echogarden 1.1.1 → 1.2.1

package/src/cli/CLIConfigFile.ts

@@ -56,10 +56,10 @@ export async function parseJSONConfigFile(path: string): Promise<ParsedConfigFil
 
 	const result: ParsedConfigFile = new Map()
 
-	for (const command in parsedJSON) {
-		const commandObj = parsedJSON[command]
+	for (const operation in parsedJSON) {
+		const operationObj = parsedJSON[operation]
 
-		const commandMap = new Map<string, string>()
+		const operationMap = new Map<string, string>()
 
 		function addFromObject(obj: any, pathPrefix: string): void {
 			for (const propertyName in obj) {
@@ -70,19 +70,19 @@ export async function parseJSONConfigFile(path: string): Promise<ParsedConfigFil
 
 				if (typeof propertyValue == 'object') {
 					if (Array.isArray(propertyValue)) {
-						commandMap.set(propertyPath, JSON.stringify(propertyValue))
+						operationMap.set(propertyPath, JSON.stringify(propertyValue))
 					} else {
 						addFromObject(propertyValue, propertyPath)
 					}
 				} else {
-					commandMap.set(propertyPath, propertyValue)
+					operationMap.set(propertyPath, propertyValue)
 				}
 			}
 		}
 
-		addFromObject(commandObj, '')
+		addFromObject(operationObj, '')
 
-		result.set(command, commandMap)
+		result.set(operation, operationMap)
 	}
 
 	return result

package/src/cli/CLIOptions.ts

@@ -0,0 +1,8 @@
+export interface CLIOptions {
+	play?: boolean
+	overwrite?: boolean
+	debug?: boolean
+	config?: string
+}
+
+export const CLIOptionsKeys: (keyof CLIOptions)[] = ['play', 'overwrite', 'debug', 'config']

package/src/cli/CLIParser.ts

@@ -1,8 +1,7 @@
-export function parseCLIArguments(command: string, args: string[]): CLIArguments {
-	const parsedArgs: CLIArguments = {
-		command,
-		commandArgs: [],
-		options: new Map(),
+export function parseCLIArguments(args: string[]): ParsedCLIArguments {
+	const parsedArgs: ParsedCLIArguments = {
+		operationArgs: [],
+		parsedArgumentsLookup: new Map(),
 	}
 
 	for (let i = 0; i < args.length; i++) {
@@ -16,26 +15,30 @@ export function parseCLIArguments(command: string, args: string[]): CLIArguments
 
 			const optionText = currentArg.substring(2)
 
-			const indexOfEquals = optionText.indexOf('=')
-			if (indexOfEquals == 0) {
+			const indexOfEqualSign = optionText.indexOf('=')
+			if (indexOfEqualSign == 0) {
 				throw new Error('An option cannot have an empty name.')
-			} else if (indexOfEquals == -1) {
-				parsedArgs.options.set(optionText, '')
+			} else if (indexOfEqualSign == -1) {
+				if (optionText.startsWith('no-')) {
+					parsedArgs.parsedArgumentsLookup.set(optionText.substring(3), 'false')
+				} else {
+					parsedArgs.parsedArgumentsLookup.set(optionText, '')
+				}
 			} else {
-				const key = optionText.substring(0, indexOfEquals)
-				const value = optionText.substring(indexOfEquals + 1)
-				parsedArgs.options.set(key, value)
+				const key = optionText.substring(0, indexOfEqualSign)
+				const value = optionText.substring(indexOfEqualSign + 1)
+
+				parsedArgs.parsedArgumentsLookup.set(key, value)
 			}
 		} else {
-			parsedArgs.commandArgs.push(currentArg)
+			parsedArgs.operationArgs.push(currentArg)
 		}
 	}
 
 	return parsedArgs
 }
 
-export type CLIArguments = {
-	command: string
-	commandArgs: string[]
-	options: Map<string, string>
+export interface ParsedCLIArguments {
+	operationArgs: string[]
+	parsedArgumentsLookup: Map<string, string>
 }

package/src/recognition/WhisperSTT.ts

@@ -35,7 +35,7 @@ export async function recognize(
 	options = extendDeep(defaultWhisperOptions, options)
 
 	if (sourceRawAudio.sampleRate != 16000) {
-		throw new Error('Source audio must have a sampling rate of 16000')
+		throw new Error('Source audio must have a sampling rate of 16000 Hz')
 	}
 
 	sourceLanguage = getShortLanguageCode(sourceLanguage)
@@ -716,7 +716,8 @@ export class Whisper {
 
 		// Start decoding loop
 		for (let decodedTokenCount = 0; decodedTokenCount < options.maxTokensPerPart!; decodedTokenCount++) {
-			const isInitialState = decodedTokens.length == initialTokens.length
+			const isInitialState = decodedTokens.length === initialTokens.length
+			const atLeastOneTextTokenDecoded = decodedTokens.slice(initialTokens.length).some(token => this.isTextToken(token))
 
 			// If not in initial state, reshape KV Cache tensor to accomodate a new output token
 			if (!isInitialState) {
@@ -759,7 +760,7 @@ export class Whisper {
 			const crossAttentionQKsForToken = makeOnnxLikeFloat32Tensor(crossAttentionQKsForTokenOnnx)
 			crossAttentionQKsForTokenOnnx.dispose()
 
-			// Compute logits
+			// Get logits
 			const resultLogitsFloatArrays = splitFloat32Array(logitsBuffer, logitsBuffer.length / decoderOutputs['logits'].dims[1])
 			const allTokenLogits = Array.from(resultLogitsFloatArrays[resultLogitsFloatArrays.length - 1])
 
@@ -768,7 +769,7 @@ export class Whisper {
 				allTokenLogits[suppressedTokenIndex] = -Infinity
 			}
 
-			if (isInitialState) {
+			if (!atLeastOneTextTokenDecoded) {
 				// If in initial state, suppress end-of-text token
 				allTokenLogits[endOfTextToken] = -Infinity
 			}
@@ -783,13 +784,21 @@ export class Whisper {
 					return false
 				}
 
+				// If this is the first token in the part, unconditionally decode a timestamp token
+				// for time 0.0
+				if (isInitialState) {
+					addToken(timestampTokensStart, timestampTokenLogits, 1.0, crossAttentionQKsForToken)
+
+					return true
+				}
+
 				const previousTokenWasTimestamp = this.isTimestampToken(decodedTokens[decodedTokens.length - 1])
 				const secondPreviousTokenWasTimestamp = this.isTimestampToken(decodedTokens[decodedTokens.length - 2])
 
 				// If there are two successive timestamp tokens decoded, or the previous timestamp was the first token,
 				// don't decode a timestamp
 				if (previousTokenWasTimestamp &&
-					(decodedTokens.length === initialTokens.length + 1) || secondPreviousTokenWasTimestamp) {
+					((decodedTokens.length === initialTokens.length + 1) || secondPreviousTokenWasTimestamp)) {
 					return false
 				}
 
@@ -854,9 +863,9 @@ export class Whisper {
 
 			let shouldDecodeEndfOfTextToken = false
 
-			// If not in initial state, and the end-of-text token's probability is sufficiently higher than
-			// the second highest ranked token, then accept end-of-text
-			if (!isInitialState) {
+			// If at least one text token was decoded, and the end-of-text token's probability is
+			// sufficiently higher than the second highest ranked token, then accept end-of-text
+			if (atLeastOneTextTokenDecoded) {
 				const endOfTextTokenLogit = nonTimestampTokenLogits[endOfTextToken]
 
 				const otherTokensLogits = nonTimestampTokenLogits.slice()
@@ -931,8 +940,8 @@ export class Whisper {
 			if (options.suppressRepetition) {
 				// Using some hardcoded constants, for now
 				const tokenWindowSize = 30
-				const thresholdMatchLength = 6
-				const thresholdCycleRepetition = 2.0
+				const thresholdMatchLength = 4
+				const thresholdCycleRepetition = 3
 
 				const filteredCandidates: typeof topCandidates = []
 
@@ -1772,7 +1781,7 @@ export function normalizeWhisperModelName(modelName: WhisperModelName, languageC
 		modelName = modelName.slice(0, modelName.length - 3) as WhisperModelName
 
 		const logger = new Logger()
-		logger.logTitledMessage(`Warning`, `The model '${originalModelName}' is English only and cannot be used to transcribe language '${languageCode}'. using '${modelName}' instead.`, chalk.yellowBright)
+		logger.logTitledMessage(`Warning`, `The model '${originalModelName}' is English only and cannot be used to transcribe language '${languageCode}'. Using '${modelName}' instead.`, chalk.yellowBright, 'warning')
 	}
 
 	return modelName

package/src/utilities/FileDownloader.ts

@@ -9,6 +9,7 @@ import path from 'node:path'
 import { extractTarball } from './Compression.js'
 import { createWriteStream, move, remove, readdir, ensureDir } from './FileSystem.js'
 import chalk from 'chalk'
+import { logLevelGreaterOrEqualTo } from '../api/GlobalOptions.js'
 
 export async function downloadAndExtractTarball(options: GaxiosOptions, targetDir: string, baseTempPath: string, displayName = 'archive') {
 	const logger = new Logger()
@@ -18,7 +19,10 @@ export async function downloadAndExtractTarball(options: GaxiosOptions, targetDi
 	const tempDirPath = path.join(baseTempPath, `/${randomID}`)
 	await ensureDir(tempDirPath)
 
+	logger.end()
+
 	await downloadFile(options, tempTarballPath, `${chalk.cyanBright('Downloading')} ${chalk.greenBright(displayName)}`)
+
 	logger.end()
 
 	logger.start(`Extracting ${displayName}`)
@@ -40,7 +44,7 @@ export async function downloadAndExtractTarball(options: GaxiosOptions, targetDi
 }
 
 export async function downloadFile(options: GaxiosOptions, targetFilePath: string, prompt = 'Downloading') {
-	const write = writeToStderr
+	const write = logLevelGreaterOrEqualTo('info') ? writeToStderr : () => {}
 
 	const downloadPromise = new OpenPromise<void>()
 

package/src/utilities/Logger.ts

@@ -1,6 +1,7 @@
 import chalk from 'chalk'
 import { Timer } from './Timer.js'
 import { logToStderr, writeToStderr, yieldToEventLoop } from './Utilities.js'
+import { LogLevel, logLevelGreaterOrEqualTo, logLevelSmallerThan } from '../api/GlobalOptions.js'
 
 let currentActiveLogger: Logger | null = null
 
@@ -23,7 +24,9 @@ export class Logger {
 			await yieldToEventLoop()
 		}
 
-		writeToStderr(`${titleColor(title)}.. `)
+		if (logLevelGreaterOrEqualTo('info')) {
+			writeToStderr(`${titleColor(title)}.. `)
+		}
 
 		this.setAsActiveLogger()
 
@@ -44,30 +47,41 @@ export class Logger {
 		if (this.active && currentActiveLogger == this) {
 			const elapsedTime = this.timer.elapsedTime
 
-			writeToStderr(`${elapsedTime.toFixed(1)}ms\n`)
+			if (logLevelGreaterOrEqualTo('info')) {
+				writeToStderr(`${elapsedTime.toFixed(1)}ms\n`)
+			}
+
 			currentActiveLogger = null
 		}
 
 		this.active = false
 	}
 
-	logDuration(message: any, startTime: number, titleColor = chalk.cyanBright) {
+	logDuration(message: any, startTime: number, titleColor = chalk.cyanBright, logLevel: LogLevel = 'info') {
 		const duration = Timer.currentTime - startTime
 
-		this.log(`${titleColor(message)}: ${duration.toFixed(1)}ms`)
+		this.log(`${titleColor(message)}: ${duration.toFixed(1)}ms`, logLevel)
 	}
 
-	logTitledMessage(title: string, content: string, titleColor = chalk.cyanBright) {
-		this.log(`${titleColor(title)}: ${content}`)
+	logTitledMessage(title: string, content: any, titleColor = chalk.cyanBright, logLevel: LogLevel = 'info') {
+		this.log(`${titleColor(title)}: ${content}`, logLevel)
 	}
 
-	log(message: any) {
+	log(message: any, logLevel: LogLevel = 'info') {
+		if (logLevelSmallerThan(logLevel)) {
+			return
+		}
+
 		if (currentActiveLogger == this || currentActiveLogger == null) {
 			logToStderr(message)
 		}
 	}
 
-	write(message: any) {
+	write(message: any, logLevel: LogLevel = 'info') {
+		if (logLevelSmallerThan(logLevel)) {
+			return
+		}
+
 		if (currentActiveLogger == this || currentActiveLogger == null) {
 			writeToStderr(message)
 		}

package/src/utilities/PackageManager.ts

@@ -4,6 +4,7 @@ import { getAppDataDir, ensureDir, existsSync, remove } from './FileSystem.js'
 import { appName } from '../api/Common.js'
 import { GaxiosOptions } from 'gaxios'
 import { getAppTempDir } from './PathUtilities.js'
+import { getGlobalOption } from '../api/GlobalOptions.js'
 
 export async function loadPackage(packageName: string) {
 	packageName = resolveToVersionedPackageNameIfNeeded(packageName)
@@ -16,13 +17,15 @@ export async function loadPackage(packageName: string) {
 		return packagePath
 	}
 
+	const packageBaseURL = getGlobalOption('packageBaseURL')
+
 	const tempPath = getAppTempDir(appName)
 
 	const headers = {
 	}
 
 	const options: GaxiosOptions = {
-		url: `${basePackageUrl}${packageName}.tar.gz`,
+		url: `${packageBaseURL}${packageName}.tar.gz`,
 		headers
 	}
 
@@ -76,8 +79,6 @@ export function resolveVersionTagForUnversionedPackageName(unversionedPackageNam
 	return packageVersionTagResolutionLookup[unversionedPackageName] || defaultVersionTag
 }
 
-const basePackageUrl = 'https://huggingface.co/echogarden/echogarden-packages/resolve/main/'
-
 const defaultVersionTag = '20230718'
 
 const packageVersionTagResolutionLookup: { [packageName: string]: string } = {

package/src/utilities/Utilities.ts

@@ -517,7 +517,7 @@ export function getTokenRepetitionScore(tokens: string[] | number[]) {
 			longestMatch = matchLength
 		}
 
-		const cycleCount = matchLength / i
+		const cycleCount = (matchLength / i) + 1
 
 		if (cycleCount > longestCycleRepetition) {
 			longestCycleRepetition = cycleCount
@@ -540,7 +540,7 @@ export async function resolveModuleScriptPath(moduleName: string) {
 export async function runOperationWithRetries<R>(
 	operationFunc: () => Promise<R>,
 	logger: Logger,
-	opreationName = 'Operation',
+	operationName = 'Operation',
 	delayBetweenRetries = 2000,
 	maxRetries = 200) {
 
@@ -560,21 +560,21 @@ export async function runOperationWithRetries<R>(
 
 			logger.setAsActiveLogger()
 
-			logger.logTitledMessage(`Error`, e.message, chalk.redBright)
-			logger.log(``)
-			logger.logTitledMessage(`${opreationName} failed`, `Trying again in ${delayBetweenRetries}ms..`, chalk.redBright)
+			logger.logTitledMessage(`Error`, e.message, chalk.redBright, 'error')
+			logger.log('', 'error')
+			logger.logTitledMessage(`${operationName} failed`, `Trying again in ${delayBetweenRetries}ms..`, chalk.redBright, 'error')
 
 			await delay(delayBetweenRetries)
 
-			logger.log(``)
-			logger.logTitledMessage(`Starting retry attempt`, `${retryIndex} / ${maxRetries}`, chalk.yellowBright)
-			logger.log(``)
+			logger.log(``, 'warning')
+			logger.logTitledMessage(`Starting retry attempt`, `${retryIndex} / ${maxRetries}`, chalk.yellowBright, 'warning')
+			logger.log(``, 'warning')
 
 			logger.unsetAsActiveLogger()
 		}
 	}
 
-	throw new Error(`${opreationName} failed after ${maxRetries} retry attempts`)
+	throw new Error(`${operationName} failed after ${maxRetries} retry attempts`)
 }
 
 export function writeToStdinInChunks(process: ChildProcessWithoutNullStreams, buffer: Buffer, chunkSize: number) {