@transloadit/node 4.10.4 → 4.10.6

package/src/cli/intentCommandSpecs.ts

@@ -1,6 +1,7 @@
 import type { z } from 'zod'
 
 import type { RobotMetaInput } from '../alphalib/types/robots/_instructions-primitives.ts'
+
 import {
   robotAudioWaveformInstructionsSchema,
   meta as robotAudioWaveformMeta,
@@ -247,6 +248,11 @@ export const intentCatalog = [
     meta: robotAudioWaveformMeta,
     schema: robotAudioWaveformInstructionsSchema,
   }),
+  defineSemanticIntent({
+    kind: 'semantic',
+    semantic: 'speech-transcribe',
+    paths: ['speech', 'transcribe'],
+  }),
   defineRobotIntent({
     kind: 'robot',
     robot: '/text/speak',

package/src/cli/intentCommands.ts

@@ -1,5 +1,4 @@
 import type { CommandClass } from 'clipanion'
-import { Command } from 'clipanion'
 import type { ZodObject, ZodRawShape, ZodTypeAny } from 'zod'
 
 import type { RobotMetaInput } from '../alphalib/types/robots/_instructions-primitives.ts'
@@ -10,14 +9,7 @@ import type {
   RobotIntentDefinition,
   SemanticIntentDefinition,
 } from './intentCommandSpecs.ts'
-import { getIntentPaths, getIntentResultStepName, intentCatalog } from './intentCommandSpecs.ts'
 import type { IntentFieldKind, IntentFieldSpec } from './intentFields.ts'
-import {
-  createIntentOption,
-  inferIntentExampleValue,
-  inferIntentFieldKind,
-  unwrapIntentSchema,
-} from './intentFields.ts'
 import type { IntentInputPolicy } from './intentInputPolicy.ts'
 import type {
   IntentCommandDefinition,
@@ -25,6 +17,16 @@ import type {
   IntentNoInputCommandDefinition,
   IntentSingleStepExecutionDefinition,
 } from './intentRuntime.ts'
+
+import { Command } from 'clipanion'
+
+import { getIntentPaths, getIntentResultStepName, intentCatalog } from './intentCommandSpecs.ts'
+import {
+  createIntentOption,
+  inferIntentExampleValue,
+  inferIntentFieldKind,
+  unwrapIntentSchema,
+} from './intentFields.ts'
 import {
   GeneratedBundledFileIntentCommand,
   GeneratedNoInputIntentCommand,

package/src/cli/intentFields.ts

@@ -1,6 +1,7 @@
+import type { z } from 'zod'
+
 import { Option } from 'clipanion'
 import * as t from 'typanion'
-import type { z } from 'zod'
 import {
   ZodArray,
   ZodBoolean,

package/src/cli/intentRuntime.ts

@@ -1,13 +1,18 @@
+import type { z } from 'zod'
+
+import type { AssembliesCreateOptions } from './commands/assemblies.ts'
+import type { SharedCliOptionDocumentation } from './fileProcessingOptions.ts'
+import type { IntentFieldSpec } from './intentFields.ts'
+import type { IntentInputPolicy } from './intentInputPolicy.ts'
+
 import { statSync } from 'node:fs'
 import { basename, dirname, join, parse, resolve } from 'node:path'
+
 import { Option } from 'clipanion'
-import type { z } from 'zod'
 
 import { prepareInputFiles } from '../inputFiles.ts'
-import type { AssembliesCreateOptions } from './commands/assemblies.ts'
 import * as assembliesCommands from './commands/assemblies.ts'
 import { AuthenticatedCommand } from './commands/BaseCommand.ts'
-import type { SharedCliOptionDocumentation } from './fileProcessingOptions.ts'
 import {
   concurrencyOption,
   countProvidedInputs,
@@ -20,9 +25,7 @@ import {
   validateSharedFileProcessingOptions,
   watchOption,
 } from './fileProcessingOptions.ts'
-import type { IntentFieldSpec } from './intentFields.ts'
 import { coerceIntentFieldValue } from './intentFields.ts'
-import type { IntentInputPolicy } from './intentInputPolicy.ts'
 import { printResultUrls } from './resultUrls.ts'
 import { getSemanticIntentDescriptor } from './semanticIntents/index.ts'
 

package/src/cli/resultUrls.ts

@@ -1,5 +1,6 @@
 import type { IOutputCtl } from './OutputCtl.ts'
 import type { NormalizedAssemblyResults } from './resultFiles.ts'
+
 import { normalizeAssemblyResults } from './resultFiles.ts'
 
 export interface ResultUrlRow {

package/src/cli/semanticIntents/imageDescribe.ts

@@ -1,9 +1,10 @@
-import { parseStringArrayValue } from '../intentFields.ts'
 import type {
   IntentDynamicStepExecutionDefinition,
   IntentOptionDefinition,
 } from '../intentRuntime.ts'
 import type { SemanticIntentDescriptor, SemanticIntentPresentation } from './index.ts'
+
+import { parseStringArrayValue } from '../intentFields.ts'
 import { parseOptionalEnumValue, parseUniqueEnumArray } from './parsing.ts'
 
 const imageDescribeFields = ['labels', 'altText', 'title', 'caption', 'description'] as const

package/src/cli/semanticIntents/imageGenerate.ts

@@ -1,6 +1,8 @@
-import { basename } from 'node:path'
 import type { IntentOptionDefinition, PreparedIntentInputs } from '../intentRuntime.ts'
 import type { SemanticIntentDescriptor, SemanticIntentPresentation } from './index.ts'
+
+import { basename } from 'node:path'
+
 import { parseOptionalEnumValue } from './parsing.ts'
 
 const defaultImageGenerateModel = 'google/nano-banana-2'

package/src/cli/semanticIntents/index.ts

@@ -4,12 +4,14 @@ import type {
   IntentRunnerKind,
   PreparedIntentInputs,
 } from '../intentRuntime.ts'
+
 import { imageDescribeSemanticIntentDescriptor } from './imageDescribe.ts'
 import { imageGenerateSemanticIntentDescriptor } from './imageGenerate.ts'
 import {
   markdownDocxSemanticIntentDescriptor,
   markdownPdfSemanticIntentDescriptor,
 } from './markdownPdf.ts'
+import { speechTranscribeSemanticIntentDescriptor } from './speechTranscribe.ts'
 
 export interface SemanticIntentPresentation {
   description: string
@@ -43,6 +45,7 @@ const semanticIntentDescriptors: Record<string, SemanticIntentDescriptor> = {
   'markdown-docx': {
     ...markdownDocxSemanticIntentDescriptor,
   },
+  'speech-transcribe': speechTranscribeSemanticIntentDescriptor,
 }
 
 export function getSemanticIntentDescriptor(name: string): SemanticIntentDescriptor {

package/src/cli/semanticIntents/markdownPdf.ts

@@ -1,5 +1,6 @@
 import type { IntentOptionDefinition } from '../intentRuntime.ts'
 import type { SemanticIntentDescriptor, SemanticIntentPresentation } from './index.ts'
+
 import { parseOptionalEnumValue } from './parsing.ts'
 
 const defaultMarkdownFormat = 'gfm'

package/src/cli/semanticIntents/speechTranscribe.ts

@@ -0,0 +1,140 @@
+import type {
+  IntentDynamicStepExecutionDefinition,
+  IntentOptionDefinition,
+} from '../intentRuntime.ts'
+import type { SemanticIntentDescriptor, SemanticIntentPresentation } from './index.ts'
+
+import { parseOptionalEnumValue } from './parsing.ts'
+
+const speechTranscribeProviders = ['aws', 'gcp', 'replicate'] as const
+const speechTranscribeFormats = ['text', 'json', 'srt', 'webvtt'] as const
+
+type SpeechTranscribeProvider = (typeof speechTranscribeProviders)[number]
+type SpeechTranscribeFormat = (typeof speechTranscribeFormats)[number]
+
+const defaultSpeechTranscribeProvider = 'replicate' satisfies SpeechTranscribeProvider
+const defaultSpeechTranscribeFormat = 'text' satisfies SpeechTranscribeFormat
+
+const speechTranscribeExecutionDefinition = {
+  kind: 'dynamic-step',
+  handler: 'speech-transcribe',
+  resultStepName: 'transcribe',
+  fields: [
+    {
+      name: 'provider',
+      kind: 'string',
+      propertyName: 'provider',
+      optionFlags: '--provider',
+      description: 'Provider to use for transcription. Defaults to replicate.',
+      required: false,
+      exampleValue: defaultSpeechTranscribeProvider,
+    },
+    {
+      name: 'format',
+      kind: 'string',
+      propertyName: 'format',
+      optionFlags: '--format',
+      description: 'Output format. Defaults to text.',
+      required: false,
+      exampleValue: defaultSpeechTranscribeFormat,
+    },
+    {
+      name: 'source_language',
+      kind: 'string',
+      propertyName: 'sourceLanguage',
+      optionFlags: '--source-language',
+      description:
+        'Spoken language as a BCP-47 code, for providers that support explicit source languages.',
+      required: false,
+      exampleValue: 'en-US',
+    },
+    {
+      name: 'target_language',
+      kind: 'string',
+      propertyName: 'targetLanguage',
+      optionFlags: '--target-language',
+      description: 'Target written language for providers that support translation.',
+      required: false,
+      exampleValue: 'en-US',
+    },
+  ] as const satisfies readonly IntentOptionDefinition[],
+} satisfies IntentDynamicStepExecutionDefinition
+
+const speechTranscribeCommandPresentation = {
+  description: 'Transcribe speech in audio or video files',
+  details:
+    'Runs `/speech/transcribe` with a text-first default and writes the transcript to `--output`.',
+  examples: [
+    [
+      'Transcribe an audio file to text',
+      'transloadit speech transcribe --input voice.opus --output voice.txt',
+    ],
+    [
+      'Generate subtitles',
+      'transloadit speech transcribe --input clip.mp4 --format webvtt --output captions.vtt',
+    ],
+  ] as Array<[string, string]>,
+} as const satisfies SemanticIntentPresentation
+
+function parseProvider(value: unknown): SpeechTranscribeProvider {
+  return (
+    parseOptionalEnumValue({
+      flagName: '--provider',
+      supportedValues: speechTranscribeProviders,
+      value,
+    }) ?? defaultSpeechTranscribeProvider
+  )
+}
+
+function parseFormat(value: unknown): SpeechTranscribeFormat {
+  return (
+    parseOptionalEnumValue({
+      flagName: '--format',
+      supportedValues: speechTranscribeFormats,
+      value,
+    }) ?? defaultSpeechTranscribeFormat
+  )
+}
+
+function parseOptionalString(value: unknown, flagName: string): string | null {
+  if (value == null || value === '') {
+    return null
+  }
+
+  if (typeof value !== 'string') {
+    throw new Error(`${flagName} must be a string`)
+  }
+
+  const trimmed = value.trim()
+  return trimmed.length > 0 ? trimmed : null
+}
+
+function createSpeechTranscribeStep(
+  rawValues: Record<string, unknown>,
+  _context: { hasInputs: boolean },
+): Record<string, unknown> {
+  const provider = parseProvider(rawValues.provider)
+  const format = parseFormat(rawValues.format)
+  const sourceLanguage = parseOptionalString(rawValues.source_language, '--source-language')
+  const targetLanguage = parseOptionalString(rawValues.target_language, '--target-language')
+
+  return {
+    robot: '/speech/transcribe',
+    use: ':original',
+    result: true,
+    provider,
+    format,
+    ...(sourceLanguage != null ? { source_language: sourceLanguage } : {}),
+    ...(targetLanguage != null ? { target_language: targetLanguage } : {}),
+  }
+}
+
+export const speechTranscribeSemanticIntentDescriptor = {
+  createStep: createSpeechTranscribeStep,
+  defaultOutputPath: 'output.txt',
+  execution: speechTranscribeExecutionDefinition,
+  inputPolicy: { kind: 'required' },
+  outputDescription: 'Write the transcript to this path or directory',
+  presentation: speechTranscribeCommandPresentation,
+  runnerKind: 'watchable',
+} as const satisfies SemanticIntentDescriptor

package/src/cli/stepsInput.ts

@@ -1,6 +1,7 @@
+import type { StepsInput } from '../alphalib/types/template.ts'
+
 import fsp from 'node:fs/promises'
 
-import type { StepsInput } from '../alphalib/types/template.ts'
 import { stepsSchema } from '../alphalib/types/template.ts'
 
 export function parseStepsInputJson(content: string): StepsInput {

package/src/cli/template-last-modified.ts

@@ -1,4 +1,5 @@
 import type { Transloadit } from '../Transloadit.ts'
+
 import { ensureError } from './types.ts'
 
 interface TemplateItem {

package/src/cli/types.ts

@@ -1,5 +1,7 @@
-import { z } from 'zod'
 import type { Steps } from '../alphalib/types/template.ts'
+
+import { z } from 'zod'
+
 import { optionalStepsSchema } from '../alphalib/types/template.ts'
 
 // Zod schemas for runtime validation

package/src/cli.ts

@@ -4,6 +4,7 @@ import { realpathSync } from 'node:fs'
 import path from 'node:path'
 import process from 'node:process'
 import { fileURLToPath } from 'node:url'
+
 import { loadProjectDotenvIntoProcessEnv } from './cli/helpers.ts'
 import { ensureError } from './cli/types.ts'
 

package/src/inputFiles.ts

@@ -1,17 +1,22 @@
 import type { LookupAddress, LookupOptions } from 'node:dns'
+import type { LookupFunction } from 'node:net'
+import type { Readable } from 'node:stream'
+
+import type { EntryObject, IPFamily } from 'cacheable-lookup'
+import type { Input as IntoStreamInput } from 'into-stream'
+
+import type { CreateAssemblyParams } from './apiTypes.ts'
+
 import * as dnsPromises from 'node:dns/promises'
 import { createWriteStream } from 'node:fs'
 import { mkdtemp, rm, writeFile } from 'node:fs/promises'
-import type { LookupFunction } from 'node:net'
 import { isIP } from 'node:net'
 import { tmpdir } from 'node:os'
 import { basename, join, parse } from 'node:path'
-import type { Readable } from 'node:stream'
 import { pipeline } from 'node:stream/promises'
-import type { EntryObject, IPFamily } from 'cacheable-lookup'
+
 import got from 'got'
-import type { Input as IntoStreamInput } from 'into-stream'
-import type { CreateAssemblyParams } from './apiTypes.ts'
+
 import { ensureUniqueCounterValue } from './ensureUniqueCounter.ts'
 
 export type InputFile =

package/src/lintAssemblyInput.ts

@@ -1,8 +1,9 @@
-import { getIndentation } from './alphalib/stepParsing.ts'
-import { mergeTemplateContent } from './alphalib/templateMerge.ts'
 import type { AssemblyInstructionsInput, StepsInput } from './alphalib/types/template.ts'
 import type { ResponseTemplateContent, TemplateContent } from './apiTypes.ts'
 
+import { getIndentation } from './alphalib/stepParsing.ts'
+import { mergeTemplateContent } from './alphalib/templateMerge.ts'
+
 const DEFAULT_INDENT = '  '
 
 const isRecord = (value: unknown): value is Record<string, unknown> =>

package/src/lintAssemblyInstructions.ts

@@ -1,8 +1,9 @@
 import type { HydratedLintIssue } from './alphalib/assembly-linter.lang.en.ts'
-import { hydrateLintIssues } from './alphalib/assembly-linter.lang.en.ts'
-import { applyFix, parseAndLint } from './alphalib/assembly-linter.ts'
 import type { AssemblyInstructionsInput, StepsInput } from './alphalib/types/template.ts'
 import type { ResponseTemplateContent, TemplateContent } from './apiTypes.ts'
+
+import { hydrateLintIssues } from './alphalib/assembly-linter.lang.en.ts'
+import { applyFix, parseAndLint } from './alphalib/assembly-linter.ts'
 import { buildLintInput, unwrapStepsOnly } from './lintAssemblyInput.ts'
 
 export type LintFatalLevel = 'error' | 'warning'

package/src/robots.ts

@@ -1,4 +1,5 @@
 import type { z } from 'zod'
+
 import { getRobotSchemaOptions, robotsMeta } from './alphalib/types/robots/_index.ts'
 
 export type RobotListOptions = {

package/src/tus.ts

@@ -1,12 +1,16 @@
+import type { Readable } from 'node:stream'
+
+import type { OnSuccessPayload, UploadOptions } from 'tus-js-client'
+
+import type { AssemblyStatus } from './alphalib/types/assemblyStatus.ts'
+import type { UploadProgress } from './Transloadit.ts'
+
 import { stat } from 'node:fs/promises'
 import { basename } from 'node:path'
-import type { Readable } from 'node:stream'
+
 import debug from 'debug'
 import pMap from 'p-map'
-import type { OnSuccessPayload, UploadOptions } from 'tus-js-client'
 import { Upload } from 'tus-js-client'
-import type { AssemblyStatus } from './alphalib/types/assemblyStatus.ts'
-import type { UploadProgress } from './Transloadit.ts'
 
 const log = debug('transloadit')
 const logWarn = debug('transloadit:warn')