chore: vendor swabble and add speech usage strings

This commit is contained in:
Peter Steinberger
2025-12-06 02:10:20 +01:00
parent 4e7d905783
commit e1c9885566
34 changed files with 1577 additions and 1 deletions

View File

@@ -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)
}
}
}

View File

@@ -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())
}
}

View File

@@ -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)
}
}
}

View File

@@ -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)
}
}
}