chore: vendor swabble and add speech usage strings

2025-12-06 02:10:20 +01:00
parent 4e7d905783
commit e1c9885566
34 changed files with 1577 additions and 1 deletions
--- a/Swabble/Sources/SwabbleCore/Config/Config.swift
+++ b/Swabble/Sources/SwabbleCore/Config/Config.swift
@@ -0,0 +1,77 @@
+import Foundation
+
+public struct SwabbleConfig: Codable, Sendable {
+    public struct Audio: Codable, Sendable {
+        public var deviceName: String = ""
+        public var deviceIndex: Int = -1
+        public var sampleRate: Double = 16000
+        public var channels: Int = 1
+    }
+
+    public struct Wake: Codable, Sendable {
+        public var enabled: Bool = true
+        public var word: String = "clawd"
+        public var aliases: [String] = ["claude"]
+    }
+
+    public struct Hook: Codable, Sendable {
+        public var command: String = ""
+        public var args: [String] = []
+        public var prefix: String = "Voice swabble from ${hostname}: "
+        public var cooldownSeconds: Double = 1
+        public var minCharacters: Int = 24
+        public var timeoutSeconds: Double = 5
+        public var env: [String: String] = [:]
+    }
+
+    public struct Logging: Codable, Sendable {
+        public var level: String = "info"
+        public var format: String = "text" // text|json placeholder
+    }
+
+    public struct Transcripts: Codable, Sendable {
+        public var enabled: Bool = true
+        public var maxEntries: Int = 50
+    }
+
+    public struct Speech: Codable, Sendable {
+        public var localeIdentifier: String = Locale.current.identifier
+        public var etiquetteReplacements: Bool = false
+    }
+
+    public var audio = Audio()
+    public var wake = Wake()
+    public var hook = Hook()
+    public var logging = Logging()
+    public var transcripts = Transcripts()
+    public var speech = Speech()
+
+    public static let defaultPath = FileManager.default
+        .homeDirectoryForCurrentUser
+        .appendingPathComponent(".config/swabble/config.json")
+
+    public init() {}
+}
+
+public enum ConfigError: Error {
+    case missingConfig
+}
+
+public enum ConfigLoader {
+    public static func load(at path: URL?) throws -> SwabbleConfig {
+        let url = path ?? SwabbleConfig.defaultPath
+        if !FileManager.default.fileExists(atPath: url.path) {
+            throw ConfigError.missingConfig
+        }
+        let data = try Data(contentsOf: url)
+        return try JSONDecoder().decode(SwabbleConfig.self, from: data)
+    }
+
+    public static func save(_ config: SwabbleConfig, at path: URL?) throws {
+        let url = path ?? SwabbleConfig.defaultPath
+        let dir = url.deletingLastPathComponent()
+        try FileManager.default.createDirectory(at: dir, withIntermediateDirectories: true)
+        let data = try JSONEncoder().encode(config)
+        try data.write(to: url)
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Hooks/HookRunner.swift
+++ b/Swabble/Sources/SwabbleCore/Hooks/HookRunner.swift
@@ -0,0 +1,75 @@
+import Foundation
+
+public struct HookJob: Sendable {
+    public let text: String
+    public let timestamp: Date
+
+    public init(text: String, timestamp: Date) {
+        self.text = text
+        self.timestamp = timestamp
+    }
+}
+
+public actor HookRunner {
+    private let config: SwabbleConfig
+    private var lastRun: Date?
+    private let hostname: String
+
+    public init(config: SwabbleConfig) {
+        self.config = config
+        self.hostname = Host.current().localizedName ?? "host"
+    }
+
+    public func shouldRun() -> Bool {
+        guard self.config.hook.cooldownSeconds > 0 else { return true }
+        if let lastRun, Date().timeIntervalSince(lastRun) < config.hook.cooldownSeconds {
+            return false
+        }
+        return true
+    }
+
+    public func run(job: HookJob) async throws {
+        guard self.shouldRun() else { return }
+        guard !self.config.hook.command.isEmpty else { throw NSError(
+            domain: "Hook",
+            code: 1,
+            userInfo: [NSLocalizedDescriptionKey: "hook command not set"]) }
+
+        let prefix = self.config.hook.prefix.replacingOccurrences(of: "${hostname}", with: self.hostname)
+        let payload = prefix + job.text
+
+        let process = Process()
+        process.executableURL = URL(fileURLWithPath: self.config.hook.command)
+        process.arguments = self.config.hook.args + [payload]
+
+        var env = ProcessInfo.processInfo.environment
+        env["SWABBLE_TEXT"] = job.text
+        env["SWABBLE_PREFIX"] = prefix
+        for (k, v) in self.config.hook.env {
+            env[k] = v
+        }
+        process.environment = env
+
+        let pipe = Pipe()
+        process.standardOutput = pipe
+        process.standardError = pipe
+
+        try process.run()
+
+        let timeoutNanos = UInt64(max(config.hook.timeoutSeconds, 0.1) * 1_000_000_000)
+        try await withThrowingTaskGroup(of: Void.self) { group in
+            group.addTask {
+                process.waitUntilExit()
+            }
+            group.addTask {
+                try await Task.sleep(nanoseconds: timeoutNanos)
+                if process.isRunning {
+                    process.terminate()
+                }
+            }
+            try await group.next()
+            group.cancelAll()
+        }
+        self.lastRun = Date()
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Speech/BufferConverter.swift
+++ b/Swabble/Sources/SwabbleCore/Speech/BufferConverter.swift
@@ -0,0 +1,50 @@
+@preconcurrency import AVFoundation
+import Foundation
+
+final class BufferConverter {
+    private final class Box<T>: @unchecked Sendable { var value: T; init(_ value: T) { self.value = value } }
+    enum ConverterError: Swift.Error {
+        case failedToCreateConverter
+        case failedToCreateConversionBuffer
+        case conversionFailed(NSError?)
+    }
+
+    private var converter: AVAudioConverter?
+
+    func convert(_ buffer: AVAudioPCMBuffer, to format: AVAudioFormat) throws -> AVAudioPCMBuffer {
+        let inputFormat = buffer.format
+        if inputFormat == format {
+            return buffer
+        }
+        if converter == nil || converter?.outputFormat != format {
+            converter = AVAudioConverter(from: inputFormat, to: format)
+            converter?.primeMethod = .none
+        }
+        guard let converter else { throw ConverterError.failedToCreateConverter }
+
+        let sampleRateRatio = converter.outputFormat.sampleRate / converter.inputFormat.sampleRate
+        let scaledInputFrameLength = Double(buffer.frameLength) * sampleRateRatio
+        let frameCapacity = AVAudioFrameCount(scaledInputFrameLength.rounded(.up))
+        guard let conversionBuffer = AVAudioPCMBuffer(pcmFormat: converter.outputFormat, frameCapacity: frameCapacity)
+        else {
+            throw ConverterError.failedToCreateConversionBuffer
+        }
+
+        var nsError: NSError?
+        let consumed = Box(false)
+        let inputBuffer = buffer
+        let status = converter.convert(to: conversionBuffer, error: &nsError) { _, statusPtr in
+            if consumed.value {
+                statusPtr.pointee = .noDataNow
+                return nil
+            }
+            consumed.value = true
+            statusPtr.pointee = .haveData
+            return inputBuffer
+        }
+        if status == .error {
+            throw ConverterError.conversionFailed(nsError)
+        }
+        return conversionBuffer
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Speech/SpeechPipeline.swift
+++ b/Swabble/Sources/SwabbleCore/Speech/SpeechPipeline.swift
@@ -0,0 +1,111 @@
+import AVFoundation
+import Foundation
+import Speech
+
+public struct SpeechSegment: Sendable {
+    public let text: String
+    public let isFinal: Bool
+}
+
+public enum SpeechPipelineError: Error {
+    case authorizationDenied
+    case analyzerFormatUnavailable
+    case transcriberUnavailable
+}
+
+/// Live microphone → SpeechAnalyzer → SpeechTranscriber pipeline.
+public actor SpeechPipeline {
+    private struct UnsafeBuffer: @unchecked Sendable { let buffer: AVAudioPCMBuffer }
+
+    private var engine = AVAudioEngine()
+    private var transcriber: SpeechTranscriber?
+    private var analyzer: SpeechAnalyzer?
+    private var inputContinuation: AsyncStream<AnalyzerInput>.Continuation?
+    private var resultTask: Task<Void, Never>?
+    private let converter = BufferConverter()
+
+    public init() {}
+
+    public func start(localeIdentifier: String, etiquette: Bool) async throws -> AsyncStream<SpeechSegment> {
+        let auth = await requestAuthorizationIfNeeded()
+        guard auth == .authorized else { throw SpeechPipelineError.authorizationDenied }
+
+        let transcriberModule = SpeechTranscriber(
+            locale: Locale(identifier: localeIdentifier),
+            transcriptionOptions: etiquette ? [.etiquetteReplacements] : [],
+            reportingOptions: [.volatileResults],
+            attributeOptions: [])
+        self.transcriber = transcriberModule
+
+        guard let analyzerFormat = await SpeechAnalyzer.bestAvailableAudioFormat(compatibleWith: [transcriberModule])
+        else {
+            throw SpeechPipelineError.analyzerFormatUnavailable
+        }
+
+        self.analyzer = SpeechAnalyzer(modules: [transcriberModule])
+        let (stream, continuation) = AsyncStream<AnalyzerInput>.makeStream()
+        self.inputContinuation = continuation
+
+        let inputNode = self.engine.inputNode
+        let inputFormat = inputNode.outputFormat(forBus: 0)
+        inputNode.removeTap(onBus: 0)
+        inputNode.installTap(onBus: 0, bufferSize: 2048, format: inputFormat) { [weak self] buffer, _ in
+            guard let self else { return }
+            let boxed = UnsafeBuffer(buffer: buffer)
+            Task { await self.handleBuffer(boxed.buffer, targetFormat: analyzerFormat) }
+        }
+
+        self.engine.prepare()
+        try self.engine.start()
+        try await self.analyzer?.start(inputSequence: stream)
+
+        guard let transcriberForStream = self.transcriber else {
+            throw SpeechPipelineError.transcriberUnavailable
+        }
+
+        return AsyncStream { continuation in
+            self.resultTask = Task {
+                do {
+                    for try await result in transcriberForStream.results {
+                        let seg = SpeechSegment(text: String(result.text.characters), isFinal: result.isFinal)
+                        continuation.yield(seg)
+                    }
+                } catch {
+                    // swallow errors and finish
+                }
+                continuation.finish()
+            }
+            continuation.onTermination = { _ in
+                Task { await self.stop() }
+            }
+        }
+    }
+
+    public func stop() async {
+        self.resultTask?.cancel()
+        self.inputContinuation?.finish()
+        self.engine.inputNode.removeTap(onBus: 0)
+        self.engine.stop()
+        try? await self.analyzer?.finalizeAndFinishThroughEndOfInput()
+    }
+
+    private func handleBuffer(_ buffer: AVAudioPCMBuffer, targetFormat: AVAudioFormat) async {
+        do {
+            let converted = try converter.convert(buffer, to: targetFormat)
+            let input = AnalyzerInput(buffer: converted)
+            self.inputContinuation?.yield(input)
+        } catch {
+            // drop on conversion failure
+        }
+    }
+
+    private func requestAuthorizationIfNeeded() async -> SFSpeechRecognizerAuthorizationStatus {
+        let current = SFSpeechRecognizer.authorizationStatus()
+        guard current == .notDetermined else { return current }
+        return await withCheckedContinuation { continuation in
+            SFSpeechRecognizer.requestAuthorization { status in
+                continuation.resume(returning: status)
+            }
+        }
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Support/AttributedString+Sentences.swift
+++ b/Swabble/Sources/SwabbleCore/Support/AttributedString+Sentences.swift
@@ -0,0 +1,63 @@
+import CoreMedia
+import Foundation
+import NaturalLanguage
+
+extension AttributedString {
+    public func sentences(maxLength: Int? = nil) -> [AttributedString] {
+        let tokenizer = NLTokenizer(unit: .sentence)
+        let string = String(characters)
+        tokenizer.string = string
+        let sentenceRanges = tokenizer.tokens(for: string.startIndex..<string.endIndex).map {
+            (
+                $0,
+                AttributedString.Index($0.lowerBound, within: self)!
+                    ..<
+                    AttributedString.Index($0.upperBound, within: self)!)
+        }
+        let ranges = sentenceRanges.flatMap { sentenceStringRange, sentenceRange in
+            let sentence = self[sentenceRange]
+            guard let maxLength, sentence.characters.count > maxLength else {
+                return [sentenceRange]
+            }
+
+            let wordTokenizer = NLTokenizer(unit: .word)
+            wordTokenizer.string = string
+            var wordRanges = wordTokenizer.tokens(for: sentenceStringRange).map {
+                AttributedString.Index($0.lowerBound, within: self)!
+                    ..<
+                    AttributedString.Index($0.upperBound, within: self)!
+            }
+            guard !wordRanges.isEmpty else { return [sentenceRange] }
+            wordRanges[0] = sentenceRange.lowerBound..<wordRanges[0].upperBound
+            wordRanges[wordRanges.count - 1] = wordRanges[wordRanges.count - 1].lowerBound..<sentenceRange.upperBound
+
+            var ranges: [Range<AttributedString.Index>] = []
+            for wordRange in wordRanges {
+                if let lastRange = ranges.last,
+                   self[lastRange].characters.count + self[wordRange].characters.count <= maxLength
+                {
+                    ranges[ranges.count - 1] = lastRange.lowerBound..<wordRange.upperBound
+                } else {
+                    ranges.append(wordRange)
+                }
+            }
+
+            return ranges
+        }
+
+        return ranges.compactMap { range in
+            let audioTimeRanges = self[range].runs.filter {
+                !String(self[$0.range].characters)
+                    .trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+            }.compactMap(\.audioTimeRange)
+            guard !audioTimeRanges.isEmpty else { return nil }
+            let start = audioTimeRanges.first!.start
+            let end = audioTimeRanges.last!.end
+            var attributes = AttributeContainer()
+            attributes[AttributeScopes.SpeechAttributes.TimeRangeAttribute.self] = CMTimeRange(
+                start: start,
+                end: end)
+            return AttributedString(self[range].characters, attributes: attributes)
+        }
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Support/Logging.swift
+++ b/Swabble/Sources/SwabbleCore/Support/Logging.swift
@@ -0,0 +1,41 @@
+import Foundation
+
+public enum LogLevel: String, Comparable, CaseIterable, Sendable {
+    case trace, debug, info, warn, error
+
+    var rank: Int {
+        switch self {
+        case .trace: 0
+        case .debug: 1
+        case .info: 2
+        case .warn: 3
+        case .error: 4
+        }
+    }
+
+    public static func < (lhs: LogLevel, rhs: LogLevel) -> Bool { lhs.rank < rhs.rank }
+}
+
+public struct Logger: Sendable {
+    public let level: LogLevel
+
+    public init(level: LogLevel) { self.level = level }
+
+    public func log(_ level: LogLevel, _ message: String) {
+        guard level >= self.level else { return }
+        let ts = ISO8601DateFormatter().string(from: Date())
+        print("[\(level.rawValue.uppercased())] \(ts) | \(message)")
+    }
+
+    public func trace(_ msg: String) { self.log(.trace, msg) }
+    public func debug(_ msg: String) { self.log(.debug, msg) }
+    public func info(_ msg: String) { self.log(.info, msg) }
+    public func warn(_ msg: String) { self.log(.warn, msg) }
+    public func error(_ msg: String) { self.log(.error, msg) }
+}
+
+extension LogLevel {
+    public init?(configValue: String) {
+        self.init(rawValue: configValue.lowercased())
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Support/OutputFormat.swift
+++ b/Swabble/Sources/SwabbleCore/Support/OutputFormat.swift
@@ -0,0 +1,45 @@
+import CoreMedia
+import Foundation
+
+public enum OutputFormat: String {
+    case txt
+    case srt
+
+    public var needsAudioTimeRange: Bool {
+        switch self {
+        case .srt: true
+        default: false
+        }
+    }
+
+    public func text(for transcript: AttributedString, maxLength: Int) -> String {
+        switch self {
+        case .txt:
+            return String(transcript.characters)
+        case .srt:
+            func format(_ timeInterval: TimeInterval) -> String {
+                let ms = Int(timeInterval.truncatingRemainder(dividingBy: 1) * 1000)
+                let s = Int(timeInterval) % 60
+                let m = (Int(timeInterval) / 60) % 60
+                let h = Int(timeInterval) / 60 / 60
+                return String(format: "%0.2d:%0.2d:%0.2d,%0.3d", h, m, s, ms)
+            }
+
+            return transcript.sentences(maxLength: maxLength).compactMap { (sentence: AttributedString) -> (
+                CMTimeRange,
+                String)? in
+                guard let timeRange = sentence.audioTimeRange else { return nil }
+                return (timeRange, String(sentence.characters))
+            }.enumerated().map { index, run in
+                let (timeRange, text) = run
+                return """
+
+                \(index + 1)
+                \(format(timeRange.start.seconds)) --> \(format(timeRange.end.seconds))
+                \(text.trimmingCharacters(in: .whitespacesAndNewlines))
+
+                """
+            }.joined().trimmingCharacters(in: .whitespacesAndNewlines)
+        }
+    }
+}
--- a/Swabble/Sources/SwabbleCore/Support/TranscriptsStore.swift
+++ b/Swabble/Sources/SwabbleCore/Support/TranscriptsStore.swift
@@ -0,0 +1,46 @@
+import Foundation
+
+public actor TranscriptsStore {
+    public static let shared = TranscriptsStore()
+
+    private var entries: [String] = []
+    private let limit = 100
+    private let fileURL: URL
+
+    public init() {
+        let dir = FileManager.default.homeDirectoryForCurrentUser
+            .appendingPathComponent("Library/Application Support/swabble", isDirectory: true)
+        try? FileManager.default.createDirectory(at: dir, withIntermediateDirectories: true)
+        self.fileURL = dir.appendingPathComponent("transcripts.log")
+        if let data = try? Data(contentsOf: fileURL),
+           let text = String(data: data, encoding: .utf8)
+        {
+            self.entries = text.split(separator: "\n").map(String.init).suffix(self.limit)
+        }
+    }
+
+    public func append(text: String) {
+        self.entries.append(text)
+        if self.entries.count > self.limit {
+            self.entries.removeFirst(self.entries.count - self.limit)
+        }
+        let body = self.entries.joined(separator: "\n")
+        try? body.write(to: self.fileURL, atomically: false, encoding: .utf8)
+    }
+
+    public func latest() -> [String] { self.entries }
+}
+
+extension String {
+    private func appendLine(to url: URL) throws {
+        let data = (self + "\n").data(using: .utf8) ?? Data()
+        if FileManager.default.fileExists(atPath: url.path) {
+            let handle = try FileHandle(forWritingTo: url)
+            try handle.seekToEnd()
+            try handle.write(contentsOf: data)
+            try handle.close()
+        } else {
+            try data.write(to: url)
+        }
+    }
+}