript)
 
             # Log transcription if requested
             if transcription_log:
@@ -615,6 +621,21 @@ async def _async_main(  # noqa: PLR0912, PLR0915, C901
     )
 
 
+def _require_transcript(result: TranscriptResult, provider: str) -> None:
+    """Do not report successful dictation when the backend produced no text."""
+    if not (result.get("transcript") or "").strip():
+        if result.get("raw_transcript"):
+            message = "LLM cleanup returned no text. Retry with --no-llm to use the raw transcript."
+        else:
+            message = f"No transcript returned by {provider}. Check the microphone and ASR server."
+        typer.echo(
+            f"{message} "
+            "If a recording was saved, retry with agent-cli transcribe --last-recording.",
+            err=True,
+        )
+        raise typer.Exit(1)
+
+
 @app.command("transcribe", rich_help_panel="Voice Commands")
 @requires_extras("audio", "llm", process_name="transcribe")
 def transcribe(  # noqa: PLR0912, PLR0911, PLR0915, C901
@@ -896,6 +917,7 @@ def transcribe(  # noqa: PLR0912, PLR0911, PLR0915, C901
             LOGGER.exception("Diarization failed")
             _print_diarization_error(exc)
             raise typer.Exit(1) from None
+        _require_transcript(result, asr_provider)
         if json_output:
             print(json.dumps(result))
         return
@@ -937,6 +959,7 @@ def transcribe(  # noqa: PLR0912, PLR0911, PLR0915, C901
     ):
         return
 
+    result = TranscriptResult(raw_transcript=None, transcript=None, llm_enabled=False)
     # Use context manager before audio setup so --stop can target startup reliably.
     try:
         with process.pid_file_context(process_name), suppress(KeyboardInterrupt):
@@ -992,5 +1015,6 @@ def transcribe(  # noqa: PLR0912, PLR0911, PLR0915, C901
         LOGGER.exception("Diarization failed")
         _print_diarization_error(exc)
         raise typer.Exit(1) from None
+    _require_transcript(result, asr_provider)
     if json_output:
         print(json.dumps(result))

{agent_cli-0.105.2 → agent_cli-0.107.0}/macos/AgentCLI/Sources/AgentCLI/AgentCommandRunner.swift

@@ -161,13 +161,13 @@ final class AgentCommandRunner: ObservableObject {
         holdTranscriptionState = .stopping
 
         let wasRecording = recordingIndicator.isRecordingCommand(.toggleTranscription)
-        beginTranscribingActivity()
         if wasRecording {
             endRecordingIndicator(for: .toggleTranscription)
             statusMessage = "Transcribing..."
         } else {
             statusMessage = "Stopping transcription as soon as it starts..."
         }
+        beginTranscribingActivity()
         stopHeldTranscriptionWhenReady()
     }
 
@@ -252,7 +252,12 @@ final class AgentCommandRunner: ObservableObject {
                 }
             }
 
-            let result = AgentRuntime.shared.runAgentCLI(arguments: commandArguments)
+            let commandResult = AgentRuntime.shared.runAgentCLI(arguments: commandArguments)
+            // A stop acknowledgement may be empty; a completed recording must contain text.
+            // Validate here too because user-installed CLIs can predate the CLI-side check.
+            let result = shouldStartRecording && command.identifier == AgentCommand.toggleTranscription.identifier
+                ? commandResult.requiringTranscript()
+                : commandResult
             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)
@@ -381,11 +386,13 @@ final class AgentCommandRunner: ObservableObject {
 
     private func beginTranscribingActivity() {
         activityTracker.beginTranscribing()
+        VoiceLevelOverlayController.shared.showTranscribing()
     }
 
     private func clearTranscribingActivityIfFinished() {
         if pendingStopRecordingCommands.isEmpty && !holdTranscriptionState.isFinishing {
             activityTracker.finishTranscribing()
+            VoiceLevelOverlayController.shared.finishTranscribing()
         }
     }
 

agent_cli-0.107.0/macos/AgentCLI/Sources/AgentCLI/CommandResult.swift

@@ -0,0 +1,17 @@
+import Foundation
+
+struct CommandResult {
+    let exitCode: Int32
+    let output: String
+
+    func requiringTranscript() -> CommandResult {
+        guard exitCode == 0, output.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty else {
+            return self
+        }
+        return CommandResult(
+            exitCode: 1,
+            output: "No transcript returned. Check the microphone and ASR server, then try again. "
+                + "If a recording was saved, recover it with agent-cli transcribe --last-recording."
+        )
+    }
+}

{agent_cli-0.105.2 → agent_cli-0.107.0}/macos/AgentCLI/Sources/AgentCLI/RecordingIndicatorController.swift

@@ -38,6 +38,7 @@ final class RecordingIndicatorController {
     }
 
     func end(for command: AgentCommand) {
+        guard isRecordingCommand(command) else { return }
         let wasRecording = isRecording
         let activeCommandCount = max(0, activeRecordingCommands[command.identifier, default: 0] - 1)
         if activeCommandCount > 0 {
@@ -48,7 +49,7 @@ final class RecordingIndicatorController {
         recordingCommandCount = max(0, recordingCommandCount - 1)
         if wasRecording && !isRecording {
             play(.finishedRecording)
-            VoiceLevelOverlayController.shared.hide()
+            VoiceLevelOverlayController.shared.endRecording()
         }
         if command.supportsLivePreviewOverlay {
             LiveTranscriptionPreview.shared.stop()

{agent_cli-0.105.2 → agent_cli-0.107.0}/macos/AgentCLI/Sources/AgentCLI/VoiceLevelOverlay.swift

@@ -3,7 +3,7 @@ import Foundation
 import SwiftUI
 
 enum VoiceLevelOverlayLayout {
-    static let pillSize = CGSize(width: 147, height: 38)
+    static let pillSize = CGSize(width: 190, height: 38)
     static let textWidth = CGFloat(420)
     static let textHeight = CGFloat(86)
     static let shadowRadius = CGFloat(13)
@@ -72,41 +72,69 @@ struct VoiceLevelOverlayView: View {
             height: VoiceLevelOverlayLayout.panelSize(showsPreviewSpace: showsPreviewSpace).height,
             alignment: .bottom
         )
-        .accessibilityLabel(Text("Voice level"))
+        .accessibilityElement(children: .ignore)
+        .accessibilityLabel(Text(meter.captureState == .recording
+            ? "Recording \(meter.recordingDuration)"
+            : meter.captureState == .transcribing
+                ? "Transcribing \(meter.transcribingDuration)"
+                : meter.captureState.label))
+        .accessibilityValue(Text(showsPreviewSpace ? preview.text : ""))
     }
 
     private var levelMeter: some View {
-        HStack(alignment: .center, spacing: 3.5) {
+        HStack(spacing: 8) {
+            if meter.captureState == .recording {
+                Text(meter.recordingDuration)
+                    .font(.system(size: 12, weight: .semibold))
+                    .monospacedDigit()
+                    .fixedSize()
+                waveform
+            } else {
+                ProgressView()
+                    .controlSize(.small)
+                Text(meter.captureState.label)
+                    .font(.system(size: 12, weight: .semibold))
+                if meter.captureState == .transcribing {
+                    Text(meter.transcribingDuration)
+                        .font(.system(size: 12, weight: .semibold))
+                        .monospacedDigit()
+                        .fixedSize()
+                }
+            }
+        }
+        .foregroundStyle(statusColor)
+        .padding(.horizontal, 14)
+        .frame(width: VoiceLevelOverlayLayout.pillSize.width, height: VoiceLevelOverlayLayout.pillSize.height)
+        .background(Capsule().fill(statusColor.opacity(isLightMode ? 0.10 : 0.14)))
+        .background(Capsule().fill(backgroundColor))
+        .overlay(Capsule().stroke(statusColor.opacity(0.55), lineWidth: 1))
+        .shadow(
+            color: shadowColor,
+            radius: VoiceLevelOverlayLayout.shadowRadius,
+            y: VoiceLevelOverlayLayout.shadowYOffset
+        )
+    }
+
+    private var waveform: some View {
+        HStack(alignment: .center, spacing: 2) {
             ForEach(Array(meter.amplitudes.enumerated()), id: \.offset) { _, amplitude in
                 Capsule()
                     .fill(
                         LinearGradient(
                             colors: [
-                                barGradientStart,
-                                barGradientEnd
+                                statusColor.opacity(0.65),
+                                statusColor
                             ],
                             startPoint: .bottom,
                             endPoint: .top
                         )
                     )
-                    .frame(width: 3.5, height: max(5, 25 * amplitude))
+                    .frame(maxWidth: .infinity)
+                    .frame(height: max(6, 30 * amplitude))
                     .animation(.easeOut(duration: 0.11), value: amplitude)
             }
         }
-        .frame(width: VoiceLevelOverlayLayout.pillSize.width, height: VoiceLevelOverlayLayout.pillSize.height)
-        .background(
-            Capsule()
-                .fill(backgroundColor)
-        )
-        .overlay(
-            Capsule()
-                .stroke(borderColor, lineWidth: 1)
-        )
-        .shadow(
-            color: shadowColor,
-            radius: VoiceLevelOverlayLayout.shadowRadius,
-            y: VoiceLevelOverlayLayout.shadowYOffset
-        )
+        .frame(maxWidth: .infinity)
     }
 
     private var isLightMode: Bool {
@@ -133,12 +161,20 @@ struct VoiceLevelOverlayView: View {
         Color.black.opacity(isLightMode ? 0.16 : 0.24)
     }
 
-    private var barGradientStart: Color {
-        isLightMode ? Color(red: 0.04, green: 0.45, blue: 0.95) : Color(red: 0.18, green: 0.82, blue: 0.92)
-    }
-
-    private var barGradientEnd: Color {
-        isLightMode ? Color(red: 0.07, green: 0.72, blue: 0.68) : Color(red: 0.64, green: 0.96, blue: 0.58)
+    private var statusColor: Color {
+        if meter.captureState == .transcribing {
+            return isLightMode
+                ? Color(red: 0.10, green: 0.32, blue: 0.70)
+                : Color(red: 0.45, green: 0.72, blue: 1.0)
+        }
+        if meter.captureState == .recording {
+            return isLightMode
+                ? Color(red: 0.04, green: 0.40, blue: 0.18)
+                : Color(red: 0.42, green: 0.91, blue: 0.58)
+        }
+        return isLightMode
+            ? Color(red: 0.62, green: 0.34, blue: 0.0)
+            : Color(red: 1.0, green: 0.74, blue: 0.25)
     }
 }
 
@@ -147,27 +183,63 @@ final class VoiceLevelOverlayController {
 
     private var panel: NSPanel?
     private var showsPreviewSpace = false
+    private var recordingShowsPreviewSpace = false
+    private var isRecording = false
+    private var isTranscribing = false
 
     private init() {}
 
     func show(showsPreviewSpace: Bool = false) {
+        isRecording = true
+        recordingShowsPreviewSpace = showsPreviewSpace
         let panel = panel ?? makePanel()
         self.panel = panel
-        if self.showsPreviewSpace != showsPreviewSpace {
-            self.showsPreviewSpace = showsPreviewSpace
-            updatePanelContent(panel)
-        }
-        panel.setContentSize(VoiceLevelOverlayLayout.panelSize(showsPreviewSpace: showsPreviewSpace))
-        position(panel)
+        setPreviewSpace(showsPreviewSpace, for: panel)
         VoiceLevelMeter.shared.start()
         panel.orderFrontRegardless()
     }
 
     func hide() {
+        isRecording = false
+        isTranscribing = false
         VoiceLevelMeter.shared.stop()
         panel?.orderOut(nil)
     }
 
+    func showTranscribing() {
+        isTranscribing = true
+        let panel = panel ?? makePanel()
+        self.panel = panel
+        VoiceLevelMeter.shared.beginTranscribing()
+        setPreviewSpace(false, for: panel)
+        panel.orderFrontRegardless()
+    }
+
+    func finishTranscribing() {
+        guard isTranscribing else { return }
+        isTranscribing = false
+        if isRecording {
+            if let panel {
+                setPreviewSpace(recordingShowsPreviewSpace, for: panel)
+            }
+            VoiceLevelMeter.shared.resumeRecording()
+        } else {
+            hide()
+        }
+    }
+
+    func endRecording() {
+        isRecording = false
+        if isTranscribing {
+            VoiceLevelMeter.shared.beginTranscribing()
+            if let panel {
+                setPreviewSpace(false, for: panel)
+            }
+        } else {
+            hide()
+        }
+    }
+
     private func makePanel() -> NSPanel {
         let panel = NSPanel(
             contentRect: NSRect(
@@ -189,6 +261,15 @@ final class VoiceLevelOverlayController {
         return panel
     }
 
+    private func setPreviewSpace(_ showsPreviewSpace: Bool, for panel: NSPanel) {
+        if self.showsPreviewSpace != showsPreviewSpace {
+            self.showsPreviewSpace = showsPreviewSpace
+            updatePanelContent(panel)
+        }
+        panel.setContentSize(VoiceLevelOverlayLayout.panelSize(showsPreviewSpace: showsPreviewSpace))
+        position(panel)
+    }
+
     private func updatePanelContent(_ panel: NSPanel) {
         panel.contentView = NSHostingView(rootView: VoiceLevelOverlayView(
             meter: VoiceLevelMeter.shared,
@@ -306,10 +387,31 @@ enum VoiceLevelLog {
     }()
 }
 
+enum VoiceCaptureState {
+    case idle
+    case connecting
+    case recording
+    case waitingForAudio
+    case transcribing
+
+    var label: String {
+        switch self {
+        case .idle: return "Ready"
+        case .connecting: return "Connecting…"
+        case .recording: return "Recording"
+        case .waitingForAudio: return "Waiting for audio…"
+        case .transcribing: return "Transcribing…"
+        }
+    }
+}
+
 final class VoiceLevelMeter: ObservableObject {
     static let shared = VoiceLevelMeter()
 
     @Published private(set) var amplitudes = VoiceLevelMeter.idleAmplitudes
+    @Published private(set) var captureState = VoiceCaptureState.idle
+    @Published private(set) var recordingDuration = "00:00"
+    @Published private(set) var transcribingDuration = "00:00"
 
     private static let barCount = 16
     private static let idleAmplitudes = Array(repeating: CGFloat(0.16), count: barCount)
@@ -319,6 +421,8 @@ final class VoiceLevelMeter: ObservableObject {
     private let levelLogURL: URL
     private let now: () -> Date
     private var timer: Timer?
+    private var recordingStartedAt: Date?
+    private var transcribingStartedAt: Date?
     private var phase = 0.0
     private var smoothedLevel = CGFloat(0.16)
 
@@ -328,12 +432,21 @@ final class VoiceLevelMeter: ObservableObject {
     }
 
     func start() {
-        guard timer == nil else { return }
+        guard timer == nil || captureState == .transcribing else { return }
         VoiceLevelLog.reset(levelLogURL)
+        captureState = .connecting
+        recordingStartedAt = nil
+        recordingDuration = "00:00"
+        transcribingStartedAt = nil
+        transcribingDuration = "00:00"
         phase = 0
         smoothedLevel = Self.idleLevel
         amplitudes = Self.idleAmplitudes
+        startPolling()
+    }
 
+    private func startPolling() {
+        timer?.invalidate()
         let timer = Timer(timeInterval: Self.pollInterval, repeats: true) { [weak self] _ in
             self?.pollLevel()
         }
@@ -342,16 +455,55 @@ final class VoiceLevelMeter: ObservableObject {
         pollLevel()
     }
 
+    func resumeRecording() {
+        guard captureState == .transcribing else { return }
+        transcribingStartedAt = nil
+        captureState = recordingStartedAt == nil ? .connecting : .waitingForAudio
+        startPolling()
+    }
+
     func stop() {
         timer?.invalidate()
         timer = nil
+        captureState = .idle
+        transcribingStartedAt = nil
+        transcribingDuration = "00:00"
+        recordingStartedAt = nil
+        recordingDuration = "00:00"
         phase = 0
         smoothedLevel = Self.idleLevel
         amplitudes = Self.idleAmplitudes
     }
 
-    private func pollLevel() {
-        let level = VoiceLevelLog.latestLevel(from: levelLogURL, now: now()) ?? Self.idleLevel
+    func beginTranscribing() {
+        guard captureState != .transcribing else { return }
+        transcribingStartedAt = now()
+        transcribingDuration = "00:00"
+        captureState = .transcribing
+        amplitudes = Self.idleAmplitudes
+        startPolling()
+    }
+
+    func pollLevel() {
+        guard timer != nil else { return }
+        let now = now()
+        if captureState == .transcribing {
+            let elapsed = max(0, Int(now.timeIntervalSince(transcribingStartedAt ?? now)))
+            transcribingDuration = String(format: "%02d:%02d", elapsed / 60, elapsed % 60)
+            return
+        }
+        guard let level = VoiceLevelLog.latestLevel(from: levelLogURL, now: now) else {
+            if captureState == .recording {
+                captureState = .waitingForAudio
+            }
+            amplitudes = Self.idleAmplitudes
+            return
+        }
+        captureState = .recording
+        let startedAt = recordingStartedAt ?? now
+        recordingStartedAt = startedAt
+        let elapsedSeconds = max(0, Int(now.timeIntervalSince(startedAt)))
+        recordingDuration = String(format: "%02d:%02d", elapsedSeconds / 60, elapsedSeconds % 60)
         updateDisplay(level: level)
     }
 

{agent_cli-0.105.2 → agent_cli-0.107.0}/macos/AgentCLI/Tests/AgentCLITests/RecordingIndicatorControllerTests.swift

@@ -4,6 +4,19 @@ import XCTest
 @testable import AgentCLI
 
 final class RecordingIndicatorControllerTests: XCTestCase {
+    func testRepeatedEndDoesNotStopAnotherRecording() {
+        let defaults = UserDefaults(suiteName: "AgentCLITests.recording-repeated-end")!
+        defaults.removePersistentDomain(forName: "AgentCLITests.recording-repeated-end")
+        let controller = RecordingIndicatorController(defaults: defaults, audioCuePlayer: RecordingCuePlayerSpy())
+        defer { controller.end(for: .voiceEdit) }
+        controller.begin(for: .toggleTranscription)
+        controller.begin(for: .voiceEdit)
+        controller.end(for: .toggleTranscription)
+        controller.end(for: .toggleTranscription)
+        XCTAssertTrue(controller.isRecording)
+        XCTAssertTrue(controller.isRecordingCommand(.voiceEdit))
+    }
+
     func testRecordingSoundsAreDisabledByDefault() {
         let defaults = UserDefaults(suiteName: "AgentCLITests.recording-sounds-default")!
         defaults.removePersistentDomain(forName: "AgentCLITests.recording-sounds-default")

agent_cli-0.107.0/macos/AgentCLI/Tests/AgentCLITests/TranscriptionResultTests.swift

@@ -0,0 +1,23 @@
+#if canImport(XCTest)
+import XCTest
+@testable import AgentCLI
+
+final class TranscriptionResultTests: XCTestCase {
+    func testEmptySuccessfulRecordingBecomesFailure() {
+        for output in ["", " \n\t"] {
+            let result = CommandResult(exitCode: 0, output: output).requiringTranscript()
+            XCTAssertNotEqual(result.exitCode, 0)
+            XCTAssertTrue(result.output.contains("No transcript"))
+        }
+    }
+
+    func testTranscriptAndBackendFailureArePreserved() {
+        let success = CommandResult(exitCode: 0, output: "Hello world").requiringTranscript()
+        XCTAssertEqual(success.exitCode, 0)
+        XCTAssertEqual(success.output, "Hello world")
+        let failure = CommandResult(exitCode: 7, output: "Connection lost").requiringTranscript()
+        XCTAssertEqual(failure.exitCode, 7)
+        XCTAssertEqual(failure.output, "Connection lost")
+    }
+}
+#endif