Compare commits

...
32 Commits
Author SHA1 Message Date
coder 83710bc1fe Document one shared runner per host [skip ci] 2026-09-10 09:05:04 +02:00
coder 7584863392 fix: isolate CI Keychain access from the production app
Build / macos (push) Successful in 27s
2026-09-10 00:33:30 +02:00
coder 31584792e6 fix: launch CI smoke app through macOS Launch Services
Build / macos (push) Failing after 1m3s
2026-09-10 00:25:09 +02:00
coder 58fbe56298 ci: retire the superseded GitHub release workflows
Build / macos (push) Failing after 1m22s
2026-09-10 00:08:31 +02:00
coder 4d87288f2f ci: stage the Mac launch check on the internal volume
Build / macos (push) Successful in 26s
2026-09-09 23:56:02 +02:00
coder cda28cc6be ci: preserve evidence when the Mac smoke launch stalls
Build / macos (push) Failing after 1m1s
2026-09-09 23:51:29 +02:00
coder df23c84474 ci: release notarized Meetingnotes updates through Gitea
Build / macos (push) Failing after 57s
2026-09-09 23:49:37 +02:00
coder 2508c77f17 ci: build and smoke test main on the Gitea Mac runner
Build / macos (push) Failing after 1m11s
2026-09-09 23:45:46 +02:00
coder 607e1236f7 fix: allow recording while meetings process 2026-08-27 16:05:23 +02:00
coder 204857cdd9 fix: infer actual system audio clock rate 2026-08-27 13:02:00 +02:00
coder 3a9ad8fe0c fix: preserve diagnostics when merging meetings 2026-08-26 17:04:15 +02:00
coder 299986ab00 fix: preserve transcription failures and audio diagnostics 2026-08-26 16:43:43 +02:00
coder 66447b2510 fix: count Core Audio callback frames explicitly 2026-08-26 14:36:32 +02:00
coder 9f3805733b fix: capture system audio as mono 2026-08-26 11:56:27 +02:00
coder 4e8fdc7603 fix: use delivered system audio format 2026-08-26 10:22:59 +02:00
coder aed34f6d28 fix: debounce automatic meeting stops 2026-08-26 10:13:39 +02:00
coder 79b2f61dfe feat: merge split meeting continuations 2026-08-26 10:09:21 +02:00
coder f6f518b0a9 fix: preserve recordings before transcription 2026-08-25 16:51:00 +02:00
coder 58b673ffcc fix: preserve system audio timing 2026-08-25 16:00:36 +02:00
coder 6cb2ae649e fix: record transcription audio as wav 2026-08-25 12:35:54 +02:00
coder 8b96d5d446 fix: persist transcription model selection 2026-08-11 22:28:35 +02:00
coder 19968160e0 fix: notarize Meetingnotes releases before publishing 2026-08-11 21:49:20 +02:00
coder 3d093f1e1d feat: add Parakeet speaker diarization 2026-08-11 21:45:51 +02:00
coder 579a91d357 fix: preserve system audio duration 2026-07-29 16:06:44 +02:00
james 19b17fc9a5 fix: transcribe long recordings in chunks 2026-07-23 14:48:25 +02:00
james 1268149114 fix: preserve system audio resampler state 2026-07-23 14:29:04 +02:00
james 0f45f2d066 feat: make audio retention configurable 2026-07-22 16:31:01 +02:00
james f1d57da227 feat: retain meeting audio for three days 2026-07-22 16:24:08 +02:00
james 94d8469eba fix: capture all system audio 2026-07-22 16:14:33 +02:00
james 628587ae92 feat: retry failed transcriptions 2026-07-20 13:13:15 +02:00
james f992457cfe feat: add menu bar mode 2026-07-17 12:46:34 +02:00
james ad995b59ad feat: generate meeting titles automatically 2026-07-17 12:15:04 +02:00
31 changed files with 2163 additions and 576 deletions
@@ -9,7 +9,7 @@ on:
jobs:
macos:
runs-on: macos-15
runs-on: macos-arm64
steps:
- uses: actions/checkout@v4
- name: Build Meetingnotes
@@ -22,6 +22,7 @@ jobs:
-derivedDataPath "$RUNNER_TEMP/DerivedData"
ARCHS="arm64 x86_64"
ONLY_ACTIVE_ARCH=NO
PRODUCT_BUNDLE_IDENTIFIER=net.jamesbone.meetingnotes.ci
CODE_SIGNING_ALLOWED=NO
build
- name: Sign test build
@@ -34,28 +35,15 @@ jobs:
codesign -d --entitlements :- "$app_path" 2>&1 \
| grep -q 'com.apple.security.network.server'
- name: Smoke test local API
run: |
defaults write net.jamesbone.meetingnotes muteDeckAPIEnabled -bool true
defaults write net.jamesbone.meetingnotes muteDeckAPIPort -int 19880
"$RUNNER_TEMP/DerivedData/Build/Products/Release/Meetingnotes.app/Contents/MacOS/Meetingnotes" >"$RUNNER_TEMP/meetingnotes.log" 2>&1 &
app_pid=$!
trap 'kill "$app_pid" 2>/dev/null || true' EXIT
for _ in {1..20}; do
if curl -fsS http://127.0.0.1:19880/api/info >"$RUNNER_TEMP/api-info.json"; then
break
fi
sleep 1
done
grep -q '"name":"MeetingDebrief"' "$RUNNER_TEMP/api-info.json"
test "$(curl -sS -o /dev/null -w '%{http_code}' http://127.0.0.1:19880/api/recording/status)" = "401"
run: python3 scripts/smoke_test_api.py "$RUNNER_TEMP/DerivedData/Build/Products/Release/Meetingnotes.app"
- name: Package test build
if: always()
run: |
app_path="$RUNNER_TEMP/DerivedData/Build/Products/Release/Meetingnotes.app"
ditto -c -k --sequesterRsrc --keepParent \
"$app_path" "$RUNNER_TEMP/Meetingnotes-macOS.zip"
- name: Upload test build
if: always()
uses: actions/upload-artifact@v4
with:
name: Meetingnotes-macOS-${{ github.sha }}
+47
View File
@@ -0,0 +1,47 @@
name: Release
on:
workflow_dispatch:
inputs:
version:
description: Version from MARKETING_VERSION, without the v prefix
required: true
type: string
permissions:
contents: write
concurrency:
group: meetingnotes-release
cancel-in-progress: false
jobs:
release:
runs-on: macos-arm64
env:
VERSION: ${{ inputs.version }}
RELEASE_BASE_URL: https://git.jamesbone.net/coder/meetingnotes
APPLE_ID: ${{ secrets.APPLE_ID }}
APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
APPLE_APP_PASSWORD: ${{ secrets.APPLE_APP_PASSWORD }}
SPARKLE_PRIVATE_KEY: ${{ secrets.SPARKLE_PRIVATE_KEY }}
steps:
- name: Require main
run: test "$GITHUB_REF" = refs/heads/main
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Build, sign, and notarize release
timeout-minutes: 45
env:
APPLE_CERTIFICATE_P12: ${{ secrets.APPLE_CERTIFICATE_P12 }}
APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
run: python3 scripts/with_signing_identity.py scripts/package_release.sh
- name: Preserve signed release artifacts
uses: actions/upload-artifact@v4
with:
name: meetingnotes-signed-release-${{ github.run_id }}
path: ${{ runner.temp }}/meetingnotes-release/release
if-no-files-found: error
retention-days: 30
- name: Publish Gitea release
env:
GITEA_SERVER_URL: ${{ github.server_url }}
GITEA_TOKEN: ${{ github.token }}
run: python3 scripts/publish_gitea_release.py
-92
View File
@@ -1,92 +0,0 @@
name: Finalize Notarization (manual)
on:
workflow_dispatch:
inputs:
release_run_id:
description: Release workflow run ID; leave blank to use the latest pending run
required: false
type: string
permissions:
actions: read
contents: write
concurrency:
group: meetingnotes-finalize-release
cancel-in-progress: false
jobs:
finalize:
if: github.event_name != 'workflow_run' || github.event.workflow_run.conclusion == 'success'
runs-on: macos-15
env:
APPLE_ID: ${{ secrets.APPLE_ID }}
APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
APPLE_APP_PASSWORD: ${{ secrets.APPLE_APP_PASSWORD }}
SPARKLE_PRIVATE_KEY: ${{ secrets.SPARKLE_PRIVATE_KEY }}
GH_TOKEN: ${{ github.token }}
steps:
- uses: actions/checkout@v7
- name: Validate release secrets
run: |
for variable in APPLE_ID APPLE_TEAM_ID APPLE_APP_PASSWORD SPARKLE_PRIVATE_KEY; do
if [[ -z "${!variable:-}" ]]; then
echo "Missing GitHub Actions secret: $variable" >&2
exit 1
fi
done
- name: Download pending signed build
id: submission
env:
MANUAL_RUN_ID: ${{ inputs.release_run_id }}
COMPLETED_RUN_ID: ${{ github.event.workflow_run.id }}
run: |
PENDING_DIR="$RUNNER_TEMP/meetingnotes-pending"
candidate_ids=()
if [[ -n "${MANUAL_RUN_ID:-}" ]]; then
candidate_ids+=("$MANUAL_RUN_ID")
elif [[ -n "${COMPLETED_RUN_ID:-}" ]]; then
candidate_ids+=("$COMPLETED_RUN_ID")
else
while IFS= read -r run_id; do
candidate_ids+=("$run_id")
done < <(gh run list --repo "$GITHUB_REPOSITORY" --workflow Release --status success --limit 20 --json databaseId --jq '.[].databaseId')
fi
for run_id in "${candidate_ids[@]}"; do
if [[ ! "$run_id" =~ ^[0-9]+$ ]]; then
echo "Invalid release run ID: $run_id" >&2
exit 1
fi
artifact_name="meetingnotes-notarization-$run_id"
artifact_count=$(gh api "repos/$GITHUB_REPOSITORY/actions/runs/$run_id/artifacts" \
--jq "[.artifacts[] | select(.name == \"$artifact_name\" and .expired == false)] | length")
if [[ "$artifact_count" == 0 ]]; then
continue
fi
rm -rf "$PENDING_DIR"
mkdir -p "$PENDING_DIR"
gh run download "$run_id" --repo "$GITHUB_REPOSITORY" --name "$artifact_name" --dir "$PENDING_DIR"
version=$(<"$PENDING_DIR/version")
if gh release view "v$version" --repo "$GITHUB_REPOSITORY" >/dev/null 2>&1; then
continue
fi
echo "found=true" >> "$GITHUB_OUTPUT"
echo "run_id=$run_id" >> "$GITHUB_OUTPUT"
echo "Using release submission from workflow run $run_id"
exit 0
done
echo "found=false" >> "$GITHUB_OUTPUT"
echo "No pending release submission was found"
- name: Check notarization and publish when accepted
if: steps.submission.outputs.found == 'true'
run: scripts/finalize_release.sh
-82
View File
@@ -1,82 +0,0 @@
name: Release
on:
workflow_dispatch:
inputs:
version:
description: Version from MARKETING_VERSION, without the v prefix
required: true
type: string
permissions:
contents: write
concurrency:
group: meetingnotes-release
cancel-in-progress: false
jobs:
release:
runs-on: macos-15
env:
VERSION: ${{ inputs.version }}
SPARKLE_PRIVATE_KEY: ${{ secrets.SPARKLE_PRIVATE_KEY }}
steps:
- uses: actions/checkout@v7
with:
fetch-depth: 0
- name: Validate release secrets
env:
APPLE_CERTIFICATE_P12: ${{ secrets.APPLE_CERTIFICATE_P12 }}
APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
run: |
for variable in APPLE_CERTIFICATE_P12 APPLE_CERTIFICATE_PASSWORD SPARKLE_PRIVATE_KEY; do
if [[ -z "${!variable:-}" ]]; then
echo "Missing GitHub Actions secret: $variable" >&2
exit 1
fi
done
- name: Import Developer ID certificate
uses: apple-actions/import-codesign-certs@v7
with:
p12-file-base64: ${{ secrets.APPLE_CERTIFICATE_P12 }}
p12-password: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
- name: Locate Developer ID identity
run: |
signing_identity=$(security find-identity -v -p codesigning | awk -F '"' '/Developer ID Application/{print $2; exit}')
if [[ -z "$signing_identity" ]]; then
echo "The .p12 does not contain a Developer ID Application identity" >&2
exit 1
fi
echo "SIGNING_IDENTITY=$signing_identity" >> "$GITHUB_ENV"
- name: Build and sign release
timeout-minutes: 30
run: scripts/package_release.sh
- name: Preserve signed release artifacts
uses: actions/upload-artifact@v4
with:
name: meetingnotes-signed-release-${{ github.run_id }}
path: ${{ runner.temp }}/meetingnotes-release/release
if-no-files-found: error
retention-days: 30
- name: Publish signed GitHub release
env:
GH_TOKEN: ${{ github.token }}
run: |
tag="v$VERSION"
if git rev-parse "$tag" >/dev/null 2>&1; then
echo "Tag already exists: $tag" >&2
exit 1
fi
gh release create "$tag" \
"$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
"$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
--target "$GITHUB_SHA" \
--title "Meetingnotes $VERSION" \
--generate-notes
+4 -4
View File
@@ -276,7 +276,7 @@
CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 21;
CURRENT_PROJECT_VERSION = 45;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES;
@@ -290,7 +290,7 @@
"@executable_path/../Frameworks",
);
MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.9;
MARKETING_VERSION = 1.1.33;
ONLY_ACTIVE_ARCH = NO;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
@@ -312,7 +312,7 @@
CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 21;
CURRENT_PROJECT_VERSION = 45;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES;
@@ -326,7 +326,7 @@
"@executable_path/../Frameworks",
);
MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.9;
MARKETING_VERSION = 1.1.33;
ONLY_ACTIVE_ARCH = YES;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
+22 -4
View File
@@ -22,7 +22,8 @@ Implemented:
- Meeting search functionality
- Abilty to edit system prompt
- Select any compatible Coder model for transcription and note generation
- Automatic start and stop from MuteDeck through a compatible local API
- Automatic start and stop from MuteDeck through a compatible local API, with a 10-second reconnect grace period
- Merge an automatically split continuation back into its previous meeting
- Auto updates
- Text formatting
- Different note templates
@@ -45,7 +46,7 @@ Later:
## Releasing a New Version
Production releases are Developer ID signed, notarized by Apple, published to
GitHub Releases, and signed for Sparkle auto-updates.
[Gitea Releases](https://git.jamesbone.net/coder/meetingnotes/releases), and signed for Sparkle auto-updates.
### Release Process
@@ -67,12 +68,12 @@ GitHub Releases, and signed for Sparkle auto-updates.
2. Commit and push the version change to `main`.
3. Run the `Release` workflow from GitHub Actions and enter the version without
3. Run the `Release` workflow from Gitea Actions on `main` and enter the version without
the `v` prefix. The workflow signs and notarizes the app, generates the
signed appcast, creates the version tag, and publishes both release assets.
The app checks
`https://github.com/superdooper86/meetingnotes/releases/latest/download/appcast.xml`
`https://git.jamesbone.net/coder/meetingnotes/releases/download/latest/appcast.xml`
and installs later releases automatically through Sparkle.
### Recovering Meetings
@@ -81,3 +82,20 @@ The first Developer ID signed build may not automatically inherit data from an
older ad-hoc signed build. In Settings, use **Import Meetings...** and select the
old `Meetings` folder. After this one-time transition, the stable signing
identity keeps the same sandbox container across updates.
### Build runner and GitHub transition
`.gitea/workflows/build.yml` builds universal macOS artifacts for pushes and pull
requests to `main`. The shared account-scoped `mac-mini` runner uses the
`macos-arm64` label. Smoke tests use a separate CI bundle identifier and temporary
launch preferences and an internal-volume staging directory. Keychain services
follow the bundle identifier, so CI never reads production credentials; the smoke
test removes its isolated API token before and after each launch. Release jobs import the original Developer ID certificate
into a temporary keychain and keep the original Sparkle signing key in Gitea
Actions secrets. They never publish from a development branch.
Version 1.1.33 moves the embedded update feed to Gitea. After its notarized archive
is verified, replace the `appcast.xml` asset on the last GitHub release with the
new appcast. Existing installations discover the Gitea download through that old
GitHub feed; after installing it, they check Gitea directly. Keep the old GitHub
repository and its release appcast available for installations that update later.
+3 -1
View File
@@ -2,6 +2,8 @@
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>LSUIElement</key>
<true/>
<key>NSAppTransportSecurity</key>
<dict>
<key>NSAllowsLocalNetworking</key>
@@ -12,7 +14,7 @@
<key>NSMicrophoneUsageDescription</key>
<string>Meetingnotes needs access to your microphone for transcription.</string>
<key>SUFeedURL</key>
<string>https://github.com/superdooper86/meetingnotes/releases/latest/download/appcast.xml</string>
<string>https://git.jamesbone.net/coder/meetingnotes/releases/download/latest/appcast.xml</string>
<key>SUPublicEDKey</key>
<string>9ZuN9G9ERB3Qoyyd/4FsF+6LMUv5jzAGP26OXAHBiW0=</string>
<key>SUEnableAutomaticChecks</key>
+612 -102
View File
@@ -3,6 +3,44 @@ import Combine
import Foundation
import SwiftUI
private enum RecoveryTranscriptionError: LocalizedError {
case noAudioFiles
case noSpeech
case requestFailed(String)
var errorDescription: String? {
switch self {
case .noAudioFiles:
return "The saved recovery audio could not be found."
case .noSpeech:
return "No speech was detected in the saved recovery audio."
case .requestFailed(let details):
return "Retry transcription failed for \(details)"
}
}
}
private struct SystemCaptureDiagnostics {
var tapAdvertisedFormat = "unavailable"
var aggregateInputFormat = "unavailable"
var selectedInputFormat = "unavailable"
var targetFormat = "unavailable"
var selectedInputSampleRate: Double?
var targetSampleRate: Double?
var inputRateInference = "unavailable"
var firstBufferLayout: String?
var callbackCount: UInt64 = 0
var inputFrameCount: UInt64 = 0
var outputFrameCount: UInt64 = 0
var discardedCallbackCount: UInt64 = 0
var firstSampleTime: Double?
var lastSampleTime: Double?
var firstHostTime: UInt64?
var lastHostTime: UInt64?
var firstCallbackAt: Date?
var lastCallbackAt: Date?
}
/// Captures microphone and system audio locally, then sends completed files to Coder.
@MainActor
final class AudioManager: NSObject, ObservableObject {
@@ -14,50 +52,44 @@ final class AudioManager: NSObject, ObservableObject {
@Published var errorMessage: String?
@Published var micAudioLevel: Float = 0
@Published var systemAudioLevel: Float = 0
private(set) var lastRecoveryAudioFolderName: String?
private var audioEngine = AVAudioEngine()
private var sessionID = UUID()
private var meetingID = UUID()
private var processTap: ProcessTap?
private let audioProcessController = AudioProcessController()
private let permission = AudioRecordingPermission()
private let tapQueue = DispatchQueue(label: "io.meetingnotes.audiotap", qos: .userInitiated)
private let audioFileLock = NSLock()
private var isTapActive = false
private var isRestartingSystemTap = false
private var isAcceptingAudio = false
private var micRetryCount = 0
private var pendingMicRestart: DispatchWorkItem?
private let maxMicRetries = 3
private var cancellables = Set<AnyCancellable>()
private var micAudioFile: AVAudioFile?
private var systemAudioFile: AVAudioFile?
private var micAudioURL: URL?
private var systemAudioURL: URL?
private var recordingStartedAt = Date()
private var systemDiagnostics = SystemCaptureDiagnostics()
private override init() {
override init() {
super.init()
observeAudioEngine()
audioProcessController.activate()
NSWorkspace.shared.publisher(for: \.runningApplications)
.debounce(for: .seconds(1), scheduler: RunLoop.main)
.sink { [weak self] _ in
guard let self, self.isTapActive else { return }
Task { await self.restartSystemAudioTapIfNeeded() }
}
.store(in: &cancellables)
}
deinit {
NotificationCenter.default.removeObserver(self)
}
func startRecording() {
func startRecording(for meetingID: UUID) {
errorMessage = nil
lastRecoveryAudioFolderName = nil
cancelCapture(removeFiles: true)
sessionID = UUID()
self.meetingID = meetingID
recordingStartedAt = Date()
systemDiagnostics = SystemCaptureDiagnostics()
do {
try prepareAudioFiles()
startMicrophoneTap()
@@ -69,6 +101,7 @@ final class AudioManager: NSObject, ObservableObject {
}
func stopRecordingAndTranscribe() async -> [TranscriptChunk] {
let completedMeetingID = meetingID
let completedSessionID = sessionID
let captureStartedAt = recordingStartedAt
let files = stopCaptureAndCloseFiles()
@@ -77,13 +110,146 @@ final class AudioManager: NSObject, ObservableObject {
isProcessing = false
}
let preRepairFileSummaries = files.map(audioFileSummary)
let repairApplied = repairHalfDurationSystemWAVIfNeeded(in: files)
let completedFiles = files.compactMap { $0 }
let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
let transcriptionFiles = preservedAudioFiles(files, in: audioFolder)
lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
writeCaptureDiagnostics(
sessionID: completedSessionID,
captureStartedAt: captureStartedAt,
preRepairFileSummaries: preRepairFileSummaries,
repairedFiles: transcriptionFiles,
repairApplied: repairApplied,
audioFolder: audioFolder
)
if let mismatch = captureDurationMismatch(in: transcriptionFiles) {
let recoveryMessage = audioFolder == nil
? " The audio remains in the app's temporary folder."
: " Audio was kept so it can be recovered."
errorMessage = "System audio timing was invalid (\(mismatch)). Transcription was stopped to avoid an out-of-order result." + recoveryMessage
return transcriptChunks.filter(\.isFinal)
}
let model = UserDefaultsManager.shared.transcriptionModel
async let micResult = transcribe(files[0], model: model)
async let systemResult = transcribe(files[1], model: model)
async let micResult = transcribe(transcriptionFiles[0], model: model, diarization: false)
async let systemResult = transcribe(transcriptionFiles[1], model: model, diarization: true)
let (micTranscription, systemTranscription) = await (micResult, systemResult)
let results = [micTranscription, systemTranscription]
var updated = transcriptChunks.filter(\.isFinal)
let (updated, failures) = buildTranscriptChunks(
from: results,
captureStartedAt: captureStartedAt,
existingChunks: transcriptChunks.filter(\.isFinal)
)
transcriptChunks = updated
if !failures.isEmpty {
let retentionDays = UserDefaultsManager.shared.audioRetentionDays
let retentionUnit = retentionDays == 1 ? "day" : "days"
let recoveryMessage = audioFolder == nil
? " The audio remains in the app's temporary folder."
: " Audio was kept for \(retentionDays) \(retentionUnit). Use Show Audio Folder in the Meetingnotes menu to find it."
errorMessage = "Transcription failed for " + failures.joined(separator: "; ") + recoveryMessage
} else if audioFolder == nil, !completedFiles.isEmpty {
errorMessage = "The transcript completed, but Meetingnotes could not move the audio into its retention folder."
}
return updated
}
func transcribeRecoveryAudio(in folder: URL, captureStartedAt: Date) async throws -> [TranscriptChunk] {
let recoveryFiles = LocalStorageManager.shared.recoveryAudioFiles(in: folder)
guard !recoveryFiles.isEmpty else {
throw RecoveryTranscriptionError.noAudioFiles
}
isProcessing = true
defer { isProcessing = false }
let model = UserDefaultsManager.shared.transcriptionModel
let sessions = Dictionary(grouping: recoveryFiles) { recoverySessionKey(for: $0.url) }
.values
.map { files in
(files: files, startedAt: recoveryCaptureStartedAt(for: files, fallback: captureStartedAt))
}
.sorted { $0.startedAt < $1.startedAt }
var chunks: [TranscriptChunk] = []
var failures: [String] = []
for session in sessions {
let micURL = session.files.first(where: { $0.source == .mic })?.url
let systemURL = session.files.first(where: { $0.source == .system })?.url
async let micResult = transcribe(micURL, model: model, diarization: false)
async let systemResult = transcribe(systemURL, model: model, diarization: true)
let (micTranscription, systemTranscription) = await (micResult, systemResult)
let result = buildTranscriptChunks(
from: [micTranscription, systemTranscription],
captureStartedAt: session.startedAt,
existingChunks: chunks
)
chunks = result.0
failures.append(contentsOf: result.1)
}
if !failures.isEmpty {
throw RecoveryTranscriptionError.requestFailed(failures.joined(separator: "; "))
}
guard !chunks.isEmpty else {
throw RecoveryTranscriptionError.noSpeech
}
return chunks
}
private func recoverySessionKey(for url: URL) -> String {
let name = url.deletingPathExtension().lastPathComponent
if name.hasSuffix("-mic") { return String(name.dropLast(4)) }
if name.hasSuffix("-system") { return String(name.dropLast(7)) }
return name
}
private func recoveryCaptureStartedAt(
for files: [(url: URL, source: AudioSource)],
fallback: Date
) -> Date {
let estimatedStarts = files.compactMap { file -> Date? in
guard let duration = audioDuration(at: file.url),
let values = try? file.url.resourceValues(forKeys: [.contentModificationDateKey, .creationDateKey]),
let finishedAt = values.contentModificationDate ?? values.creationDate else {
return nil
}
return finishedAt.addingTimeInterval(-duration)
}
return estimatedStarts.min() ?? fallback
}
func cancelRecording() {
cancelCapture(removeFiles: true)
lastRecoveryAudioFolderName = nil
}
private func transcribe(
_ fileURL: URL?,
model: String,
diarization: Bool
) async -> Result<CoderAPIClient.Transcription, Error>? {
guard let fileURL else { return nil }
do {
return .success(try await CoderAPIClient.shared.transcribe(
fileURL: fileURL,
model: model,
diarization: diarization,
maxSpeakerCount: 4
))
} catch {
return .failure(error)
}
}
private func buildTranscriptChunks(
from results: [Result<CoderAPIClient.Transcription, Error>?],
captureStartedAt: Date,
existingChunks: [TranscriptChunk]
) -> ([TranscriptChunk], [String]) {
var updated = existingChunks
var failures: [String] = []
for (source, result) in zip([AudioSource.mic, .system], results) {
guard let result else { continue }
@@ -101,6 +267,7 @@ final class AudioManager: NSObject, ObservableObject {
updated.append(TranscriptChunk(
timestamp: captureStartedAt.addingTimeInterval(max(0, segment.start)),
source: source,
speaker: source == .system ? segment.speaker : nil,
text: text,
isFinal: true
))
@@ -114,44 +281,23 @@ final class AudioManager: NSObject, ObservableObject {
if $0.timestamp != $1.timestamp { return $0.timestamp < $1.timestamp }
return $0.source.rawValue < $1.source.rawValue
}
transcriptChunks = updated
let completedFiles = files.compactMap { $0 }
if failures.isEmpty {
removeAudioFiles(completedFiles)
} else {
let recoveryFolder = preserveAudioFiles(completedFiles, sessionID: completedSessionID)
let recoveryMessage = recoveryFolder == nil
? " The audio remains in the app's temporary folder."
: " Audio was saved in Documents/Meetingnotes-Recovery/\(completedSessionID.uuidString)."
errorMessage = "Transcription failed for " + failures.joined(separator: "; ") + recoveryMessage
}
return updated
}
func cancelRecording() {
cancelCapture(removeFiles: true)
}
private func transcribe(_ fileURL: URL?, model: String) async -> Result<CoderAPIClient.Transcription, Error>? {
guard let fileURL else { return nil }
do {
return .success(try await CoderAPIClient.shared.transcribe(fileURL: fileURL, model: model))
} catch {
return .failure(error)
}
return (updated, failures)
}
private func prepareAudioFiles() throws {
let settings: [String: Any] = [
AVFormatIDKey: kAudioFormatMPEG4AAC,
AVFormatIDKey: kAudioFormatLinearPCM,
AVSampleRateKey: 16_000,
AVNumberOfChannelsKey: 1,
AVEncoderBitRateKey: 48_000
AVLinearPCMBitDepthKey: 16,
AVLinearPCMIsFloatKey: false,
AVLinearPCMIsBigEndianKey: false,
AVLinearPCMIsNonInterleaved: false
]
let base = FileManager.default.temporaryDirectory
let id = sessionID.uuidString
let micURL = base.appendingPathComponent("meetingnotes-\(id)-mic.m4a")
let systemURL = base.appendingPathComponent("meetingnotes-\(id)-system.m4a")
let micURL = base.appendingPathComponent("meetingnotes-\(id)-mic.wav")
let systemURL = base.appendingPathComponent("meetingnotes-\(id)-system.wav")
let newMicAudioFile = try AVAudioFile(
forWriting: micURL,
settings: settings,
@@ -239,8 +385,7 @@ final class AudioManager: NSObject, ObservableObject {
return
}
let processIDs = audioProcessController.processes.map(\.objectID)
let newTap = ProcessTap(target: .systemAudio(processObjectIDs: processIDs))
let newTap = ProcessTap(target: .systemAudio)
newTap.activate()
if let tapError = newTap.errorMessage {
errorMessage = "Failed to activate system audio capture: \(tapError)"
@@ -264,21 +409,8 @@ final class AudioManager: NSObject, ObservableObject {
}
}
private func restartSystemAudioTapIfNeeded() async {
let next = Set(audioProcessController.processes.map(\.objectID))
let current: Set<AudioObjectID>
if case .systemAudio(let processIDs) = processTap?.target {
current = Set(processIDs)
} else {
current = []
}
if next != current { await restartSystemAudioTap() }
}
private func restartSystemAudioTap() async {
guard isRecording else { return }
isRestartingSystemTap = true
defer { isRestartingSystemTap = false }
if isTapActive {
processTap?.invalidate()
processTap = nil
@@ -301,43 +433,209 @@ final class AudioManager: NSObject, ObservableObject {
private func startTapIO(_ tap: ProcessTap) throws {
guard var description = tap.tapStreamDescription,
let inputFormat = AVAudioFormat(streamDescription: &description),
let targetFormat = systemAudioFile?.processingFormat else {
let advertisedInputFormat = AVAudioFormat(streamDescription: &description),
let targetFormat = systemAudioFile?.processingFormat,
advertisedInputFormat.sampleRate > 0 else {
throw NSError(domain: "AudioManager", code: -1, userInfo: [NSLocalizedDescriptionKey: "Unsupported system audio format"])
}
try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in
guard let self,
let converter = AVAudioConverter(from: inputFormat, to: targetFormat) else { return }
systemDiagnostics.tapAdvertisedFormat = streamDescriptionSummary(tap.tapAdvertisedStreamDescription)
systemDiagnostics.aggregateInputFormat = streamDescriptionSummary(tap.aggregateInputStreamDescription)
systemDiagnostics.targetFormat = audioFormatSummary(targetFormat)
systemDiagnostics.targetSampleRate = targetFormat.sampleRate
var inputFormat: AVAudioFormat?
var converter: AVAudioConverter?
var pendingBuffer: AVAudioPCMBuffer?
var pendingInputTime: AudioTimeStamp?
var pendingNow: AudioTimeStamp?
try tap.run(on: tapQueue) { [weak self] inNow, inputData, inInputTime, _, _ in
guard let self else { return }
if inputFormat == nil {
guard let callbackFormat = self.inputFormat(
for: inputData,
sampleRate: advertisedInputFormat.sampleRate
),
let currentBuffer = self.copyAudioBuffer(from: inputData, format: callbackFormat) else {
self.systemDiagnostics.discardedCallbackCount += 1
return
}
guard let previousBuffer = pendingBuffer,
let previousInputTime = pendingInputTime,
let previousNow = pendingNow else {
pendingBuffer = currentBuffer
pendingInputTime = inInputTime.pointee
pendingNow = inNow.pointee
return
}
let selectedSampleRate = self.effectiveInputSampleRate(
advertisedSampleRate: advertisedInputFormat.sampleRate,
previousFrameLength: previousBuffer.frameLength,
previousTimestamp: previousInputTime,
currentTimestamp: inInputTime.pointee,
previousHostTimestamp: previousNow,
currentHostTimestamp: inNow.pointee
)
inputFormat = self.inputFormat(
for: inputData,
sampleRate: selectedSampleRate
)
if let inputFormat {
self.systemDiagnostics.selectedInputFormat = self.audioFormatSummary(inputFormat)
self.systemDiagnostics.selectedInputSampleRate = inputFormat.sampleRate
converter = AVAudioConverter(from: inputFormat, to: targetFormat)
}
guard let inputFormat, let converter else { return }
self.processAudioBuffer(
{ self.copyAudioBuffer(from: previousBuffer.audioBufferList, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system,
callbackTimestamp: previousInputTime,
ioTimestamp: pendingNow
)
pendingBuffer = nil
pendingInputTime = nil
pendingNow = nil
self.processAudioBuffer(
{ self.copyAudioBuffer(from: currentBuffer.audioBufferList, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system,
callbackTimestamp: inInputTime.pointee,
ioTimestamp: inNow.pointee
)
return
}
guard let inputFormat, let converter else { return }
// The tap queue is serial. Reusing the converter preserves its
// resampler state instead of discarding audio at every callback.
self.processAudioBuffer(
{ self.copyAudioBuffer(from: inputData, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system
source: .system,
callbackTimestamp: inInputTime.pointee,
ioTimestamp: inNow.pointee
)
} invalidationHandler: { [weak self] _ in
guard let self, !self.isRestartingSystemTap, self.isRecording else { return }
guard let self, self.isRecording else { return }
Task { await self.restartSystemAudioTap() }
}
}
private func inputFormat(
for inputData: UnsafePointer<AudioBufferList>,
sampleRate: Double
) -> AVAudioFormat? {
let buffers = UnsafeMutableAudioBufferListPointer(
UnsafeMutablePointer(mutating: inputData)
)
let channelCount = buffers.reduce(UInt32(0)) { $0 + $1.mNumberChannels }
guard channelCount > 0 else { return nil }
// HAL I/O proc samples use the canonical Float32 representation. The
// tap's stream description can advertise a different common format;
// using that to interpret the callback bytes can halve the frame count
// (for example, treating four-byte Float32 samples as eight-byte
// Float64 samples). The callback's AudioBufferList is authoritative for
// its channel layout, while its sample rate comes from the input stream.
let isInterleaved = buffers.count == 1 && channelCount > 1
return AVAudioFormat(
commonFormat: .pcmFormatFloat32,
sampleRate: sampleRate,
channels: AVAudioChannelCount(channelCount),
interleaved: isInterleaved
)
}
private func effectiveInputSampleRate(
advertisedSampleRate: Double,
previousFrameLength: AVAudioFrameCount,
previousTimestamp: AudioTimeStamp,
currentTimestamp: AudioTimeStamp,
previousHostTimestamp: AudioTimeStamp,
currentHostTimestamp: AudioTimeStamp
) -> Double {
let sampleTimeDelta = currentTimestamp.mSampleTime - previousTimestamp.mSampleTime
guard advertisedSampleRate > 0, previousFrameLength > 0 else {
systemDiagnostics.inputRateInference = "advertised (invalid advertised rate or frame count)"
return advertisedSampleRate
}
guard currentHostTimestamp.mHostTime > previousHostTimestamp.mHostTime else {
systemDiagnostics.inputRateInference = "advertised (host timestamps unavailable)"
return advertisedSampleRate
}
let hostTimeDelta = currentHostTimestamp.mHostTime - previousHostTimestamp.mHostTime
let hostNanoseconds = AudioConvertHostTimeToNanos(hostTimeDelta)
guard hostNanoseconds > 0 else {
systemDiagnostics.inputRateInference = "advertised (host-time conversion failed)"
return advertisedSampleRate
}
let inferredSampleRate = Double(previousFrameLength) * 1_000_000_000 / Double(hostNanoseconds)
let plausibleRange = (advertisedSampleRate * 0.25)...(advertisedSampleRate * 1.25)
guard inferredSampleRate.isFinite, plausibleRange.contains(inferredSampleRate) else {
systemDiagnostics.inputRateInference = String(
format: "advertised (invalid host inference %.3f from %u frames / %llu ns; sampleTimeDelta=%.3f)",
inferredSampleRate,
previousFrameLength,
hostNanoseconds,
sampleTimeDelta
)
return advertisedSampleRate
}
let commonSampleRates: [Double] = [8_000, 11_025, 12_000, 16_000, 22_050, 24_000, 32_000, 44_100, 48_000, 88_200, 96_000]
let selectedSampleRate = commonSampleRates
.min(by: { abs($0 - inferredSampleRate) < abs($1 - inferredSampleRate) })
.flatMap { abs($0 - inferredSampleRate) / $0 <= 0.01 ? $0 : nil }
?? inferredSampleRate
systemDiagnostics.inputRateInference = String(
format: "source=hostTime,advertised=%.3f,inferred=%.3f,selected=%.3f,previousFrames=%u,hostNanoseconds=%llu,sampleTimeDelta=%.3f",
advertisedSampleRate,
inferredSampleRate,
selectedSampleRate,
previousFrameLength,
hostNanoseconds,
sampleTimeDelta
)
return selectedSampleRate
}
private func copyAudioBuffer(
from inputData: UnsafePointer<AudioBufferList>,
format: AVAudioFormat
) -> AVAudioPCMBuffer? {
guard let borrowedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
bufferListNoCopy: inputData,
deallocator: nil
), borrowedBuffer.frameLength > 0,
let ownedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
frameCapacity: borrowedBuffer.frameLength
) else { return nil }
ownedBuffer.frameLength = borrowedBuffer.frameLength
let sourceBuffers = UnsafeMutableAudioBufferListPointer(
UnsafeMutablePointer(mutating: inputData)
)
if systemDiagnostics.firstBufferLayout == nil {
systemDiagnostics.firstBufferLayout = sourceBuffers.enumerated().map { index, buffer in
"buffer\(index):channels=\(buffer.mNumberChannels),bytes=\(buffer.mDataByteSize)"
}.joined(separator: "; ")
}
let frameLengths = sourceBuffers.compactMap { source -> AVAudioFrameCount? in
let bytesPerFrame = Int(source.mNumberChannels) * MemoryLayout<Float32>.size
guard bytesPerFrame > 0,
Int(source.mDataByteSize).isMultiple(of: bytesPerFrame) else { return nil }
return AVAudioFrameCount(Int(source.mDataByteSize) / bytesPerFrame)
}
guard frameLengths.count == sourceBuffers.count,
let frameLength = frameLengths.first,
frameLength > 0,
frameLengths.allSatisfy({ $0 == frameLength }),
let ownedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
frameCapacity: frameLength
) else { return nil }
// Set the frame count explicitly instead of asking AVAudioPCMBuffer to
// infer it from potentially inconsistent tap metadata.
ownedBuffer.frameLength = frameLength
let destinationBuffers = UnsafeMutableAudioBufferListPointer(
ownedBuffer.mutableAudioBufferList
)
@@ -360,20 +658,44 @@ final class AudioManager: NSObject, ObservableObject {
_ inputBufferProvider: () -> AVAudioPCMBuffer?,
converter: AVAudioConverter,
targetFormat: AVAudioFormat,
source: AudioSource
source: AudioSource,
callbackTimestamp: AudioTimeStamp? = nil,
ioTimestamp: AudioTimeStamp? = nil
) {
// The system callback copies its borrowed Core Audio memory while this
// lock prevents teardown, then conversion operates on the owned copy.
audioFileLock.lock()
defer { audioFileLock.unlock() }
guard isAcceptingAudio,
let inputBuffer = inputBufferProvider(),
inputBuffer.frameLength > 0 else { return }
guard isAcceptingAudio else { return }
if source == .system {
systemDiagnostics.callbackCount += 1
let callbackAt = Date()
if systemDiagnostics.firstCallbackAt == nil { systemDiagnostics.firstCallbackAt = callbackAt }
systemDiagnostics.lastCallbackAt = callbackAt
if let callbackTimestamp {
if systemDiagnostics.firstSampleTime == nil { systemDiagnostics.firstSampleTime = callbackTimestamp.mSampleTime }
systemDiagnostics.lastSampleTime = callbackTimestamp.mSampleTime
}
if let ioTimestamp {
if systemDiagnostics.firstHostTime == nil { systemDiagnostics.firstHostTime = ioTimestamp.mHostTime }
systemDiagnostics.lastHostTime = ioTimestamp.mHostTime
}
}
guard let inputBuffer = inputBufferProvider(), inputBuffer.frameLength > 0 else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
if source == .system {
systemDiagnostics.inputFrameCount += UInt64(inputBuffer.frameLength)
}
updateAudioLevel(inputBuffer, source: source)
let ratio = targetFormat.sampleRate / inputBuffer.format.sampleRate
let capacity = max(1, AVAudioFrameCount(ceil(Double(inputBuffer.frameLength) * ratio)))
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: capacity) else { return }
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: capacity) else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
var suppliedInput = false
var conversionError: NSError?
let status = converter.convert(to: outputBuffer, error: &conversionError) { _, outputStatus in
@@ -385,7 +707,10 @@ final class AudioManager: NSObject, ObservableObject {
outputStatus.pointee = .haveData
return inputBuffer
}
guard status != .error, conversionError == nil, outputBuffer.frameLength > 0 else { return }
guard status != .error, conversionError == nil, outputBuffer.frameLength > 0 else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
do {
switch source {
@@ -393,6 +718,7 @@ final class AudioManager: NSObject, ObservableObject {
try micAudioFile?.write(from: outputBuffer)
case .system:
try systemAudioFile?.write(from: outputBuffer)
systemDiagnostics.outputFrameCount += UInt64(outputBuffer.frameLength)
}
} catch {
DispatchQueue.main.async { [weak self] in
@@ -463,29 +789,213 @@ final class AudioManager: NSObject, ObservableObject {
for url in urls { try? FileManager.default.removeItem(at: url) }
}
private func preserveAudioFiles(_ urls: [URL], sessionID: UUID) -> URL? {
guard !urls.isEmpty,
let documents = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask).first else {
return nil
private func preserveAudioFiles(_ urls: [URL], meetingID: UUID) -> URL? {
LocalStorageManager.shared.preserveAudioFiles(urls, for: meetingID)
}
private func preservedAudioFiles(_ urls: [URL?], in folder: URL?) -> [URL?] {
urls.map { sourceURL in
guard let sourceURL else { return nil }
guard let folder else { return sourceURL }
let preservedURL = folder.appendingPathComponent(sourceURL.lastPathComponent)
return FileManager.default.fileExists(atPath: preservedURL.path) ? preservedURL : sourceURL
}
let folder = documents
.appendingPathComponent("Meetingnotes-Recovery", isDirectory: true)
.appendingPathComponent(sessionID.uuidString, isDirectory: true)
}
private func captureDurationMismatch(in files: [URL?]) -> String? {
guard files.count >= 2,
let micDuration = audioDuration(at: files[0]),
let systemDuration = audioDuration(at: files[1]),
micDuration >= 60 else { return nil }
let ratio = systemDuration / micDuration
guard (0.45...0.55).contains(ratio) || (1.8...2.2).contains(ratio) else { return nil }
return String(format: "mic %.1fs, system %.1fs", micDuration, systemDuration)
}
private func repairHalfDurationSystemWAVIfNeeded(in files: [URL?]) -> Bool {
guard files.count >= 2,
let micDuration = audioDuration(at: files[0]),
let systemURL = files[1],
let systemDuration = audioDuration(at: systemURL),
micDuration >= 60,
systemURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame,
(0.48...0.52).contains(systemDuration / micDuration) else { return false }
do {
try FileManager.default.createDirectory(at: folder, withIntermediateDirectories: true)
try halveWAVSampleRate(at: systemURL)
return true
} catch {
return nil
return false
}
var preservedCount = 0
for url in urls {
do {
try FileManager.default.moveItem(at: url, to: folder.appendingPathComponent(url.lastPathComponent))
preservedCount += 1
} catch {
continue
}
private func halveWAVSampleRate(at url: URL) throws {
let handle = try FileHandle(forUpdating: url)
defer { try? handle.close() }
try handle.seek(toOffset: 0)
guard let riffHeader = try handle.read(upToCount: 12),
riffHeader.count == 12,
String(data: riffHeader[0..<4], encoding: .ascii) == "RIFF",
String(data: riffHeader[8..<12], encoding: .ascii) == "WAVE" else {
throw NSError(domain: "AudioManager", code: -2, userInfo: [NSLocalizedDescriptionKey: "Invalid WAV header"])
}
var offset: UInt64 = 12
while true {
try handle.seek(toOffset: offset)
guard let chunkHeader = try handle.read(upToCount: 8), chunkHeader.count == 8 else { break }
let chunkID = String(data: chunkHeader[0..<4], encoding: .ascii)
let chunkSize = UInt32(chunkHeader[4])
| (UInt32(chunkHeader[5]) << 8)
| (UInt32(chunkHeader[6]) << 16)
| (UInt32(chunkHeader[7]) << 24)
let chunkDataOffset = offset + 8
if chunkID == "fmt ", chunkSize >= 16 {
try handle.seek(toOffset: chunkDataOffset)
guard let format = try handle.read(upToCount: 16), format.count == 16 else { break }
let audioFormat = UInt16(format[0]) | (UInt16(format[1]) << 8)
let blockAlign = UInt16(format[12]) | (UInt16(format[13]) << 8)
let sampleRate = UInt32(format[4])
| (UInt32(format[5]) << 8)
| (UInt32(format[6]) << 16)
| (UInt32(format[7]) << 24)
guard audioFormat == 1, sampleRate >= 16_000, sampleRate.isMultiple(of: 2) else {
throw NSError(domain: "AudioManager", code: -3, userInfo: [NSLocalizedDescriptionKey: "Unsupported WAV format"])
}
let correctedSampleRate = sampleRate / 2
let correctedByteRate = correctedSampleRate * UInt32(blockAlign)
try handle.seek(toOffset: chunkDataOffset + 4)
try handle.write(contentsOf: littleEndianData(correctedSampleRate))
try handle.write(contentsOf: littleEndianData(correctedByteRate))
try handle.synchronize()
return
}
offset = chunkDataOffset + UInt64(chunkSize) + UInt64(chunkSize % 2)
}
return preservedCount > 0 ? folder : nil
throw NSError(domain: "AudioManager", code: -4, userInfo: [NSLocalizedDescriptionKey: "WAV format chunk was not found"])
}
private func littleEndianData(_ value: UInt32) -> Data {
var littleEndianValue = value.littleEndian
return withUnsafeBytes(of: &littleEndianValue) { Data($0) }
}
private func audioDuration(at url: URL?) -> TimeInterval? {
guard let url,
let file = try? AVAudioFile(forReading: url),
file.processingFormat.sampleRate > 0 else { return nil }
return Double(file.length) / file.processingFormat.sampleRate
}
private func streamDescriptionSummary(_ description: AudioStreamBasicDescription?) -> String {
guard let description else { return "unavailable" }
return String(
format: "sampleRate=%.3f,formatID=%u,flags=%u,bytesPerPacket=%u,framesPerPacket=%u,bytesPerFrame=%u,channels=%u,bitsPerChannel=%u",
description.mSampleRate,
description.mFormatID,
description.mFormatFlags,
description.mBytesPerPacket,
description.mFramesPerPacket,
description.mBytesPerFrame,
description.mChannelsPerFrame,
description.mBitsPerChannel
)
}
private func audioFormatSummary(_ format: AVAudioFormat) -> String {
"sampleRate=\(format.sampleRate),channels=\(format.channelCount),commonFormat=\(format.commonFormat.rawValue),interleaved=\(format.isInterleaved)"
}
private func audioFileSummary(_ url: URL?) -> String {
guard let url else { return "missing" }
guard let file = try? AVAudioFile(forReading: url), file.processingFormat.sampleRate > 0 else {
return "\(url.lastPathComponent):unreadable"
}
let duration = Double(file.length) / file.processingFormat.sampleRate
return String(
format: "%@:sampleRate=%.3f,channels=%u,frames=%lld,duration=%.6f",
url.lastPathComponent,
file.processingFormat.sampleRate,
file.processingFormat.channelCount,
file.length,
duration
)
}
private func writeCaptureDiagnostics(
sessionID: UUID,
captureStartedAt: Date,
preRepairFileSummaries: [String],
repairedFiles: [URL?],
repairApplied: Bool,
audioFolder: URL?
) {
guard let audioFolder else { return }
let callbackDuration = systemDiagnostics.firstCallbackAt.flatMap { first in
systemDiagnostics.lastCallbackAt.map { $0.timeIntervalSince(first) }
}
let sampleTimeDelta = systemDiagnostics.firstSampleTime.flatMap { first in
systemDiagnostics.lastSampleTime.map { $0 - first }
}
let hostTimeDelta = systemDiagnostics.firstHostTime.flatMap { first in
systemDiagnostics.lastHostTime.map { $0 >= first ? $0 - first : 0 }
}
let inputFrameDuration = systemDiagnostics.selectedInputSampleRate.flatMap { sampleRate in
sampleRate > 0 ? Double(systemDiagnostics.inputFrameCount) / sampleRate : nil
}
let outputFrameDuration = systemDiagnostics.targetSampleRate.flatMap { sampleRate in
sampleRate > 0 ? Double(systemDiagnostics.outputFrameCount) / sampleRate : nil
}
let observedInputRate = callbackDuration.flatMap { duration in
duration > 0 ? Double(systemDiagnostics.inputFrameCount) / duration : nil
}
let preRepairMic = preRepairFileSummaries.indices.contains(0) ? preRepairFileSummaries[0] : "missing"
let preRepairSystem = preRepairFileSummaries.indices.contains(1) ? preRepairFileSummaries[1] : "missing"
let postRepairMic = repairedFiles.indices.contains(0) ? audioFileSummary(repairedFiles[0]) : "missing"
let postRepairSystem = repairedFiles.indices.contains(1) ? audioFileSummary(repairedFiles[1]) : "missing"
let callbackDurationLine = callbackDuration.map { String(format: "callbackWallDuration=%.6f", $0) } ?? "callbackWallDuration=unavailable"
let inputFrameDurationLine = inputFrameDuration.map { String(format: "inputFrameDurationAtSelectedRate=%.6f", $0) } ?? "inputFrameDurationAtSelectedRate=unavailable"
let outputFrameDurationLine = outputFrameDuration.map { String(format: "outputFrameDurationAtTargetRate=%.6f", $0) } ?? "outputFrameDurationAtTargetRate=unavailable"
let observedInputRateLine = observedInputRate.map { String(format: "observedInputFramesPerSecond=%.3f", $0) } ?? "observedInputFramesPerSecond=unavailable"
let sampleTimeDeltaLine = sampleTimeDelta.map { String(format: "sampleTimeDelta=%.6f", $0) } ?? "sampleTimeDelta=unavailable"
let hostTimeDeltaLine = hostTimeDelta.map { "hostTimeDelta=\($0)" } ?? "hostTimeDelta=unavailable"
let lines: [String] = [
"Meetingnotes system capture diagnostics",
"sessionID=\(sessionID.uuidString)",
"createdAt=\(ISO8601DateFormatter().string(from: Date()))",
String(format: "captureWallDuration=%.6f", Date().timeIntervalSince(captureStartedAt)),
"tapAdvertisedFormat=\(systemDiagnostics.tapAdvertisedFormat)",
"aggregateInputFormat=\(systemDiagnostics.aggregateInputFormat)",
"selectedInputFormat=\(systemDiagnostics.selectedInputFormat)",
"inputRateInference=\(systemDiagnostics.inputRateInference)",
"targetFormat=\(systemDiagnostics.targetFormat)",
"firstBufferLayout=\(systemDiagnostics.firstBufferLayout ?? "unavailable")",
"callbackCount=\(systemDiagnostics.callbackCount)",
"inputFrameCount=\(systemDiagnostics.inputFrameCount)",
"outputFrameCount=\(systemDiagnostics.outputFrameCount)",
"discardedCallbackCount=\(systemDiagnostics.discardedCallbackCount)",
callbackDurationLine,
inputFrameDurationLine,
outputFrameDurationLine,
observedInputRateLine,
sampleTimeDeltaLine,
hostTimeDeltaLine,
"repairApplied=\(repairApplied)",
"preRepairMic=\(preRepairMic)",
"preRepairSystem=\(preRepairSystem)",
"postRepairMic=\(postRepairMic)",
"postRepairSystem=\(postRepairSystem)"
]
let diagnosticsURL = audioFolder.appendingPathComponent("audio-diagnostics-\(sessionID.uuidString).txt")
try? lines.joined(separator: "\n").appending("\n").write(
to: diagnosticsURL,
atomically: true,
encoding: .utf8
)
}
private func resetAudioLevels() {
+1 -1
View File
@@ -8,7 +8,7 @@ import Security
class KeychainHelper {
static let shared = KeychainHelper()
private let serviceName = "net.jamesbone.meetingnotes"
private let serviceName = Bundle.main.bundleIdentifier ?? "net.jamesbone.meetingnotes"
private init() {}
@@ -2,6 +2,7 @@
// Handles local storage of meetings and app data
import Foundation
import AppKit
/// Manages local file storage for meetings and app data
class LocalStorageManager {
@@ -15,6 +16,7 @@ class LocalStorageManager {
private let documentsDirectory: URL
private let meetingsDirectory: URL
private let templatesDirectory: URL
private let recoveryDirectory: URL
private init() {
// Get the app's documents directory
@@ -27,11 +29,17 @@ class LocalStorageManager {
// Create templates subdirectory
templatesDirectory = documentsDirectory.appendingPathComponent("Templates")
recoveryDirectory = documentsDirectory.appendingPathComponent("Meetingnotes Audio")
// Ensure directories exist
try? FileManager.default.createDirectory(at: meetingsDirectory,
withIntermediateDirectories: true)
try? FileManager.default.createDirectory(at: templatesDirectory,
withIntermediateDirectories: true)
try? FileManager.default.createDirectory(at: recoveryDirectory,
withIntermediateDirectories: true)
migrateLegacyRecoveryAudio()
purgeExpiredAudioFolders()
}
// MARK: - Meeting Management
@@ -144,6 +152,253 @@ class LocalStorageManager {
}
}
func mergeMeeting(_ continuation: Meeting, into previous: Meeting) -> Meeting? {
guard continuation.id != previous.id, continuation.date >= previous.date else { return nil }
let previousFolder = recoveryDirectory.appendingPathComponent(previous.id.uuidString, isDirectory: true)
let continuationFolder = recoveryAudioFolder(named: continuation.id.uuidString)
var copiedAudioURLs: [URL] = []
var createdPreviousFolder = false
if let continuationFolder {
if !FileManager.default.fileExists(atPath: previousFolder.path) {
do {
try FileManager.default.createDirectory(at: previousFolder, withIntermediateDirectories: true)
createdPreviousFolder = true
} catch {
return nil
}
}
let recoveryFiles = recoveryAudioFiles(in: continuationFolder).map(\.url)
+ recoveryDiagnosticFiles(in: continuationFolder)
for recoveryFile in recoveryFiles {
let destination = previousFolder.appendingPathComponent(recoveryFile.lastPathComponent)
if FileManager.default.fileExists(atPath: destination.path) {
guard FileManager.default.contentsEqual(
atPath: recoveryFile.path,
andPath: destination.path
) else {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
continue
}
do {
try FileManager.default.copyItem(at: recoveryFile, to: destination)
copiedAudioURLs.append(destination)
} catch {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
}
}
var merged = previous
var seenChunkIDs = Set<UUID>()
merged.transcriptChunks = (previous.transcriptChunks + continuation.transcriptChunks)
.filter { seenChunkIDs.insert($0.id).inserted }
.sorted {
if $0.timestamp == $1.timestamp {
return $0.id.uuidString < $1.id.uuidString
}
return $0.timestamp < $1.timestamp
}
merged.userNotes = mergedText(previous.userNotes, continuation.userNotes)
merged.generatedNotes = mergedText(previous.generatedNotes, continuation.generatedNotes)
merged.templateId = previous.templateId ?? continuation.templateId
merged.transcriptionError = previous.transcriptionError ?? continuation.transcriptionError
if !recoveryAudioFiles(in: previousFolder).isEmpty {
merged.recoveryAudioFolderName = previous.id.uuidString
}
guard saveMeeting(merged) else {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
guard deleteMeeting(continuation) else {
_ = saveMeeting(previous)
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
if let continuationFolder {
try? FileManager.default.removeItem(at: continuationFolder)
}
return merged
}
private func mergedText(_ first: String, _ second: String) -> String {
let first = first.trimmingCharacters(in: .whitespacesAndNewlines)
let second = second.trimmingCharacters(in: .whitespacesAndNewlines)
guard !first.isEmpty else { return second }
guard !second.isEmpty, second != first else { return first }
return first + "\n\n---\n\n" + second
}
private func rollbackMergedAudio(_ copiedURLs: [URL], removeFolder folder: URL?) {
for url in copiedURLs {
try? FileManager.default.removeItem(at: url)
}
if let folder, recoveryAudioFiles(in: folder).isEmpty {
try? FileManager.default.removeItem(at: folder)
}
}
// MARK: - Recovery Audio
func preserveAudioFiles(_ urls: [URL], for meetingID: UUID) -> URL? {
guard !urls.isEmpty else { return nil }
purgeExpiredAudioFolders()
let folder = recoveryDirectory.appendingPathComponent(meetingID.uuidString, isDirectory: true)
do {
try FileManager.default.createDirectory(at: folder, withIntermediateDirectories: true)
} catch {
return nil
}
var preservedCount = 0
for url in urls {
let destination = folder.appendingPathComponent(url.lastPathComponent)
do {
if FileManager.default.fileExists(atPath: destination.path) {
try FileManager.default.removeItem(at: destination)
}
try FileManager.default.moveItem(at: url, to: destination)
preservedCount += 1
} catch {
continue
}
}
return preservedCount > 0 ? folder : nil
}
func purgeExpiredAudioFolders(now: Date = Date()) {
guard let folders = try? FileManager.default.contentsOfDirectory(
at: recoveryDirectory,
includingPropertiesForKeys: [.isDirectoryKey],
options: [.skipsHiddenFiles]
) else { return }
let retentionInterval = TimeInterval(UserDefaultsManager.shared.audioRetentionDays) * 24 * 60 * 60
let expirationDate = now.addingTimeInterval(-retentionInterval)
for folder in folders {
guard (try? folder.resourceValues(forKeys: [.isDirectoryKey]).isDirectory) == true else { continue }
let audioFiles = recoveryAudioFiles(in: folder)
let newestDate = audioFiles.compactMap { file -> Date? in
let values = try? file.url.resourceValues(forKeys: [.contentModificationDateKey, .creationDateKey])
return values?.contentModificationDate ?? values?.creationDate
}.max()
guard let newestDate else {
try? FileManager.default.removeItem(at: folder)
continue
}
if newestDate < expirationDate {
try? FileManager.default.removeItem(at: folder)
}
}
}
@MainActor
func showAudioFolderInFinder(_ folder: URL? = nil) {
let target = folder ?? recoveryDirectory
try? FileManager.default.createDirectory(at: target, withIntermediateDirectories: true)
NSWorkspace.shared.open(target)
}
func recoveryAudioFolder(named name: String) -> URL? {
let trimmedName = name.trimmingCharacters(in: .whitespacesAndNewlines)
guard !trimmedName.isEmpty,
URL(fileURLWithPath: trimmedName).lastPathComponent == trimmedName else {
return nil
}
let folder = recoveryDirectory.appendingPathComponent(trimmedName, isDirectory: true)
var isDirectory: ObjCBool = false
guard FileManager.default.fileExists(atPath: folder.path, isDirectory: &isDirectory),
isDirectory.boolValue,
!recoveryAudioFiles(in: folder).isEmpty else {
return nil
}
return folder
}
func recoveryAudioFiles(in folder: URL) -> [(url: URL, source: AudioSource)] {
guard let files = try? FileManager.default.contentsOfDirectory(
at: folder,
includingPropertiesForKeys: [.creationDateKey, .contentModificationDateKey, .isRegularFileKey],
options: [.skipsHiddenFiles]
) else {
return []
}
return files.compactMap { url in
let name = url.lastPathComponent.lowercased()
guard ["m4a", "mp3", "wav", "flac", "webm", "mp4"].contains(url.pathExtension.lowercased()) else {
return nil
}
if name.contains("-mic.") {
return (url, .mic)
}
if name.contains("-system.") {
return (url, .system)
}
return nil
}
}
private func recoveryDiagnosticFiles(in folder: URL) -> [URL] {
guard let files = try? FileManager.default.contentsOfDirectory(
at: folder,
includingPropertiesForKeys: [.isRegularFileKey],
options: [.skipsHiddenFiles]
) else {
return []
}
return files.filter { url in
let values = try? url.resourceValues(forKeys: [.isRegularFileKey])
return values?.isRegularFile == true
&& url.pathExtension.caseInsensitiveCompare("txt") == .orderedSame
&& url.lastPathComponent.lowercased().hasPrefix("audio-diagnostics-")
}
}
func findRecoveryAudioFolder(for meeting: Meeting) -> URL? {
let canonicalName = meeting.id.uuidString
guard meeting.recoveryAudioFolderName == nil
|| meeting.recoveryAudioFolderName?.caseInsensitiveCompare(canonicalName) == .orderedSame else {
return nil
}
return recoveryAudioFolder(named: canonicalName)
}
func deleteRecoveryAudioFolder(_ folder: URL) {
guard folder.deletingLastPathComponent().standardizedFileURL == recoveryDirectory.standardizedFileURL else {
return
}
try? FileManager.default.removeItem(at: folder)
}
private func migrateLegacyRecoveryAudio() {
let legacyDirectory = documentsDirectory.appendingPathComponent("Meetingnotes-Recovery", isDirectory: true)
guard let folders = try? FileManager.default.contentsOfDirectory(
at: legacyDirectory,
includingPropertiesForKeys: [.isDirectoryKey],
options: [.skipsHiddenFiles]
) else { return }
for folder in folders {
let destination = recoveryDirectory.appendingPathComponent(folder.lastPathComponent, isDirectory: true)
guard !FileManager.default.fileExists(atPath: destination.path) else { continue }
try? FileManager.default.moveItem(at: folder, to: destination)
}
try? FileManager.default.removeItem(at: legacyDirectory)
}
/// Imports meeting JSON files from a folder selected by the user.
func importMeetings(from directory: URL) throws -> MeetingImportResult {
let didStartAccess = directory.startAccessingSecurityScopedResource()
@@ -2,6 +2,12 @@ import Foundation
import SwiftUI
import Combine
struct RecordingCompletion {
let chunks: [TranscriptChunk]
let recoveryAudioFolderName: String?
let transcriptionError: String?
}
/// Manages recording sessions at the app level to persist across navigation
@MainActor
class RecordingSessionManager: ObservableObject {
@@ -14,9 +20,11 @@ class RecordingSessionManager: ObservableObject {
@Published var errorMessage: String?
@Published var activeRecordingTranscriptChunksUpdated: [TranscriptChunk] = []
private let audioManager = AudioManager.shared
private var audioManager = AudioManager.shared
private var cancellables = Set<AnyCancellable>()
private var audioManagerCancellables = Set<AnyCancellable>()
private let transcriptUpdateSubject = PassthroughSubject<[TranscriptChunk], Never>()
private var processingMeetingIds = Set<UUID>()
// Store transcript chunks for the active recording session
private var activeRecordingTranscriptChunks: [TranscriptChunk] = []
@@ -27,24 +35,19 @@ class RecordingSessionManager: ObservableObject {
}
private func setupAudioManagerBindings() {
audioManagerCancellables.removeAll()
// Bind to audio manager state
audioManager.$isRecording
.sink { [weak self] isRecording in
self?.isRecording = isRecording
}
.store(in: &cancellables)
audioManager.$isProcessing
.sink { [weak self] isProcessing in
self?.isProcessing = isProcessing
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
audioManager.$errorMessage
.sink { [weak self] errorMessage in
self?.errorMessage = errorMessage
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
// When transcript chunks change, store them for the active recording and send to debouncer
audioManager.$transcriptChunks
@@ -55,7 +58,7 @@ class RecordingSessionManager: ObservableObject {
self.transcriptUpdateSubject.send(newChunks)
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
}
private func setupDebouncedSaving() {
@@ -81,25 +84,43 @@ class RecordingSessionManager: ObservableObject {
activeMeetingId = meetingId
recordingStartedAt = Date()
audioManager.startRecording()
audioManager.startRecording(for: meetingId)
}
func stopRecording() async -> [TranscriptChunk] {
func stopRecording() async -> RecordingCompletion {
print("🛑 Stopping recording for meeting: \(activeMeetingId?.uuidString ?? "unknown")")
guard let meetingId = activeMeetingId else {
audioManager.cancelRecording()
recordingStartedAt = nil
return []
return RecordingCompletion(chunks: [], recoveryAudioFolderName: nil, transcriptionError: nil)
}
recordingStartedAt = nil
let chunks = await audioManager.stopRecordingAndTranscribe()
activeRecordingTranscriptChunks = chunks
activeRecordingTranscriptChunksUpdated = chunks
updateActiveMeetingTranscript(meetingId: meetingId, chunks: chunks)
let completedAudioManager = audioManager
audioManager = AudioManager()
setupAudioManagerBindings()
activeMeetingId = nil
activeRecordingTranscriptChunks = []
return chunks
recordingStartedAt = nil
processingMeetingIds.insert(meetingId)
isProcessing = true
defer {
processingMeetingIds.remove(meetingId)
isProcessing = !processingMeetingIds.isEmpty
}
let chunks = await completedAudioManager.stopRecordingAndTranscribe()
let completion = RecordingCompletion(
chunks: chunks,
recoveryAudioFolderName: completedAudioManager.lastRecoveryAudioFolderName,
transcriptionError: completedAudioManager.errorMessage
)
updateActiveMeetingTranscript(
meetingId: meetingId,
chunks: chunks,
recoveryAudioFolderName: completion.recoveryAudioFolderName,
transcriptionError: completion.transcriptionError
)
return completion
}
func cancelRecording() {
@@ -113,13 +134,26 @@ class RecordingSessionManager: ObservableObject {
return isRecording && activeMeetingId == meetingId
}
private func updateActiveMeetingTranscript(meetingId: UUID, chunks: [TranscriptChunk]) {
func isProcessingMeeting(_ meetingId: UUID) -> Bool {
processingMeetingIds.contains(meetingId)
}
private func updateActiveMeetingTranscript(
meetingId: UUID,
chunks: [TranscriptChunk],
recoveryAudioFolderName: String? = nil,
transcriptionError: String? = nil
) {
// Load all meetings
var meetings = LocalStorageManager.shared.loadMeetings()
// Find and update the active meeting
if let index = meetings.firstIndex(where: { $0.id == meetingId }) {
meetings[index].transcriptChunks = chunks
if let recoveryAudioFolderName {
meetings[index].recoveryAudioFolderName = recoveryAudioFolderName
}
meetings[index].transcriptionError = transcriptionError
// Save the updated meeting
let success = LocalStorageManager.shared.saveMeeting(meetings[index])
@@ -23,6 +23,7 @@ class UserDefaultsManager {
static let transcriptionModel = "transcriptionModel"
static let muteDeckAPIEnabled = "muteDeckAPIEnabled"
static let muteDeckAPIPort = "muteDeckAPIPort"
static let audioRetentionDays = "audioRetentionDays"
}
// MARK: - User Blurb
@@ -78,7 +79,7 @@ class UserDefaultsManager {
}
var transcriptionModel: String {
get { userDefaults.string(forKey: Keys.transcriptionModel) ?? "groq/whisper-large-v3-turbo" }
get { userDefaults.string(forKey: Keys.transcriptionModel) ?? "local-parakeet/parakeet-tdt-0.6b-v3" }
set { userDefaults.set(newValue, forKey: Keys.transcriptionModel) }
}
@@ -94,4 +95,12 @@ class UserDefaultsManager {
}
set { userDefaults.set(newValue, forKey: Keys.muteDeckAPIPort) }
}
var audioRetentionDays: Int {
get {
guard userDefaults.object(forKey: Keys.audioRetentionDays) != nil else { return 3 }
return min(max(userDefaults.integer(forKey: Keys.audioRetentionDays), 1), 365)
}
set { userDefaults.set(min(max(newValue, 1), 365), forKey: Keys.audioRetentionDays) }
}
}
+131 -1
View File
@@ -6,12 +6,14 @@
//
import SwiftUI
import AppKit
import Sparkle
import PostHog
@main
struct MeetingnotesApp: App {
private let updaterController: SPUStandardUpdaterController
@StateObject private var recordingSessionManager = RecordingSessionManager.shared
init() {
updaterController = SPUStandardUpdaterController(updaterDelegate: nil, userDriverDelegate: nil)
@@ -34,17 +36,145 @@ struct MeetingnotesApp: App {
}
var body: some Scene {
WindowGroup {
WindowGroup("Meetingnotes", id: "main") {
ContentView()
.frame(minWidth: 700, minHeight: 400)
.background(MainWindowConfigurator())
}
.windowResizability(.automatic)
.defaultSize(width: 1000, height: 600)
.commands {
CommandGroup(after: .appInfo) {
CheckForUpdatesView(updater: updaterController.updater)
Divider()
Button("Show Audio Folder") {
LocalStorageManager.shared.showAudioFolderInFinder()
}
}
}
MenuBarExtra {
MeetingnotesMenu(
recordingSessionManager: recordingSessionManager,
updater: updaterController.updater
)
} label: {
Image(systemName: recordingSessionManager.isRecording ? "record.circle.fill" : "waveform")
.accessibilityLabel(recordingSessionManager.isRecording ? "Meetingnotes recording" : "Meetingnotes")
}
}
}
private struct MeetingnotesMenu: View {
@ObservedObject var recordingSessionManager: RecordingSessionManager
let updater: SPUUpdater
@Environment(\.openWindow) private var openWindow
var body: some View {
if recordingSessionManager.isRecording {
Label("Recording", systemImage: "record.circle.fill")
} else if recordingSessionManager.isProcessing {
Label("Processing audio", systemImage: "hourglass")
}
Button {
MainWindowController.shared.show(using: openWindow)
} label: {
Label("Open Meetingnotes", systemImage: "macwindow")
}
Button {
LocalStorageManager.shared.showAudioFolderInFinder()
} label: {
Label("Show Audio Folder", systemImage: "folder")
}
Divider()
Button {
updater.checkForUpdates()
} label: {
Label("Check for Updates...", systemImage: "arrow.triangle.2.circlepath")
}
.keyboardShortcut("u")
Button {
NSApplication.shared.terminate(nil)
} label: {
Label("Quit Meetingnotes", systemImage: "power")
}
.keyboardShortcut("q")
}
}
@MainActor
private final class MainWindowController: NSObject {
static let shared = MainWindowController()
private weak var window: NSWindow?
func configure(_ window: NSWindow) {
if self.window !== window {
if let previousWindow = self.window {
NotificationCenter.default.removeObserver(
self,
name: NSWindow.willMiniaturizeNotification,
object: previousWindow
)
}
self.window = window
NotificationCenter.default.addObserver(
self,
selector: #selector(windowWillMiniaturize(_:)),
name: NSWindow.willMiniaturizeNotification,
object: window
)
}
if let minimizeButton = window.standardWindowButton(.miniaturizeButton) {
minimizeButton.target = self
minimizeButton.action = #selector(hideWindow(_:))
}
}
func show(using openWindow: OpenWindowAction) {
if let window {
window.makeKeyAndOrderFront(nil)
} else {
openWindow(id: "main")
}
NSApplication.shared.activate(ignoringOtherApps: true)
}
@objc private func hideWindow(_ sender: NSButton) {
sender.window?.orderOut(nil)
}
@objc private func windowWillMiniaturize(_ notification: Notification) {
guard let window = notification.object as? NSWindow else { return }
DispatchQueue.main.async {
window.deminiaturize(nil)
window.orderOut(nil)
}
}
}
private struct MainWindowConfigurator: NSViewRepresentable {
func makeNSView(context: Context) -> NSView {
let view = NSView()
configureWindow(for: view)
return view
}
func updateNSView(_ nsView: NSView, context: Context) {
configureWindow(for: nsView)
}
private func configureWindow(for view: NSView) {
DispatchQueue.main.async {
guard let window = view.window else { return }
MainWindowController.shared.configure(window)
}
}
}
+27 -4
View File
@@ -48,30 +48,46 @@ struct TranscriptChunk: Codable, Identifiable, Hashable {
let id: UUID
let timestamp: Date
let source: AudioSource
let speaker: Int?
let text: String
let isFinal: Bool
init(id: UUID = UUID(), timestamp: Date = Date(), source: AudioSource, text: String, isFinal: Bool = false) {
init(id: UUID = UUID(), timestamp: Date = Date(), source: AudioSource, speaker: Int? = nil, text: String, isFinal: Bool = false) {
self.id = id
self.timestamp = timestamp
self.source = source
self.speaker = speaker
self.text = text
self.isFinal = isFinal
}
var displayName: String {
if source == .mic { return "Me" }
if let speaker { return "Speaker \(speaker)" }
return source.displayName
}
}
struct CollapsedTranscriptChunk: Identifiable {
let id: UUID
let timestamp: Date
let source: AudioSource
let speaker: Int?
let combinedText: String
init(id: UUID = UUID(), timestamp: Date, source: AudioSource, combinedText: String) {
init(id: UUID = UUID(), timestamp: Date, source: AudioSource, speaker: Int? = nil, combinedText: String) {
self.id = id
self.timestamp = timestamp
self.source = source
self.speaker = speaker
self.combinedText = combinedText
}
var displayName: String {
if source == .mic { return "Me" }
if let speaker { return "Speaker \(speaker)" }
return source.displayName
}
}
struct Meeting: Codable, Identifiable, Hashable {
@@ -82,6 +98,8 @@ struct Meeting: Codable, Identifiable, Hashable {
var userNotes: String
var generatedNotes: String
var templateId: UUID? // Add property to track per-meeting template
var recoveryAudioFolderName: String?
var transcriptionError: String?
// MARK: - Data versioning
/// Version of this Meeting record on disk. Useful for migration.
var dataVersion: Int
@@ -95,6 +113,8 @@ struct Meeting: Codable, Identifiable, Hashable {
userNotes: String = "",
generatedNotes: String = "",
templateId: UUID? = nil,
recoveryAudioFolderName: String? = nil,
transcriptionError: String? = nil,
dataVersion: Int = Meeting.currentDataVersion) {
self.id = id
self.date = date
@@ -103,6 +123,8 @@ struct Meeting: Codable, Identifiable, Hashable {
self.userNotes = userNotes
self.generatedNotes = generatedNotes
self.templateId = templateId
self.recoveryAudioFolderName = recoveryAudioFolderName
self.transcriptionError = transcriptionError
self.dataVersion = dataVersion
}
@@ -112,7 +134,7 @@ struct Meeting: Codable, Identifiable, Hashable {
var transcript: String {
return transcriptChunks
.filter { $0.isFinal }
.map { "[\($0.source.rawValue)] \($0.text)" }
.map { "[\($0.displayName)] \($0.text)" }
.joined(separator: " ")
}
@@ -121,7 +143,7 @@ struct Meeting: Codable, Identifiable, Hashable {
let finalChunks = transcriptChunks.filter { $0.isFinal }
return finalChunks.map { chunk in
"[\(TranscriptTimestampFormatter.string(from: chunk.timestamp))] \(chunk.source.copyPrefix): \(chunk.text)"
"[\(TranscriptTimestampFormatter.string(from: chunk.timestamp))] \(chunk.displayName): \(chunk.text)"
}.joined(separator: "\n")
}
@@ -132,6 +154,7 @@ struct Meeting: Codable, Identifiable, Hashable {
id: chunk.id,
timestamp: chunk.timestamp,
source: chunk.source,
speaker: chunk.speaker,
combinedText: chunk.text
)
}
+5
View File
@@ -54,6 +54,11 @@ struct Settings: Codable {
set { UserDefaultsManager.shared.muteDeckAPIPort = newValue }
}
var audioRetentionDays: Int {
get { UserDefaultsManager.shared.audioRetentionDays }
set { UserDefaultsManager.shared.audioRetentionDays = newValue }
}
// System prompt default loading
static func defaultSystemPrompt() -> String {
guard let path = Bundle.main.path(forResource: "DefaultSystemPrompt", ofType: "txt"),
@@ -116,6 +116,47 @@ extension AudioObjectID {
try read(kAudioTapPropertyFormat, defaultValue: AudioStreamBasicDescription())
}
/// Reads the virtual format of the first input stream exposed by this device.
///
/// An aggregate device can adapt a tap to the active output hardware. Its
/// input stream format is therefore the format delivered to the I/O proc,
/// which can differ from the tap object's originally advertised format.
func readInputStreamBasicDescription() throws -> AudioStreamBasicDescription {
var streamsAddress = AudioObjectPropertyAddress(
mSelector: kAudioDevicePropertyStreams,
mScope: kAudioObjectPropertyScopeGlobal,
mElement: kAudioObjectPropertyElementMain
)
var dataSize: UInt32 = 0
var status = AudioObjectGetPropertyDataSize(self, &streamsAddress, 0, nil, &dataSize)
guard status == noErr else {
throw "Error reading device stream list size: \(status)"
}
var streamIDs = [AudioObjectID](
repeating: .unknown,
count: Int(dataSize) / MemoryLayout<AudioObjectID>.size
)
status = AudioObjectGetPropertyData(self, &streamsAddress, 0, nil, &dataSize, &streamIDs)
guard status == noErr else {
throw "Error reading device stream list: \(status)"
}
for streamID in streamIDs {
let direction: UInt32 = try streamID.read(
kAudioStreamPropertyDirection,
defaultValue: 0
)
guard direction == 1 else { continue }
return try streamID.read(
kAudioStreamPropertyVirtualFormat,
defaultValue: AudioStreamBasicDescription()
)
}
throw "Device has no input stream."
}
private func requireSystemObject() throws {
if self != .system { throw "Only supported for the system object." }
}
+28 -9
View File
@@ -5,7 +5,7 @@ import AVFoundation
enum TapTarget {
case singleProcess(AudioProcess)
case systemAudio(processObjectIDs: [AudioObjectID])
case systemAudio
var displayName: String {
switch self {
@@ -63,6 +63,10 @@ final class ProcessTap {
@ObservationIgnored
private(set) var tapStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private(set) var tapAdvertisedStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private(set) var aggregateInputStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private var invalidationHandler: InvalidationHandler?
@ObservationIgnored
@@ -137,12 +141,12 @@ final class ProcessTap {
case .singleProcess(let process):
tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID])
logger.debug("Configuring tap for single process objectID: \(process.objectID)")
case .systemAudio(let processObjectIDs):
if processObjectIDs.isEmpty {
logger.warning("System audio tap configured with an empty list of processObjectIDs. This might not capture any audio or behave unexpectedly.")
}
tapDescription = CATapDescription(monoMixdownOfProcesses: processObjectIDs)
logger.debug("Configuring tap for system audio output using \(processObjectIDs.count) explicit processes.")
case .systemAudio:
// The transcription file is mono, so ask Core Audio for a mono
// mixdown at the source. This avoids interpreting a stereo HAL
// buffer as half as many frames before the 16 kHz conversion.
tapDescription = CATapDescription(monoGlobalTapButExcludeProcesses: [])
logger.info("Configuring a mono global system audio tap.")
}
tapDescription.uuid = UUID()
@@ -248,8 +252,23 @@ final class ProcessTap {
do {
logger.debug("Attempting to read audio tap stream basic description for tapID #\(tapID)...")
self.tapStreamDescription = try tapID.readAudioTapStreamBasicDescription()
logger.debug("Successfully read tap stream description: \(String(describing: self.tapStreamDescription))")
let advertisedDescription = try tapID.readAudioTapStreamBasicDescription()
self.tapAdvertisedStreamDescription = advertisedDescription
// The aggregate device may adapt the tap to the active hardware's
// sample rate. Its input stream is the format actually delivered
// to the I/O proc, so use that rather than the tap's pre-aggregate
// advertisement. Using the latter can halve the written duration
// when, for example, a 48 kHz tap is delivered at 24 kHz.
do {
let aggregateDescription = try aggregateDeviceID.readInputStreamBasicDescription()
self.aggregateInputStreamDescription = aggregateDescription
self.tapStreamDescription = aggregateDescription
logger.info("Using aggregate input stream description: \(String(describing: self.tapStreamDescription), privacy: .public); tap advertised: \(String(describing: advertisedDescription), privacy: .public)")
} catch {
self.tapStreamDescription = advertisedDescription
logger.warning("Could not read aggregate input stream format; using tap format: \(error, privacy: .public)")
}
} catch {
logger.error("Failed to read audio tap stream basic description for tapID #\(tapID): \(error)")
throw error // Propagate error
+269 -8
View File
@@ -1,3 +1,4 @@
import AVFoundation
import Foundation
struct CoderModel: Codable, Identifiable, Hashable {
@@ -20,6 +21,7 @@ struct CoderModel: Codable, Identifiable, Hashable {
var supportsChat: Bool { capabilities.isEmpty || capabilities.contains("chat") }
var supportsTranscription: Bool { capabilities.contains("audio_transcription") }
var supportsSpeakerDiarization: Bool { capabilities.contains("speaker_diarization") }
}
enum CoderAPIError: LocalizedError {
@@ -28,6 +30,7 @@ enum CoderAPIError: LocalizedError {
case missingModel(String)
case invalidResponse
case serviceError(Int, String)
case audioPreparationFailed(String, String)
var errorDescription: String? {
switch self {
@@ -41,6 +44,8 @@ enum CoderAPIError: LocalizedError {
return "Coder returned an invalid response."
case .serviceError(let status, let message):
return "Coder request failed (\(status)): \(message)"
case .audioPreparationFailed(let filename, let message):
return "Could not prepare \(filename) for transcription: \(message)"
}
}
}
@@ -53,6 +58,7 @@ final class CoderAPIClient {
let start: TimeInterval
let end: TimeInterval
let text: String
let speaker: Int?
}
let text: String
@@ -69,10 +75,25 @@ final class CoderAPIClient {
}
private struct TranscriptionResponse: Decodable {
struct Word: Decodable {
let word: String
let start: TimeInterval
let end: TimeInterval
let speaker: Int?
}
let text: String
let segments: [Transcription.Segment]?
let words: [Word]?
}
private struct AudioChunk {
let url: URL
let offset: TimeInterval
let isTemporary: Bool
}
private let transcriptionChunkDuration: TimeInterval = 3 * 60
private let transcriptionSession: URLSession
private init() {
@@ -97,7 +118,11 @@ final class CoderAPIClient {
)
}
func streamChat(systemPrompt: String, model: String) -> AsyncThrowingStream<String, Error> {
func streamChat(
systemPrompt: String,
userPrompt: String = "Create the meeting notes now.",
model: String
) -> AsyncThrowingStream<String, Error> {
AsyncThrowingStream { continuation in
Task {
do {
@@ -113,7 +138,7 @@ final class CoderAPIClient {
"model": selectedModel,
"messages": [
["role": "system", "content": systemPrompt],
["role": "user", "content": "Create the meeting notes now."]
["role": "user", "content": userPrompt]
],
"stream": true
])
@@ -147,15 +172,90 @@ final class CoderAPIClient {
}
}
func transcribe(fileURL: URL, model: String, language: String = "en") async throws -> Transcription {
func transcribe(
fileURL: URL,
model: String,
language: String = "en",
diarization: Bool = false,
maxSpeakerCount: Int = 4
) async throws -> Transcription {
let selectedModel = model.trimmingCharacters(in: .whitespacesAndNewlines)
guard !selectedModel.isEmpty else { throw CoderAPIError.missingModel("transcription") }
let apiKey = try requiredAPIKey(KeychainHelper.shared.getCoderAPIKey() ?? "")
let chunks: [AudioChunk]
do {
chunks = try makeAudioChunks(from: fileURL, preserveSpeakerIdentity: diarization)
} catch {
throw CoderAPIError.audioPreparationFailed(fileURL.lastPathComponent, error.localizedDescription)
}
defer {
for chunk in chunks where chunk.isTemporary {
try? FileManager.default.removeItem(at: chunk.url)
}
}
var textParts: [String] = []
var segments: [Transcription.Segment] = []
var lastNormalizedText = ""
var consecutiveDuplicateCount = 0
for chunk in chunks {
let transcription = try await transcribeChunk(
chunk.url,
model: selectedModel,
language: language,
apiKey: apiKey,
diarization: diarization,
maxSpeakerCount: maxSpeakerCount
)
if transcription.segments.isEmpty {
let text = transcription.text.trimmingCharacters(in: .whitespacesAndNewlines)
if !text.isEmpty {
textParts.append(text)
segments.append(.init(start: chunk.offset, end: chunk.offset, text: text, speaker: nil))
}
continue
}
for segment in transcription.segments {
let text = segment.text.trimmingCharacters(in: .whitespacesAndNewlines)
guard !text.isEmpty else { continue }
let normalized = text.lowercased()
if normalized == lastNormalizedText {
consecutiveDuplicateCount += 1
} else {
lastNormalizedText = normalized
consecutiveDuplicateCount = 1
}
guard consecutiveDuplicateCount <= 2 else { continue }
textParts.append(text)
segments.append(.init(
start: segment.start + chunk.offset,
end: segment.end + chunk.offset,
text: text,
speaker: segment.speaker
))
}
}
return Transcription(text: textParts.joined(separator: "\n"), segments: segments)
}
private func transcribeChunk(
_ fileURL: URL,
model: String,
language: String,
apiKey: String,
diarization: Bool,
maxSpeakerCount: Int
) async throws -> Transcription {
let boundary = "Meetingnotes-\(UUID().uuidString)"
let bodyURL = try makeMultipartBody(
audioURL: fileURL,
model: selectedModel,
model: model,
language: language,
diarization: diarization,
maxSpeakerCount: maxSpeakerCount,
boundary: boundary
)
defer { try? FileManager.default.removeItem(at: bodyURL) }
@@ -168,10 +268,160 @@ final class CoderAPIClient {
let size = attributes[.size] as? NSNumber {
request.setValue(size.stringValue, forHTTPHeaderField: "Content-Length")
}
let (data, response) = try await transcriptionSession.upload(for: request, fromFile: bodyURL)
let (data, response) = try await uploadTranscription(request: request, bodyURL: bodyURL)
try validate(response: response, data: data)
let decoded = try JSONDecoder().decode(TranscriptionResponse.self, from: data)
return Transcription(text: decoded.text, segments: decoded.segments ?? [])
let segments = decoded.segments ?? segments(from: decoded.words ?? [])
return Transcription(text: decoded.text, segments: segments)
}
private func uploadTranscription(request: URLRequest, bodyURL: URL) async throws -> (Data, URLResponse) {
let retryDelays: [UInt64] = [3, 10]
for attempt in 0...retryDelays.count {
do {
return try await transcriptionSession.upload(for: request, fromFile: bodyURL)
} catch {
guard attempt < retryDelays.count, isRetryableTranscriptionError(error) else {
throw error
}
try await Task.sleep(nanoseconds: retryDelays[attempt] * 1_000_000_000)
}
}
throw CoderAPIError.invalidResponse
}
private func isRetryableTranscriptionError(_ error: Error) -> Bool {
guard let urlError = error as? URLError else { return false }
switch urlError.code {
case .networkConnectionLost, .cannotConnectToHost, .timedOut:
return true
default:
return false
}
}
private func makeAudioChunks(from fileURL: URL, preserveSpeakerIdentity: Bool) throws -> [AudioChunk] {
let input = try AVAudioFile(forReading: fileURL)
let format = input.processingFormat
guard format.sampleRate > 0 else { throw CoderAPIError.invalidResponse }
let framesPerChunk = preserveSpeakerIdentity
? max(1, input.length)
: AVAudioFramePosition(format.sampleRate * transcriptionChunkDuration)
if fileURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame,
input.fileFormat.streamDescription.pointee.mFormatID == kAudioFormatLinearPCM,
input.length <= framesPerChunk {
return [AudioChunk(url: fileURL, offset: 0, isTemporary: false)]
}
var chunks: [AudioChunk] = []
var frameOffset: AVAudioFramePosition = 0
do {
while frameOffset < input.length {
let frameCount = min(framesPerChunk, input.length - frameOffset)
let chunkURL = FileManager.default.temporaryDirectory
.appendingPathComponent("meetingnotes-transcription-\(UUID().uuidString).wav")
try writeAudioChunk(
from: input,
frameCount: frameCount,
format: format,
to: chunkURL
)
chunks.append(AudioChunk(
url: chunkURL,
offset: Double(frameOffset) / format.sampleRate,
isTemporary: true
))
frameOffset += frameCount
}
return chunks
} catch {
for chunk in chunks {
try? FileManager.default.removeItem(at: chunk.url)
}
throw error
}
}
private func writeAudioChunk(
from input: AVAudioFile,
frameCount: AVAudioFramePosition,
format: AVAudioFormat,
to outputURL: URL
) throws {
let settings: [String: Any] = [
AVFormatIDKey: kAudioFormatLinearPCM,
AVSampleRateKey: format.sampleRate,
AVNumberOfChannelsKey: format.channelCount,
AVLinearPCMBitDepthKey: 16,
AVLinearPCMIsFloatKey: false,
AVLinearPCMIsBigEndianKey: false,
AVLinearPCMIsNonInterleaved: false
]
let output = try AVAudioFile(
forWriting: outputURL,
settings: settings,
commonFormat: format.commonFormat,
interleaved: format.isInterleaved
)
var remaining = frameCount
while remaining > 0 {
let requestedFrames = AVAudioFrameCount(min(remaining, 8_192))
guard let buffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: requestedFrames) else {
throw CoderAPIError.invalidResponse
}
try input.read(into: buffer, frameCount: requestedFrames)
guard buffer.frameLength > 0 else { break }
try output.write(from: buffer)
remaining -= AVAudioFramePosition(buffer.frameLength)
}
}
private func segments(from words: [TranscriptionResponse.Word]) -> [Transcription.Segment] {
var result: [Transcription.Segment] = []
var currentWords: [String] = []
var currentStart: TimeInterval?
var currentEnd: TimeInterval = 0
var currentSpeaker: Int?
func flush() {
guard let start = currentStart, !currentWords.isEmpty else { return }
result.append(.init(
start: start,
end: currentEnd,
text: currentWords.joined(separator: " "),
speaker: currentSpeaker
))
currentWords.removeAll(keepingCapacity: true)
currentStart = nil
currentEnd = 0
currentSpeaker = nil
}
for word in words {
let text = word.word.trimmingCharacters(in: .whitespacesAndNewlines)
guard !text.isEmpty else { continue }
let speakerChanged = currentStart != nil && word.speaker != currentSpeaker
let longPause = currentStart != nil && word.start - currentEnd > 1.5
if speakerChanged || longPause { flush() }
if currentStart == nil {
currentStart = word.start
currentSpeaker = word.speaker
}
currentWords.append(text)
currentEnd = word.end
let sentenceEnded = text.last.map { ".!?".contains($0) } ?? false
let duration = currentEnd - (currentStart ?? currentEnd)
if currentWords.count >= 40 || (sentenceEnded && (currentWords.count >= 12 || duration >= 8)) {
flush()
}
}
flush()
return result
}
private func endpoint(baseURL: String, path: String) throws -> URL {
@@ -203,7 +453,14 @@ final class CoderAPIClient {
}
}
private func makeMultipartBody(audioURL: URL, model: String, language: String, boundary: String) throws -> URL {
private func makeMultipartBody(
audioURL: URL,
model: String,
language: String,
diarization: Bool,
maxSpeakerCount: Int,
boundary: String
) throws -> URL {
let bodyURL = FileManager.default.temporaryDirectory.appendingPathComponent("meetingnotes-upload-\(UUID().uuidString).body")
_ = FileManager.default.createFile(atPath: bodyURL.path, contents: nil)
let output = try FileHandle(forWritingTo: bodyURL)
@@ -215,7 +472,11 @@ final class CoderAPIClient {
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\n\(model)\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"language\"\r\n\r\n\(language)\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"response_format\"\r\n\r\nverbose_json\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"file\"; filename=\"\(audioURL.lastPathComponent)\"\r\nContent-Type: audio/mp4\r\n\r\n")
if diarization {
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"diarization\"\r\n\r\ntrue\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"max_speaker_count\"\r\n\r\n\(maxSpeakerCount)\r\n")
}
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"file\"; filename=\"\(audioURL.lastPathComponent)\"\r\nContent-Type: audio/wav\r\n\r\n")
let input = try FileHandle(forReadingFrom: audioURL)
defer { try? input.close() }
while let chunk = try input.read(upToCount: 1 << 20), !chunk.isEmpty {
+51 -14
View File
@@ -182,13 +182,13 @@ private enum LocalAPIRouter {
}
private enum LocalRecordingError: LocalizedError {
case processing
case stopping
case saveFailed
var errorDescription: String? {
switch self {
case .processing:
return "The previous meeting is still processing."
case .stopping:
return "The current recording is still stopping."
case .saveFailed:
return "Could not create a meeting for this recording."
}
@@ -200,18 +200,20 @@ private final class LocalRecordingController {
static let shared = LocalRecordingController()
private let recordingManager = RecordingSessionManager.shared
private let stopGraceNanoseconds: UInt64 = 10_000_000_000
private var isStopping = false
private var pendingStopTask: Task<Void, Never>?
private init() {}
func statusPayload() -> [String: Any] {
let state: String
if isStopping || recordingManager.isProcessing {
state = "processing"
} else if recordingManager.isRecording {
if recordingManager.isRecording {
state = "recording"
} else if recordingManager.activeMeetingId != nil {
state = "starting"
} else if isStopping || recordingManager.isProcessing {
state = "processing"
} else {
state = "idle"
}
@@ -237,8 +239,14 @@ private final class LocalRecordingController {
}
func startRecording() throws -> [String: Any] {
if isStopping || recordingManager.isProcessing {
throw LocalRecordingError.processing
if let pendingStopTask {
pendingStopTask.cancel()
self.pendingStopTask = nil
isStopping = false
return statusPayload()
}
if isStopping {
throw LocalRecordingError.stopping
}
if recordingManager.activeMeetingId != nil {
return statusPayload()
@@ -258,14 +266,22 @@ private final class LocalRecordingController {
return statusPayload()
}
isStopping = true
Task { [weak self] in
pendingStopTask = Task { [weak self] in
try? await Task.sleep(nanoseconds: self?.stopGraceNanoseconds ?? 0)
guard !Task.isCancelled else { return }
await self?.finishRecording()
}
return statusPayload()
}
func cancelRecording() -> [String: Any] {
guard !isStopping else { return statusPayload() }
if let pendingStopTask {
pendingStopTask.cancel()
self.pendingStopTask = nil
isStopping = false
} else if isStopping {
return statusPayload()
}
let meetingID = recordingManager.activeMeetingId
recordingManager.cancelRecording()
if let meetingID,
@@ -277,15 +293,18 @@ private final class LocalRecordingController {
}
private func finishRecording() async {
pendingStopTask = nil
let meetingID = recordingManager.activeMeetingId
let chunks = await recordingManager.stopRecording()
isStopping = false
let completion = await recordingManager.stopRecording()
guard let meetingID,
var meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == meetingID }) else {
isStopping = false
return
}
meeting.transcriptChunks = chunks
meeting.transcriptChunks = completion.chunks
meeting.recoveryAudioFolderName = completion.recoveryAudioFolderName
meeting.transcriptionError = completion.transcriptionError
let templates = LocalStorageManager.shared.loadTemplates()
if meeting.templateId == nil {
meeting.templateId = LocalStorageManager.shared.preferredTemplateID(in: templates)
@@ -309,13 +328,31 @@ private final class LocalRecordingController {
generatedNotes = ""
}
}
var meetingChanged = false
if !generatedNotes.isEmpty {
meeting.generatedNotes = generatedNotes
meetingChanged = true
}
if meeting.title.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
do {
if let title = try await NotesGenerator.shared.generateMeetingTitle(
meeting: meeting,
generatedNotes: generatedNotes,
templateId: meeting.templateId
) {
meeting.title = title
meetingChanged = true
}
} catch {
// The recording and generated notes remain valid without a generated title.
print("Meeting title generation failed: \(error)")
}
}
if meetingChanged {
_ = LocalStorageManager.shared.saveMeeting(meeting)
NotificationCenter.default.post(name: .meetingSaved, object: meeting)
}
}
isStopping = false
}
}
@@ -106,6 +106,88 @@ class NotesGenerator {
}
}
/// Generates a short title after meeting notes have been created.
func generateMeetingTitle(
meeting: Meeting,
generatedNotes: String,
templateId: UUID?
) async throws -> String? {
let templates = LocalStorageManager.shared.loadTemplates()
let template = templateId.flatMap { id in
templates.first(where: { $0.id == id })
}
let meetingContext: String
if let template {
meetingContext = "\(template.title): \(template.context)"
} else {
meetingContext = "General meeting"
}
let trimmedNotes = generatedNotes.trimmingCharacters(in: .whitespacesAndNewlines)
let meetingContent = trimmedNotes.isEmpty ? meeting.formattedTranscript : trimmedNotes
guard !meetingContent.isEmpty else { return nil }
let systemPrompt = """
Create a concise, descriptive title for a recorded meeting from its context and content.
Return only the title in 3 to 8 words. Do not use quotation marks, Markdown, or a trailing period.
Avoid generic titles such as Meeting, Discussion, Meeting Notes, or Untitled Meeting.
"""
let userPrompt = """
Meeting context:
\(meetingContext)
Meeting content:
\(meetingContent)
"""
var generatedTitle = ""
let stream = CoderAPIClient.shared.streamChat(
systemPrompt: systemPrompt,
userPrompt: userPrompt,
model: UserDefaultsManager.shared.notesModel
)
for try await content in stream {
generatedTitle += content
}
return normalizedTitle(generatedTitle)
}
private func normalizedTitle(_ value: String) -> String? {
guard var title = value
.split(whereSeparator: \Character.isNewline)
.map(String.init)
.first(where: { !$0.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty })?
.trimmingCharacters(in: .whitespacesAndNewlines) else {
return nil
}
while title.hasPrefix("#") {
title.removeFirst()
title = title.trimmingCharacters(in: .whitespaces)
}
if title.lowercased().hasPrefix("title:") {
title = String(title.dropFirst("title:".count))
.trimmingCharacters(in: .whitespacesAndNewlines)
}
title = title.trimmingCharacters(in: CharacterSet(charactersIn: "\"'"))
if title.hasSuffix(".") {
title.removeLast()
}
if title.count > 80 {
let prefix = String(title.prefix(80))
if let lastSpace = prefix.lastIndex(of: " ") {
title = String(prefix[..<lastSpace])
} else {
title = prefix
}
}
title = title.trimmingCharacters(in: .whitespacesAndNewlines)
return title.isEmpty ? nil : title
}
/// Validates if the Coder service token is configured
/// - Returns: True if API key exists, false otherwise
func isConfigured() -> Bool {
@@ -67,6 +67,28 @@ class MeetingListViewModel: ObservableObject {
_ = LocalStorageManager.shared.deleteMeeting(meeting)
}
func previousMeeting(for meeting: Meeting) -> Meeting? {
meetings
.filter { $0.id != meeting.id && $0.date < meeting.date }
.max { $0.date < $1.date }
}
func mergeIntoPrevious(_ meeting: Meeting) -> Meeting? {
guard let previous = previousMeeting(for: meeting),
let merged = LocalStorageManager.shared.mergeMeeting(meeting, into: previous) else {
errorMessage = "The meetings could not be merged. Their original records and audio were kept."
return nil
}
meetings.removeAll { $0.id == meeting.id || $0.id == previous.id }
meetings.append(merged)
meetings.sort { $0.date > $1.date }
NotificationCenter.default.post(name: .meetingSaved, object: merged)
NotificationCenter.default.post(name: .meetingDeleted, object: meeting)
PostHogSDK.shared.capture("meetings_merged")
return merged
}
func createNewMeeting() -> Meeting {
let newMeeting = Meeting(templateId: LocalStorageManager.shared.preferredTemplateID())
meetings.insert(newMeeting, at: 0)
+92 -9
View File
@@ -12,7 +12,7 @@ extension Notification.Name {
enum MeetingViewTab: String, CaseIterable {
case myNotes = "My Notes"
case transcript = "Transcript"
case enhancedNotes = "Enhanced Notes"
case enhancedNotes = "Meeting Notes"
}
@@ -25,6 +25,8 @@ class MeetingViewModel: ObservableObject {
@Published private var recordingStateChanged = false // Trigger SwiftUI updates
@Published var isValidatingKey = false // Indicates API key validation in progress
@Published var isStartingRecording = false // Indicates recording start in progress
@Published var isRetryingTranscription = false
@Published private(set) var recoveryAudioFolderURL: URL?
// Computed property to determine if Generate button should animate
var shouldAnimateGenerateButton: Bool {
@@ -44,7 +46,13 @@ class MeetingViewModel: ObservableObject {
}
var isProcessing: Bool {
return recordingSessionManager.isProcessing && recordingSessionManager.activeMeetingId == meeting.id
return isRetryingTranscription ||
recordingSessionManager.isProcessingMeeting(meeting.id)
}
var canRetryTranscription: Bool {
recoveryAudioFolderURL != nil &&
!isRecording && !isProcessing && !isStartingRecording && !isValidatingKey
}
@Published var selectedTab: MeetingViewTab = .transcript // Default to transcript tab
@@ -76,6 +84,7 @@ class MeetingViewModel: ObservableObject {
// Load templates and selected template
loadTemplates()
refreshRecoveryAudioFolder()
// Observe template selection: save to meeting and regenerate notes on changes (skip initial)
$selectedTemplateId
.dropFirst()
@@ -128,18 +137,27 @@ class MeetingViewModel: ObservableObject {
self.meeting.transcriptChunks = recordingSessionManager.getTranscriptChunks(for: meeting.id)
}
// Listen for final transcript updates for this meeting.
// Listen for transcript updates emitted while this meeting is recording.
recordingSessionManager.$activeRecordingTranscriptChunksUpdated
.dropFirst()
.sink { [weak self] updatedChunks in
guard let self = self else { return }
// Only update if this meeting is the active recording
if recordingSessionManager.isRecordingMeeting(self.meeting.id) {
if recordingSessionManager.activeMeetingId == self.meeting.id {
self.meeting.transcriptChunks = updatedChunks
}
}
.store(in: &cancellables)
NotificationCenter.default.publisher(for: .meetingSaved)
.compactMap { $0.object as? Meeting }
.filter { [weak self] in $0.id == self?.meeting.id }
.sink { [weak self] savedMeeting in
guard let self, !self.isDeleted, self.meeting != savedMeeting else { return }
self.meeting = savedMeeting
self.refreshRecoveryAudioFolder()
}
.store(in: &cancellables)
// Auto-save when meeting properties change
@@ -203,8 +221,12 @@ class MeetingViewModel: ObservableObject {
func stopRecording() {
isStartingRecording = true
Task {
let chunks = await recordingSessionManager.stopRecording()
meeting.transcriptChunks = chunks
let completion = await recordingSessionManager.stopRecording()
meeting.transcriptChunks = completion.chunks
meeting.recoveryAudioFolderName = completion.recoveryAudioFolderName
meeting.transcriptionError = completion.transcriptionError
errorMessage = completion.transcriptionError
refreshRecoveryAudioFolder()
saveMeeting()
if !meeting.formattedTranscript.isEmpty {
await generateNotes()
@@ -213,6 +235,47 @@ class MeetingViewModel: ObservableObject {
}
}
func retryTranscription() {
guard let recoveryAudioFolderURL, canRetryTranscription else { return }
isRetryingTranscription = true
errorMessage = nil
Task {
defer { isRetryingTranscription = false }
do {
let chunks = try await AudioManager.shared.transcribeRecoveryAudio(
in: recoveryAudioFolderURL,
captureStartedAt: meeting.date
)
meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
meeting.transcriptionError = nil
selectedTab = .transcript
guard saveMeeting() else {
throw CocoaError(.fileWriteUnknown)
}
await generateNotes()
} catch {
errorMessage = error.localizedDescription
meeting.transcriptionError = error.localizedDescription
_ = saveMeeting()
print("Retry transcription failed: \(error)")
}
}
}
private func refreshRecoveryAudioFolder() {
recoveryAudioFolderURL = LocalStorageManager.shared.findRecoveryAudioFolder(for: meeting)
meeting.recoveryAudioFolderName = recoveryAudioFolderURL?.lastPathComponent
}
func showAudioInFinder() {
guard let recoveryAudioFolderURL else { return }
LocalStorageManager.shared.showAudioFolderInFinder(recoveryAudioFolderURL)
}
func loadTemplates() {
templates = LocalStorageManager.shared.loadTemplates()
@@ -260,20 +323,40 @@ class MeetingViewModel: ObservableObject {
// Only save if there was no error
if !hasError {
await generateTitleIfNeeded()
saveMeeting()
}
isGeneratingNotes = false
}
func saveMeeting() {
if isDeleted { return }
private func generateTitleIfNeeded() async {
guard meeting.title.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty else { return }
do {
if let title = try await NotesGenerator.shared.generateMeetingTitle(
meeting: meeting,
generatedNotes: meeting.generatedNotes,
templateId: selectedTemplateId
) {
meeting.title = title
}
} catch {
// A title is helpful but should never prevent completed notes from being saved.
print("Meeting title generation failed: \(error)")
}
}
@discardableResult
func saveMeeting() -> Bool {
if isDeleted { return false }
print("💾 Saving meeting: \(meeting.id)")
let success = LocalStorageManager.shared.saveMeeting(meeting)
print("💾 Save result: \(success ? "SUCCESS" : "FAILED")")
if success {
NotificationCenter.default.post(name: .meetingSaved, object: meeting)
}
return success
}
func copyCurrentTabContent() {
@@ -71,6 +71,7 @@ class SettingsViewModel: ObservableObject {
// via computed properties when they're modified
let coderSaved = KeychainHelper.shared.saveCoderAPIKey(settings.coderAPIKey)
LocalAPIServer.shared.applyConfiguration()
LocalStorageManager.shared.purgeExpiredAudioFolders()
if showMessage {
if coderSaved {
+164 -23
View File
@@ -6,6 +6,8 @@ struct MeetingListView: View {
@StateObject private var recordingSessionManager = RecordingSessionManager.shared
@State private var selectedMeeting: Meeting?
@State private var navigationPath = NavigationPath()
@State private var recordingFailureMessage: String?
@State private var failedMeetingID: UUID?
var body: some View {
NavigationSplitView {
@@ -23,6 +25,38 @@ struct MeetingListView: View {
.background(Color.clear)
}
}
.onReceive(recordingSessionManager.$activeMeetingId.compactMap { $0 }) { meetingID in
if let meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == meetingID }) {
selectedMeeting = meeting
}
}
.onReceive(NotificationCenter.default.publisher(for: .meetingSaved)) { notification in
guard let savedMeeting = notification.object as? Meeting,
selectedMeeting?.id == savedMeeting.id else { return }
selectedMeeting = savedMeeting
}
.onReceive(recordingSessionManager.$errorMessage.compactMap { $0 }) { message in
failedMeetingID = recordingSessionManager.activeMeetingId
recordingFailureMessage = message
}
.alert("Transcription Failed", isPresented: Binding(
get: { recordingFailureMessage != nil },
set: { if !$0 { recordingFailureMessage = nil } }
)) {
if failedMeetingID != nil {
Button("View Meeting") {
selectFailedMeeting()
recordingFailureMessage = nil
recordingSessionManager.errorMessage = nil
}
}
Button("OK") {
recordingFailureMessage = nil
recordingSessionManager.errorMessage = nil
}
} message: {
Text(recordingFailureMessage ?? "")
}
}
private var sidebarContent: some View {
@@ -83,10 +117,19 @@ struct MeetingListView: View {
NavigationStack(path: $navigationPath) {
Group {
if let selectedMeeting = selectedMeeting {
MeetingDetailContentView(meeting: selectedMeeting, onDelete: {
// When a meeting is deleted from the detail view, clear the selection
self.selectedMeeting = nil
})
MeetingDetailContentView(
meeting: selectedMeeting,
mergeCandidate: viewModel.previousMeeting(for: selectedMeeting),
onMerge: { meeting in
guard let merged = viewModel.mergeIntoPrevious(meeting) else { return false }
self.selectedMeeting = merged
return true
},
onDelete: {
// When a meeting is deleted from the detail view, clear the selection
self.selectedMeeting = nil
}
)
.id(selectedMeeting.id) // Force recreation when selection changes
} else {
ContentUnavailableView(
@@ -152,6 +195,14 @@ struct MeetingListView: View {
return DayGroup(day: dayString, date: date, meetings: meetings.sorted { $0.date > $1.date })
}.sorted { $0.date > $1.date }
}
private func selectFailedMeeting() {
guard let failedMeetingID,
let meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == failedMeetingID }) else {
return
}
selectedMeeting = meeting
}
}
struct DayGroup {
@@ -173,7 +224,14 @@ struct MeetingRowView: View {
.foregroundColor(.red)
.font(.headline)
}
Text(meeting.title.isEmpty ? "Untitled meeting" : meeting.title)
if meeting.transcriptionError != nil {
Image(systemName: "exclamationmark.triangle.fill")
.foregroundColor(.orange)
.accessibilityLabel("Transcription failed")
}
Text(meeting.title.isEmpty
? (meeting.transcriptionError == nil ? "Untitled meeting" : "Transcription failed")
: meeting.title)
.font(.headline)
.lineLimit(1)
}
@@ -208,12 +266,12 @@ struct CollapsedTranscriptChunkView: View {
.font(.caption)
.foregroundColor(chunk.source == .mic ? .blue : .orange)
Text(chunk.source.displayName)
Text(chunk.displayName)
.font(.caption)
.fontWeight(.medium)
.foregroundColor(chunk.source == .mic ? .blue : .orange)
}
.frame(width: 50, alignment: .leading)
.frame(width: 78, alignment: .leading)
// Transcript text
Text(chunk.combinedText)
@@ -229,12 +287,22 @@ struct MeetingDetailContentView: View {
@StateObject private var viewModel: MeetingViewModel
@StateObject private var recordingSessionManager = RecordingSessionManager.shared
@State private var showDeleteAlert = false
@State private var showMergeAlert = false
@State private var isEditing = false
@State private var showCopyConfirmation = false
let mergeCandidate: Meeting?
let onMerge: (Meeting) -> Bool
let onDelete: () -> Void
init(meeting: Meeting, onDelete: @escaping () -> Void) {
init(
meeting: Meeting,
mergeCandidate: Meeting?,
onMerge: @escaping (Meeting) -> Bool,
onDelete: @escaping () -> Void
) {
self._viewModel = StateObject(wrappedValue: MeetingViewModel(meeting: meeting))
self.mergeCandidate = mergeCandidate
self.onMerge = onMerge
self.onDelete = onDelete
}
@@ -259,6 +327,34 @@ struct MeetingDetailContentView: View {
// Ellipsis menu
Menu {
if viewModel.recoveryAudioFolderURL != nil {
Button {
viewModel.retryTranscription()
} label: {
Label("Retry Transcription", systemImage: "arrow.clockwise")
}
.disabled(!viewModel.canRetryTranscription)
Button {
viewModel.showAudioInFinder()
} label: {
Label("Show Audio in Finder", systemImage: "folder")
}
Divider()
}
if mergeCandidate != nil {
Button {
showMergeAlert = true
} label: {
Label("Merge into Previous Meeting", systemImage: "arrow.triangle.merge")
}
.disabled(recordingSessionManager.isRecording || recordingSessionManager.isProcessing)
Divider()
}
Button("Delete Meeting", role: .destructive) {
showDeleteAlert = true
}
@@ -327,7 +423,7 @@ struct MeetingDetailContentView: View {
}
.buttonStyle(.plain)
.disabled(viewModel.meeting.transcript.isEmpty || viewModel.isGeneratingNotes || viewModel.isRecording || viewModel.isProcessing || viewModel.isStartingRecording)
.help("Generate enhanced notes using a template")
.help("Generate meeting notes using a template")
// Recording Button
Button(action: {
@@ -374,7 +470,7 @@ struct MeetingDetailContentView: View {
Spacer()
// Edit/Preview button (for My Notes and Enhanced Notes)
// Edit/Preview button (for My Notes and Meeting Notes)
if viewModel.selectedTab == .myNotes || viewModel.selectedTab == .enhancedNotes {
Button(action: {
isEditing.toggle()
@@ -444,6 +540,23 @@ struct MeetingDetailContentView: View {
} message: {
Text("Are you sure you want to delete this meeting? This action cannot be undone.")
}
.alert("Merge into Previous Meeting?", isPresented: $showMergeAlert) {
Button("Merge", role: .destructive) {
// Prevent the disappearing detail view from auto-saving the
// continuation after the storage layer removes it.
viewModel.isDeleted = true
if !onMerge(viewModel.meeting) {
viewModel.isDeleted = false
viewModel.errorMessage = "The meetings could not be merged. Their original records and audio were kept."
}
}
Button("Cancel", role: .cancel) { }
} message: {
let previousTitle = mergeCandidate?.title.isEmpty == false
? mergeCandidate?.title ?? "the previous meeting"
: "the previous meeting"
Text("This combines this meeting's transcript, notes, and saved audio into \"\(previousTitle)\", keeps the earlier meeting's title and time, then removes this continuation.")
}
.onDisappear {
// A failed recording may still be empty. Keep it until the user
// explicitly deletes it so app updates cannot erase history.
@@ -478,24 +591,52 @@ struct MeetingDetailContentView: View {
}
private var transcriptView: some View {
ScrollView {
if viewModel.meeting.collapsedTranscriptChunks.isEmpty {
Text("Transcript will appear here...")
.frame(maxWidth: .infinity, alignment: .leading)
.padding()
.foregroundColor(.secondary)
} else {
LazyVStack(alignment: .leading, spacing: 4) {
ForEach(viewModel.meeting.collapsedTranscriptChunks) { chunk in
CollapsedTranscriptChunkView(chunk: chunk)
VStack(alignment: .leading, spacing: 10) {
if let transcriptionError = viewModel.meeting.transcriptionError {
HStack(alignment: .top, spacing: 10) {
Image(systemName: "exclamationmark.triangle.fill")
.foregroundColor(.orange)
VStack(alignment: .leading, spacing: 4) {
Text("Transcription failed")
.font(.headline)
Text(transcriptionError)
.font(.caption)
.foregroundColor(.secondary)
}
Spacer()
if viewModel.recoveryAudioFolderURL != nil {
Button("Retry") {
viewModel.retryTranscription()
}
.disabled(!viewModel.canRetryTranscription)
}
}
.padding()
.background(Color.orange.opacity(0.08))
.cornerRadius(8)
}
ScrollView {
if viewModel.meeting.collapsedTranscriptChunks.isEmpty {
Text(viewModel.meeting.transcriptionError == nil
? "Transcript will appear here..."
: "No transcript was produced. The recording was kept and can be retried.")
.frame(maxWidth: .infinity, alignment: .leading)
.padding()
.foregroundColor(.secondary)
} else {
LazyVStack(alignment: .leading, spacing: 4) {
ForEach(viewModel.meeting.collapsedTranscriptChunks) { chunk in
CollapsedTranscriptChunkView(chunk: chunk)
}
}
.padding()
}
}
.frame(maxHeight: .infinity)
.background(Color.gray.opacity(0.05))
.cornerRadius(8)
}
.frame(maxHeight: .infinity)
.background(Color.gray.opacity(0.05))
.cornerRadius(8)
}
private var enhancedNotesView: some View {
+30 -1
View File
@@ -58,6 +58,18 @@ struct SettingsView: View {
Text(model.displayName).tag(model.id)
}
}
if let selectedModel = viewModel.coderModels.first(where: { $0.id == viewModel.settings.transcriptionModel }) {
if selectedModel.supportsSpeakerDiarization {
Label("Remote participants are labeled Speaker 14; your microphone is labeled Me.", systemImage: "person.2.wave.2")
.font(.caption)
.foregroundColor(.secondary)
} else {
Label("Remote participants are labeled Them; your microphone is labeled Me.", systemImage: "person.2")
.font(.caption)
.foregroundColor(.secondary)
}
}
}
Text("The token is stored locally in Keychain. Audio and note generation are sent only to this Coder service.")
@@ -128,6 +140,23 @@ struct SettingsView: View {
Text("Meeting Storage")
.font(.headline)
LabeledContent("Audio retention") {
Stepper(
value: $viewModel.settings.audioRetentionDays,
in: 1...365
) {
Text("\(viewModel.settings.audioRetentionDays) \(viewModel.settings.audioRetentionDays == 1 ? "day" : "days")")
.monospacedDigit()
.frame(minWidth: 70, alignment: .trailing)
}
}
Button {
LocalStorageManager.shared.showAudioFolderInFinder()
} label: {
Label("Show Audio Folder", systemImage: "folder")
}
Button {
showingMeetingImporter = true
} label: {
@@ -226,7 +255,7 @@ struct SettingsView: View {
// Link to GitHub repository
Link("GitHub",
destination: URL(string: "https://github.com/superdooper86/meetingnotes")!)
destination: URL(string: "https://git.jamesbone.net/coder/meetingnotes")!)
.foregroundColor(.blue)
// Link to landing page
@@ -1,34 +0,0 @@
#!/usr/bin/env bash
set -euo pipefail
REPOSITORY="${REPOSITORY:-superdooper86/meetingnotes}"
if ! command -v gh >/dev/null 2>&1; then
echo "Install GitHub CLI first: brew install gh" >&2
exit 1
fi
gh auth status >/dev/null
read -r -p "Developer ID certificate (.p12) path: " certificate_path
if [[ ! -f "$certificate_path" ]]; then
echo "Certificate not found: $certificate_path" >&2
exit 1
fi
read -r -s -p "Certificate export password: " certificate_password
printf '\n'
read -r -p "Apple ID email: " apple_id
read -r -p "Apple Developer Team ID: " team_id
read -r -s -p "Apple app-specific password: " app_password
printf '\n'
base64 < "$certificate_path" | gh secret set APPLE_CERTIFICATE_P12 -R "$REPOSITORY"
printf '%s' "$certificate_password" | gh secret set APPLE_CERTIFICATE_PASSWORD -R "$REPOSITORY"
printf '%s' "$apple_id" | gh secret set APPLE_ID -R "$REPOSITORY"
printf '%s' "$team_id" | gh secret set APPLE_TEAM_ID -R "$REPOSITORY"
printf '%s' "$app_password" | gh secret set APPLE_APP_PASSWORD -R "$REPOSITORY"
unset certificate_password app_password
echo "Apple release secrets configured for $REPOSITORY."
-139
View File
@@ -1,139 +0,0 @@
#!/usr/bin/env bash
set -euo pipefail
APP_NAME="Meetingnotes"
RUNNER_TEMP="${RUNNER_TEMP:-/tmp}"
PENDING_DIR="${PENDING_DIR:-$RUNNER_TEMP/meetingnotes-pending}"
WORK_ROOT="${WORK_ROOT:-$RUNNER_TEMP/meetingnotes-finalize}"
APP_PATH="$WORK_ROOT/$APP_NAME.app"
RELEASE_DIR="$WORK_ROOT/release"
VERSION_PATH="$PENDING_DIR/version"
COMMIT_SHA_PATH="$PENDING_DIR/commit-sha"
SUBMISSION_PATH="$PENDING_DIR/notary-submission.json"
PRE_NOTARY_ZIP="$PENDING_DIR/$APP_NAME-pre-notary.zip"
required_variables=(
APPLE_ID
APPLE_TEAM_ID
APPLE_APP_PASSWORD
SPARKLE_PRIVATE_KEY
GH_TOKEN
GITHUB_REPOSITORY
)
for variable in "${required_variables[@]}"; do
if [[ -z "${!variable:-}" ]]; then
echo "Missing required environment variable: $variable" >&2
exit 1
fi
done
for path in "$VERSION_PATH" "$COMMIT_SHA_PATH" "$SUBMISSION_PATH" "$PRE_NOTARY_ZIP" "$PENDING_DIR/generate_appcast"; do
if [[ ! -e "$path" ]]; then
echo "Missing release submission artifact: $path" >&2
exit 1
fi
done
VERSION=$(<"$VERSION_PATH")
COMMIT_SHA=$(<"$COMMIT_SHA_PATH")
SUBMISSION_ID=$(plutil -extract id raw -o - "$SUBMISSION_PATH")
TAG="v$VERSION"
if gh release view "$TAG" --repo "$GITHUB_REPOSITORY" >/dev/null 2>&1; then
echo "Release already exists: $TAG"
exit 0
fi
STATUS_PATH="$WORK_ROOT/notary-status.json"
rm -rf "$WORK_ROOT"
mkdir -p "$RELEASE_DIR"
notary_info_succeeded=false
for attempt in 1 2 3; do
if xcrun notarytool info "$SUBMISSION_ID" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" \
--output-format json > "$STATUS_PATH"; then
notary_info_succeeded=true
break
fi
echo "Notary status check failed ($attempt/3); retrying"
sleep 15
done
if [[ "$notary_info_succeeded" != true ]]; then
echo "Unable to query Apple notarization status after three attempts" >&2
exit 1
fi
NOTARY_STATUS=$(plutil -extract status raw -o - "$STATUS_PATH")
echo "Notarization status for $SUBMISSION_ID: $NOTARY_STATUS"
case "$NOTARY_STATUS" in
"In Progress")
if [[ -n "${GITHUB_STEP_SUMMARY:-}" ]]; then
printf 'Apple is still processing Meetingnotes %s. Submission: `%s`. The scheduled workflow will check again.\n' \
"$VERSION" "$SUBMISSION_ID" >> "$GITHUB_STEP_SUMMARY"
fi
exit 0
;;
Accepted)
;;
Invalid|Rejected)
xcrun notarytool log "$SUBMISSION_ID" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" || true
exit 1
;;
*)
echo "Unexpected notarization status: $NOTARY_STATUS" >&2
exit 1
;;
esac
ditto -x -k "$PRE_NOTARY_ZIP" "$WORK_ROOT"
if [[ ! -d "$APP_PATH" ]]; then
echo "Signed app was not found after extracting $PRE_NOTARY_ZIP" >&2
exit 1
fi
APP_VERSION=$(plutil -extract CFBundleShortVersionString raw -o - "$APP_PATH/Contents/Info.plist")
if [[ "$APP_VERSION" != "$VERSION" ]]; then
echo "Signed app version $APP_VERSION does not match release version $VERSION" >&2
exit 1
fi
xcrun stapler staple "$APP_PATH"
xcrun stapler validate "$APP_PATH"
spctl --assess --type execute --verbose=2 "$APP_PATH"
ARCHIVE_NAME="$APP_NAME-$VERSION.zip"
ARCHIVE_PATH="$RELEASE_DIR/$ARCHIVE_NAME"
ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
GENERATE_APPCAST="$PENDING_DIR/generate_appcast"
chmod +x "$GENERATE_APPCAST"
DOWNLOAD_URL="https://github.com/$GITHUB_REPOSITORY/releases/download/$TAG/"
printf '%s' "$SPARKLE_PRIVATE_KEY" | "$GENERATE_APPCAST" "$RELEASE_DIR" \
--ed-key-file - \
--download-url-prefix "$DOWNLOAD_URL" \
--maximum-deltas 0 \
-o "$RELEASE_DIR/appcast.xml"
grep -q "$DOWNLOAD_URL$ARCHIVE_NAME" "$RELEASE_DIR/appcast.xml"
grep -q 'sparkle:edSignature=' "$RELEASE_DIR/appcast.xml"
gh release \
create "$TAG" \
"$ARCHIVE_PATH" \
"$RELEASE_DIR/appcast.xml" \
--repo "$GITHUB_REPOSITORY" \
--target "$COMMIT_SHA" \
--title "Meetingnotes $VERSION" \
--generate-notes
echo "Published Meetingnotes $VERSION"
+38 -6
View File
@@ -14,8 +14,13 @@ APP_PATH="$DERIVED_DATA/Build/Products/Release/$APP_NAME.app"
required_variables=(
VERSION
SIGNING_IDENTITY
SIGNING_KEYCHAIN
APPLE_ID
APPLE_TEAM_ID
APPLE_APP_PASSWORD
SPARKLE_PRIVATE_KEY
GITHUB_REPOSITORY
RELEASE_BASE_URL
)
for variable in "${required_variables[@]}"; do
@@ -54,12 +59,12 @@ SPARKLE_FRAMEWORK="$APP_PATH/Contents/Frameworks/Sparkle.framework"
SPARKLE_CONTENTS="$SPARKLE_FRAMEWORK/Versions/B"
sign_component() {
codesign --force --timestamp --options runtime --sign "$SIGNING_IDENTITY" "$1"
codesign --force --timestamp --options runtime --keychain "$SIGNING_KEYCHAIN" --sign "$SIGNING_IDENTITY" "$1"
}
sign_component "$SPARKLE_CONTENTS/XPCServices/Installer.xpc"
if [[ -d "$SPARKLE_CONTENTS/XPCServices/Downloader.xpc" ]]; then
codesign --force --timestamp --options runtime \
codesign --force --timestamp --options runtime --keychain "$SIGNING_KEYCHAIN" \
--preserve-metadata=entitlements \
--sign "$SIGNING_IDENTITY" \
"$SPARKLE_CONTENTS/XPCServices/Downloader.xpc"
@@ -68,7 +73,7 @@ sign_component "$SPARKLE_CONTENTS/Autoupdate"
sign_component "$SPARKLE_CONTENTS/Updater.app"
sign_component "$SPARKLE_FRAMEWORK"
codesign --force --timestamp --options runtime \
codesign --force --timestamp --options runtime --keychain "$SIGNING_KEYCHAIN" \
--entitlements meetingnotes/meetingnotes.entitlements \
--sign "$SIGNING_IDENTITY" \
"$APP_PATH"
@@ -80,13 +85,40 @@ ARCHIVE_NAME="$APP_NAME-$VERSION.zip"
ARCHIVE_PATH="$RELEASE_DIR/$ARCHIVE_NAME"
ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
NOTARY_RESULT="$BUILD_ROOT/notary-result.json"
xcrun notarytool submit "$ARCHIVE_PATH" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" \
--wait \
--timeout 20m \
--output-format json > "$NOTARY_RESULT"
NOTARY_STATUS=$(plutil -extract status raw -o - "$NOTARY_RESULT")
if [[ "$NOTARY_STATUS" != "Accepted" ]]; then
submission_id=$(plutil -extract id raw -o - "$NOTARY_RESULT")
xcrun notarytool log "$submission_id" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" || true
echo "Apple notarization failed with status: $NOTARY_STATUS" >&2
exit 1
fi
xcrun stapler staple "$APP_PATH"
xcrun stapler validate "$APP_PATH"
spctl --assess --type execute --verbose=2 "$APP_PATH"
rm -f "$ARCHIVE_PATH"
ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
GENERATE_APPCAST=$(find "$DERIVED_DATA/SourcePackages/artifacts" -type f -name generate_appcast -print -quit)
if [[ -z "$GENERATE_APPCAST" ]]; then
echo "Sparkle generate_appcast tool was not found" >&2
exit 1
fi
DOWNLOAD_URL="https://github.com/$GITHUB_REPOSITORY/releases/download/v$VERSION/"
DOWNLOAD_URL="$RELEASE_BASE_URL/releases/download/v$VERSION/"
printf '%s' "$SPARKLE_PRIVATE_KEY" | "$GENERATE_APPCAST" "$RELEASE_DIR" \
--ed-key-file - \
--download-url-prefix "$DOWNLOAD_URL" \
@@ -97,8 +129,8 @@ grep -q "$DOWNLOAD_URL$ARCHIVE_NAME" "$RELEASE_DIR/appcast.xml"
grep -q 'sparkle:edSignature=' "$RELEASE_DIR/appcast.xml"
if [[ -n "${GITHUB_STEP_SUMMARY:-}" ]]; then
printf 'Built and Developer ID-signed Meetingnotes %s. The GitHub release is ready to publish.\n' \
printf 'Built, Developer ID-signed, notarized, and stapled Meetingnotes %s. The Gitea release is ready to publish.\n' \
"$VERSION" >> "$GITHUB_STEP_SUMMARY"
fi
echo "Signed release artifacts are ready in $RELEASE_DIR"
echo "Signed and notarized release artifacts are ready in $RELEASE_DIR"
+45
View File
@@ -0,0 +1,45 @@
#!/usr/bin/env python3
"""Publish both signed assets together, keeping incomplete uploads as a draft."""
import json
import os
import urllib.error
import urllib.parse
import urllib.request
from pathlib import Path
base = os.environ["GITEA_SERVER_URL"].rstrip("/") + "/api/v1/repos/" + os.environ["GITHUB_REPOSITORY"]
token = os.environ["GITEA_TOKEN"]
version = os.environ["VERSION"]
class NoRedirect(urllib.request.HTTPRedirectHandler):
def redirect_request(self, *args, **kwargs):
return None
def request(path, method="GET", data=None, binary=False):
body = data if binary else None if data is None else json.dumps(data).encode()
req = urllib.request.Request(base + path, method=method, data=body, headers={
"Authorization": "token " + token,
"Content-Type": "application/octet-stream" if binary else "application/json",
"User-Agent": "Meetingnotes-release",
})
with urllib.request.build_opener(NoRedirect()).open(req, timeout=120) as response:
raw = response.read()
return json.loads(raw) if raw else None
release_dir = Path(os.environ["RUNNER_TEMP"]) / "meetingnotes-release/release"
assets = [release_dir / f"Meetingnotes-{version}.zip", release_dir / "appcast.xml"]
for asset in assets:
if not asset.is_file() or not asset.stat().st_size:
raise RuntimeError(f"Missing release artifact: {asset.name}")
release = request("/releases", "POST", {
"tag_name": "v" + version, "target_commitish": os.environ["GITHUB_SHA"],
"name": "Meetingnotes " + version, "draft": True, "prerelease": False,
"body": "Developer ID signed, Apple notarized, and signed for Sparkle updates.\n\nBuilt from main at `" + os.environ["GITHUB_SHA"] + "`.",
})
for asset in assets:
request(f"/releases/{release['id']}/assets?name=" + urllib.parse.quote(asset.name), "POST", asset.read_bytes(), binary=True)
request(f"/releases/{release['id']}", "PATCH", {"draft": False})
print("Published Meetingnotes " + version)
+77
View File
@@ -0,0 +1,77 @@
#!/usr/bin/env python3
"""Check the CI app's unauthenticated API without changing saved preferences."""
import json
import plistlib
import socket
import subprocess
import sys
import tempfile
import time
import urllib.error
import urllib.request
from pathlib import Path
# Launch from the Mac's internal volume; the runner checkout is on removable storage.
cache = Path.home() / "Library/Caches/meetingnotes-ci"
cache.mkdir(parents=True, exist_ok=True)
with tempfile.TemporaryDirectory(prefix="smoke-", dir=cache) as directory:
staged = Path(directory) / "Meetingnotes.app"
subprocess.run(["ditto", sys.argv[1], str(staged)], check=True)
app = staged / "Contents/MacOS/Meetingnotes"
with (staged / "Contents/Info.plist").open("rb") as info_file:
bundle_id = plistlib.load(info_file)["CFBundleIdentifier"]
assert bundle_id == "net.jamesbone.meetingnotes.ci", "Smoke tests require the isolated CI app"
def clear_test_token():
result = subprocess.run(
["security", "delete-generic-password", "-s", bundle_id, "-a", "muteDeckAPIToken"],
capture_output=True, timeout=10,
)
if result.returncode not in (0, 44): # 44: item does not exist
raise RuntimeError("Could not clear the CI Keychain token")
clear_test_token()
with socket.socket() as listener:
listener.bind(("127.0.0.1", 0))
port = listener.getsockname()[1]
with tempfile.TemporaryFile() as log:
process = subprocess.Popen(
[str(app), "-muteDeckAPIEnabled", "YES", "-muteDeckAPIPort", str(port), "-hasCompletedOnboarding", "YES", "-hasAcceptedTerms", "YES", "-SUEnableAutomaticChecks", "NO"],
stdout=log, stderr=subprocess.STDOUT,
)
try:
for attempt in range(30):
if process.poll() is not None:
raise RuntimeError("CI app exited before the API became ready")
try:
with urllib.request.urlopen(f"http://127.0.0.1:{port}/api/info", timeout=2) as response:
info = json.load(response)
break
except (urllib.error.URLError, TimeoutError):
time.sleep(1)
else:
raise RuntimeError("CI API did not become ready")
assert info["name"] == "MeetingDebrief", "Unexpected API identity"
try:
urllib.request.urlopen(f"http://127.0.0.1:{port}/api/recording/status", timeout=2)
except urllib.error.HTTPError as error:
assert error.code == 401, f"Unexpected status: {error.code}"
else:
raise RuntimeError("Recording status allowed an unauthenticated request")
print("Local API readiness and authentication checks passed")
except Exception:
if process.poll() is None:
sample = subprocess.run(["sample", str(process.pid), "1", "1"], capture_output=True, text=True, timeout=10)
print(sample.stdout.split("Binary Images:")[0], file=sys.stderr)
log.seek(0)
print(log.read().decode(errors="replace")[-8000:], file=sys.stderr)
raise
finally:
if process.poll() is None:
process.terminate()
try:
process.wait(timeout=10)
except subprocess.TimeoutExpired:
process.kill()
process.wait()
clear_test_token()
+43
View File
@@ -0,0 +1,43 @@
#!/usr/bin/env python3
"""Import the CI certificate into an isolated keychain for one release command."""
import base64
import os
import re
import secrets
import subprocess
import sys
import tempfile
from pathlib import Path
def security(*args):
result = subprocess.run(["security", *args], capture_output=True, text=True)
if result.returncode:
raise RuntimeError(f"security {args[0]} failed (output withheld to protect credentials)")
return result.stdout
with tempfile.TemporaryDirectory(prefix="meetingnotes-signing-", dir=os.environ["RUNNER_TEMP"]) as directory:
keychain = str(Path(directory) / "release.keychain-db")
certificate = Path(directory) / "certificate.p12"
certificate.write_bytes(base64.b64decode(os.environ["APPLE_CERTIFICATE_P12"]))
certificate.chmod(0o600)
password = secrets.token_urlsafe(32)
created = False
try:
security("create-keychain", "-p", password, keychain)
created = True
security("set-keychain-settings", "-lut", "21600", keychain)
security("unlock-keychain", "-p", password, keychain)
security("import", str(certificate), "-k", keychain, "-P", os.environ["APPLE_CERTIFICATE_PASSWORD"], "-T", "/usr/bin/codesign", "-T", "/usr/bin/security")
security("set-key-partition-list", "-S", "apple-tool:,apple:", "-s", "-k", password, keychain)
identities = security("find-identity", "-v", "-p", "codesigning", keychain)
match = re.search(r'([0-9A-F]{40}) "Developer ID Application:', identities)
if not match:
raise RuntimeError("The certificate contains no valid Developer ID Application identity")
environment = dict(os.environ, SIGNING_IDENTITY=match[1], SIGNING_KEYCHAIN=keychain)
status = subprocess.run(sys.argv[1:], env=environment).returncode
finally:
if created:
security("delete-keychain", keychain)
sys.exit(status)