@wovin/tranz 0.1.36 → 0.2.0

package/dist/{providers.min.js → providers.js}

@@ -245,6 +245,16 @@ var MistralProvider = class {
     if (timestampGranularity) {
       formData.append("timestamp_granularities", timestampGranularity);
     }
+    if (params.contextBias && params.contextBias.length > 0) {
+      if (params.contextBias.length > VOXTRAL_LIMITS.maxContextBiasingTerms) {
+        throw new Error(
+          `contextBias has ${params.contextBias.length} terms; Voxtral limit is ${VOXTRAL_LIMITS.maxContextBiasingTerms}`
+        );
+      }
+      for (const term of params.contextBias) {
+        formData.append("context_bias[]", term);
+      }
+    }
     const response = await fetch("https://api.mistral.ai/v1/audio/transcriptions", {
       method: "POST",
       headers: {
@@ -260,14 +270,27 @@ var MistralProvider = class {
     if (!result?.text) {
       return { text: "", error: "No transcription returned", rawResponse: result };
     }
-    const words = result.words || result.segments?.flatMap((seg) => seg.words || []);
+    const segments = Array.isArray(result.segments) && result.segments.length > 0 ? result.segments.map((seg) => ({
+      startMs: Math.round((seg.start ?? 0) * 1e3),
+      endMs: Math.round((seg.end ?? 0) * 1e3),
+      text: seg.text ?? "",
+      ...seg.speaker_id !== void 0 ? { diarization: seg.speaker_id } : {}
+    })) : void 0;
+    let words;
+    if (Array.isArray(result.words) && result.words.length > 0) {
+      words = result.words;
+    } else if (Array.isArray(result.segments)) {
+      const nested = result.segments.flatMap((seg) => seg.words ?? []);
+      if (nested.length > 0) words = nested;
+    }
     const duration = result.usage?.prompt_audio_seconds;
     return {
       text: result.text,
       language: result.language ?? params.language,
       model: result.model,
       duration,
-      words,
+      ...words ? { words } : {},
+      ...segments ? { segments } : {},
       rawResponse: result
     };
   }
@@ -620,6 +643,10 @@ async function autoSplitAudio(audioPath, outputDir, config = {}) {
 }
 
 // src/utils/audio/merge-results.ts
+function prefixChunkLabel(chunkIndex, value) {
+  if (value === void 0) return void 0;
+  return `chunk${chunkIndex}/${String(value)}`;
+}
 function mergeTranscriptionResults(results, segments) {
   if (results.length === 0) {
     return {
@@ -628,10 +655,7 @@ function mergeTranscriptionResults(results, segments) {
     };
   }
   if (results.length === 1) {
-    return {
-      ...results[0],
-      totalSegments: 1
-    };
+    return results[0];
   }
   const errors = results.map((r, i) => r.error ? `Segment ${i}: ${r.error}` : null).filter(Boolean);
   if (errors.length > 0) {
@@ -652,12 +676,26 @@ function mergeTranscriptionResults(results, segments) {
         start: (word.start || 0) + segment.startSec,
         end: (word.end || 0) + segment.startSec,
         confidence: word.confidence,
-        speaker: word.speaker
+        speaker: prefixChunkLabel(i, word.speaker)
+      });
+    }
+  }
+  const mergedSegments = [];
+  for (let i = 0; i < results.length; i++) {
+    const result = results[i];
+    const chunkOffsetMs = Math.round(segments[i].startSec * 1e3);
+    if (!result.segments) continue;
+    for (const seg of result.segments) {
+      mergedSegments.push({
+        startMs: seg.startMs + chunkOffsetMs,
+        endMs: seg.endMs + chunkOffsetMs,
+        text: seg.text,
+        ...seg.diarization !== void 0 ? { diarization: prefixChunkLabel(i, seg.diarization) } : {}
       });
     }
   }
   const totalDuration = segments.reduce((sum, seg) => sum + seg.durationSec, 0);
-  const segmentMeta = results.map((r, i) => ({
+  const audioChunks = results.map((r, i) => ({
     index: i,
     startSec: segments[i].startSec,
     endSec: segments[i].endSec,
@@ -665,24 +703,23 @@ function mergeTranscriptionResults(results, segments) {
   }));
   const mergedRawResponse = {
     merged: true,
-    segmentCount: results.length,
-    segments: results.map((r, i) => ({
+    chunkCount: results.length,
+    chunks: results.map((r, i) => ({
       index: i,
       startSec: segments[i].startSec,
       rawResponse: r.rawResponse
-    })),
-    words: mergedWords
+    }))
   };
   const firstResult = results[0];
   return {
     text: mergedText,
-    words: mergedWords,
     duration: totalDuration,
     language: firstResult.language,
     model: firstResult.model,
     rawResponse: mergedRawResponse,
-    segments: segmentMeta,
-    totalSegments: results.length
+    audioChunks,
+    ...mergedWords.length > 0 ? { words: mergedWords } : {},
+    ...mergedSegments.length > 0 ? { segments: mergedSegments } : {}
   };
 }
 
@@ -784,6 +821,7 @@ function createMistralTranscriber(config) {
         duration: knownDuration,
         language,
         model = defaultModel,
+        contextBias,
         diarize = true,
         timestamps = language ? void 0 : "segment",
         autoSplit,
@@ -803,9 +841,10 @@ function createMistralTranscriber(config) {
           model,
           language,
           diarize,
-          timestampGranularity: timestamps
+          timestampGranularity: timestamps,
+          contextBias
         });
-        return { ...result, totalSegments: 1 };
+        return result;
       }
       if (audioUrl) {
         if (autoSplit === false) {
@@ -816,9 +855,10 @@ function createMistralTranscriber(config) {
             model,
             language,
             diarize,
-            timestampGranularity: timestamps
+            timestampGranularity: timestamps,
+            contextBias
           });
-          return { ...result2, totalSegments: 1 };
+          return result2;
         }
         let duration2 = knownDuration;
         if (duration2 === void 0) {
@@ -840,9 +880,10 @@ function createMistralTranscriber(config) {
             model,
             language,
             diarize,
-            timestampGranularity: timestamps
+            timestampGranularity: timestamps,
+            contextBias
           });
-          return { ...result2, totalSegments: 1 };
+          return result2;
         }
         log.info(`Downloading URL to temp file for processing...`);
         const outDir2 = splitOutputDir || path3.join(os.tmpdir(), `tranz-${Date.now()}`);
@@ -855,6 +896,7 @@ function createMistralTranscriber(config) {
           model,
           diarize,
           timestamps,
+          contextBias,
           autoSplit: true,
           splitOutputDir: outDir2,
           logger: customLogger,
@@ -881,9 +923,10 @@ function createMistralTranscriber(config) {
           model,
           language,
           diarize,
-          timestampGranularity: timestamps
+          timestampGranularity: timestamps,
+          contextBias
         });
-        return { ...result, totalSegments: 1 };
+        return result;
       }
       log.info(`Duration ${duration.toFixed(1)}s > ${maxDuration}s, splitting audio...`);
       const outDir = splitOutputDir || path3.join(os.tmpdir(), `tranz-split-${Date.now()}`);
@@ -902,7 +945,8 @@ function createMistralTranscriber(config) {
           model,
           language,
           diarize,
-          timestampGranularity: timestamps
+          timestampGranularity: timestamps,
+          contextBias
         });
         results.push(result);
       }

package/dist/utils/audio/merge-results.d.ts

@@ -11,21 +11,23 @@ export interface WordData {
     start: number;
     end: number;
     confidence?: number;
-    speaker?: string;
+    speaker?: string | number;
 }
 /**
- * Merged transcription result with segment information
+ * Metadata describing one audio chunk in an auto-split + merge run.
+ */
+export interface AudioChunk {
+    index: number;
+    startSec: number;
+    endSec: number;
+    text: string;
+}
+/**
+ * Merged transcription result with chunk-level metadata.
  */
 export interface MergedTranscriptionResult extends TranscriptionResult {
-    /** Segment metadata for reference */
-    segments?: {
-        index: number;
-        startSec: number;
-        endSec: number;
-        text: string;
-    }[];
-    /** Total segments that were merged */
-    totalSegments?: number;
+    /** Audio chunks that were transcribed independently and merged. Absent when no split happened. */
+    audioChunks?: AudioChunk[];
 }
 /**
  * Merge multiple transcription results from audio segments into one
@@ -40,7 +42,7 @@ export declare function mergeTranscriptionResults(results: TranscriptionResult[]
  * Format merged results with optional segment markers in the text
  *
  * @param result - Merged transcription result
- * @param includeMarkers - Whether to include [Segment N] markers
+ * @param includeMarkers - Whether to include [Chunk N] markers
  * @returns Formatted text
  */
 export declare function formatMergedText(result: MergedTranscriptionResult, includeMarkers?: boolean): string;

package/dist/utils/audio/merge-results.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"merge-results.d.ts","sourceRoot":"","sources":["../../../src/utils/audio/merge-results.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,+BAA+B,CAAA;AACxE,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,YAAY,CAAA;AAE9C;;GAEG;AACH,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,MAAM,CAAA;IACZ,KAAK,EAAE,MAAM,CAAA;IACb,GAAG,EAAE,MAAM,CAAA;IACX,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,OAAO,CAAC,EAAE,MAAM,CAAA;CACjB;AAED;;GAEG;AACH,MAAM,WAAW,yBAA0B,SAAQ,mBAAmB;IACpE,qCAAqC;IACrC,QAAQ,CAAC,EAAE;QACT,KAAK,EAAE,MAAM,CAAA;QACb,QAAQ,EAAE,MAAM,CAAA;QAChB,MAAM,EAAE,MAAM,CAAA;QACd,IAAI,EAAE,MAAM,CAAA;KACb,EAAE,CAAA;IACH,sCAAsC;IACtC,aAAa,CAAC,EAAE,MAAM,CAAA;CACvB;AAED;;;;;;;GAOG;AACH,wBAAgB,yBAAyB,CACvC,OAAO,EAAE,mBAAmB,EAAE,EAC9B,QAAQ,EAAE,YAAY,EAAE,GACvB,yBAAyB,CAqF3B;AAED;;;;;;GAMG;AACH,wBAAgB,gBAAgB,CAC9B,MAAM,EAAE,yBAAyB,EACjC,cAAc,GAAE,OAAe,GAC9B,MAAM,CAeR"}
+{"version":3,"file":"merge-results.d.ts","sourceRoot":"","sources":["../../../src/utils/audio/merge-results.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,mBAAmB,EAAqB,MAAM,+BAA+B,CAAA;AAC3F,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,YAAY,CAAA;AAE9C;;GAEG;AACH,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,MAAM,CAAA;IACZ,KAAK,EAAE,MAAM,CAAA;IACb,GAAG,EAAE,MAAM,CAAA;IACX,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,OAAO,CAAC,EAAE,MAAM,GAAG,MAAM,CAAA;CAC1B;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,MAAM,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,MAAM,EAAE,MAAM,CAAA;IACd,IAAI,EAAE,MAAM,CAAA;CACb;AAED;;GAEG;AACH,MAAM,WAAW,yBAA0B,SAAQ,mBAAmB;IACpE,kGAAkG;IAClG,WAAW,CAAC,EAAE,UAAU,EAAE,CAAA;CAC3B;AAYD;;;;;;;GAOG;AACH,wBAAgB,yBAAyB,CACvC,OAAO,EAAE,mBAAmB,EAAE,EAC9B,QAAQ,EAAE,YAAY,EAAE,GACvB,yBAAyB,CAmG3B;AAED;;;;;;GAMG;AACH,wBAAgB,gBAAgB,CAC9B,MAAM,EAAE,yBAAyB,EACjC,cAAc,GAAE,OAAe,GAC9B,MAAM,CAeR"}

package/dist/utils/transcription/format.d.ts

@@ -11,4 +11,31 @@ export declare function formatTranscriptWithPauses(transcript: string, words: Ar
     end: number;
     confidence: number;
 }>, shortPauseThreshold?: number, longPauseThreshold?: number): string;
+import type { TranscriptionResult } from './providers.ts';
+import type { MergedTranscriptionResult } from '../audio/merge-results.ts';
+export interface FormatMarkdownOptions {
+    /** Silence gap (seconds) that ends a paragraph. Default 1.5. */
+    gapSec?: number;
+    /** Include `· Speaker N` in each paragraph header when diarization labels are present. Default true. */
+    speakerLabel?: boolean;
+    /** Prepend an `# <source>` title + bulleted metadata block. Default false. */
+    includeHeader?: boolean;
+    /** Source filename to use in the `# ` title and `Source:` line (when includeHeader=true). */
+    source?: string;
+    /** Total audio duration in seconds — used for `Duration:` line and for picking mm:ss vs h:mm:ss formatting. */
+    durationSec?: number;
+}
+/**
+ * Format a transcription result as readable Markdown with timestamped paragraphs.
+ *
+ * Groups adjacent segments into paragraphs, starting a new paragraph on either
+ * a silence gap ≥ `gapSec` OR a change in diarization label. Each paragraph is
+ * preceded by `**[mm:ss · Speaker N]**` (or `**[h:mm:ss · Speaker N]**` for
+ * audio ≥ 1h). The speaker suffix is dropped when no diarization labels are
+ * present or all segments share the same label.
+ *
+ * If `segments` is missing/empty, falls back to emitting `result.text` as a
+ * single (un-timestamped) paragraph.
+ */
+export declare function formatTranscriptAsMarkdown(result: TranscriptionResult | MergedTranscriptionResult, opts?: FormatMarkdownOptions): string;
 //# sourceMappingURL=format.d.ts.map

package/dist/utils/transcription/format.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"format.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/format.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,wBAAgB,0BAA0B,CACzC,UAAU,EAAE,MAAM,EAClB,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,GAAG,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,CAAC,EAC9E,mBAAmB,SAAM,EACzB,kBAAkB,SAAM,GACtB,MAAM,CAgDR"}
+{"version":3,"file":"format.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/format.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,wBAAgB,0BAA0B,CACzC,UAAU,EAAE,MAAM,EAClB,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,GAAG,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,CAAC,EAC9E,mBAAmB,SAAM,EACzB,kBAAkB,SAAM,GACtB,MAAM,CAgDR;AAED,OAAO,KAAK,EAAE,mBAAmB,EAAqB,MAAM,gBAAgB,CAAA;AAC5E,OAAO,KAAK,EAAE,yBAAyB,EAAE,MAAM,2BAA2B,CAAA;AAE1E,MAAM,WAAW,qBAAqB;IACrC,gEAAgE;IAChE,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,wGAAwG;IACxG,YAAY,CAAC,EAAE,OAAO,CAAA;IACtB,8EAA8E;IAC9E,aAAa,CAAC,EAAE,OAAO,CAAA;IACvB,6FAA6F;IAC7F,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,+GAA+G;IAC/G,WAAW,CAAC,EAAE,MAAM,CAAA;CACpB;AAsBD;;;;;;;;;;;GAWG;AACH,wBAAgB,0BAA0B,CACzC,MAAM,EAAE,mBAAmB,GAAG,yBAAyB,EACvD,IAAI,GAAE,qBAA0B,GAC9B,MAAM,CAuFR"}

package/dist/utils/transcription/providers.d.ts

@@ -2,6 +2,26 @@
  * Transcription provider types and interfaces
  * Defines the contract for all transcription providers
  */
+/**
+ * A single transcription segment (one diarized turn, or one segment-granularity unit).
+ *
+ * Timestamps are integer milliseconds — normalized at the SDK boundary so consumers
+ * can pass straight into the wovin annotation schema (see docs/annotation-schema.md).
+ *
+ * `diarization` is the anonymous, per-recording diarization label as returned by
+ * the provider (Mistral: `"speaker_1"`, Deepgram: `0`, AssemblyAI: `"A"`, …).
+ * It is NOT a real-world speaker identity — that's a separate (future) `speakerId` field.
+ *
+ * When `mergeTranscriptionResults` joins multiple chunks, `diarization` is rewritten
+ * as `` `chunk${index}/${value}` `` because per-chunk labels are not comparable
+ * across chunks.
+ */
+export interface TranscriptSegment {
+    startMs: number;
+    endMs: number;
+    text: string;
+    diarization?: string | number;
+}
 /**
  * Result object returned from transcription operations
  * Contains the transcribed text and optional provider-specific metadata
@@ -15,8 +35,10 @@ export interface TranscriptionResult {
     error?: string;
     /** Confidence score of the transcription (0-1) */
     confidence?: number;
-    /** Word-level data (timing, confidence, etc.) */
+    /** Word-level data — populated only when granularity='word' or the provider returns it. Left undefined otherwise (not `[]`). */
     words?: any[];
+    /** Segment-level data — populated when granularity='segment' (or the provider returns it). */
+    segments?: TranscriptSegment[];
     /** Duration of audio in seconds */
     duration?: number;
     /** Detected or specified language code */
@@ -63,6 +85,13 @@ export interface TranscribeParams {
     diarize?: boolean;
     /** Timestamp granularity for transcription (Mistral-specific) */
     timestampGranularity?: 'segment' | 'word';
+    /**
+     * Context biasing terms (Voxtral/Mistral-specific).
+     * Up to `VOXTRAL_LIMITS.maxContextBiasingTerms` (100) custom-vocabulary terms
+     * passed to the Voxtral transcribe endpoint as `context_bias[]` form fields.
+     * Ignored by non-Mistral providers.
+     */
+    contextBias?: string[];
     /** Path to model file (Whisper-specific) */
     modelPath?: string;
     /** Output directory for results (Whisper-specific) */

package/dist/utils/transcription/providers.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"providers.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/providers.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAUH;;;GAGG;AACH,MAAM,WAAW,mBAAmB;IAClC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAA;IACZ,sEAAsE;IACtE,WAAW,CAAC,EAAE,GAAG,CAAA;IACjB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,kDAAkD;IAClD,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,iDAAiD;IACjD,KAAK,CAAC,EAAE,GAAG,EAAE,CAAA;IACb,mCAAmC;IACnC,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,6DAA6D;IAC7D,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,+BAA+B;IAC/B,IAAI,EAAE,MAAM,CAAA;IACZ,+DAA+D;IAC/D,mBAAmB,CAAC,EAAE,MAAM,CAAA;IAC5B;;;;OAIG;IACH,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC,CAAA;CACnE;AAED;;;GAGG;AACH,MAAM,WAAW,gBAAgB;IAC/B,2CAA2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yDAAyD;IACzD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,yDAAyD;IACzD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,qDAAqD;IACrD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,oDAAoD;IACpD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iEAAiE;IACjE,oBAAoB,CAAC,EAAE,SAAS,GAAG,MAAM,CAAA;IACzC,4CAA4C;IAC5C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,sDAAsD;IACtD,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,oCAAoC;IACpC,MAAM,CAAC,EAAE,GAAG,CAAA;CACb;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,SAAS,GAAG,SAAS,GAAG,SAAS,CAAA;AAE5D;;;;;GAKG;AACH,wBAAgB,cAAc,CAAC,YAAY,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,GAAG,GAAG,qBAAqB,CAW9F;AASD;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEhB,OAAO,CAAC,QAAQ,CAAQ;IAExB,MAAM,CAAC,QAAQ;;;;;;;;;;;;;;;MAkBd;gBAEW,MAAM,CAAC,EAAE,GAAG;IAKlB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;YA+F1D,4BAA4B;CAkC3C;AAED;;;GAGG;AACH,eAAO,MAAM,cAAc;IACzB,2EAA2E;;IAE3E,4CAA4C;;IAE5C,uCAAuC;;CAExC,CAAA;AAED,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAChB,mBAAmB,SAAqC;IAElD,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAwGzE;AAED;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEV,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAkGzE"}
+{"version":3,"file":"providers.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/providers.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAUH;;;;;;;;;;;;;GAaG;AACH,MAAM,WAAW,iBAAiB;IAChC,OAAO,EAAE,MAAM,CAAA;IACf,KAAK,EAAE,MAAM,CAAA;IACb,IAAI,EAAE,MAAM,CAAA;IACZ,WAAW,CAAC,EAAE,MAAM,GAAG,MAAM,CAAA;CAG9B;AAED;;;GAGG;AACH,MAAM,WAAW,mBAAmB;IAClC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAA;IACZ,sEAAsE;IACtE,WAAW,CAAC,EAAE,GAAG,CAAA;IACjB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,kDAAkD;IAClD,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,gIAAgI;IAChI,KAAK,CAAC,EAAE,GAAG,EAAE,CAAA;IACb,8FAA8F;IAC9F,QAAQ,CAAC,EAAE,iBAAiB,EAAE,CAAA;IAC9B,mCAAmC;IACnC,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,6DAA6D;IAC7D,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,+BAA+B;IAC/B,IAAI,EAAE,MAAM,CAAA;IACZ,+DAA+D;IAC/D,mBAAmB,CAAC,EAAE,MAAM,CAAA;IAC5B;;;;OAIG;IACH,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC,CAAA;CACnE;AAED;;;GAGG;AACH,MAAM,WAAW,gBAAgB;IAC/B,2CAA2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yDAAyD;IACzD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,yDAAyD;IACzD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,qDAAqD;IACrD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,oDAAoD;IACpD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iEAAiE;IACjE,oBAAoB,CAAC,EAAE,SAAS,GAAG,MAAM,CAAA;IACzC;;;;;OAKG;IACH,WAAW,CAAC,EAAE,MAAM,EAAE,CAAA;IACtB,4CAA4C;IAC5C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,sDAAsD;IACtD,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,oCAAoC;IACpC,MAAM,CAAC,EAAE,GAAG,CAAA;CACb;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,SAAS,GAAG,SAAS,GAAG,SAAS,CAAA;AAE5D;;;;;GAKG;AACH,wBAAgB,cAAc,CAAC,YAAY,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,GAAG,GAAG,qBAAqB,CAW9F;AASD;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEhB,OAAO,CAAC,QAAQ,CAAQ;IAExB,MAAM,CAAC,QAAQ;;;;;;;;;;;;;;;MAkBd;gBAEW,MAAM,CAAC,EAAE,GAAG;IAKlB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;YA+F1D,4BAA4B;CAkC3C;AAED;;;GAGG;AACH,eAAO,MAAM,cAAc;IACzB,2EAA2E;;IAE3E,4CAA4C;;IAE5C,uCAAuC;;CAExC,CAAA;AAED,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAChB,mBAAmB,SAAqC;IAElD,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAuIzE;AAED;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEV,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAkGzE"}

package/dist/utils/transcription/transcribe.d.ts

@@ -27,6 +27,11 @@ export interface TranscribeOptions {
     diarize?: boolean;
     /** Timestamp granularity: 'word' | 'segment' (default: 'segment' when diarize=true, disabled if language set) */
     timestamps?: 'word' | 'segment';
+    /**
+     * Context biasing terms — up to `VOXTRAL_LIMITS.maxContextBiasingTerms` (100)
+     * custom-vocabulary entries passed to Voxtral as `context_bias[]`. Mistral only.
+     */
+    contextBias?: string[];
     /** Auto-split long audio (default: true). For URLs, detects duration first. */
     autoSplit?: boolean;
     /** Output directory for split segments (default: system temp) */

package/dist/utils/transcription/transcribe.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"transcribe.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/transcribe.ts"],"names":[],"mappings":"AAAA;;GAEG;AASH,OAAO,EAA6B,KAAK,yBAAyB,EAAE,MAAM,2BAA2B,CAAA;AAErG,kDAAkD;AAClD,MAAM,WAAW,gBAAgB;IAC/B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,KAAK,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;CAC7B;AAQD,MAAM,WAAW,iBAAiB;IAChC,yBAAyB;IACzB,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yEAAyE;IACzE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,mFAAmF;IACnF,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,kDAAkD;IAClD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,iDAAiD;IACjD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iHAAiH;IACjH,UAAU,CAAC,EAAE,MAAM,GAAG,SAAS,CAAA;IAC/B,+EAA+E;IAC/E,SAAS,CAAC,EAAE,OAAO,CAAA;IACnB,iEAAiE;IACjE,cAAc,CAAC,EAAE,MAAM,CAAA;IACvB,uCAAuC;IACvC,MAAM,CAAC,EAAE,gBAAgB,CAAA;IACzB,mCAAmC;IACnC,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAED,MAAM,WAAW,wBAAwB;IACvC,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAA;IACd,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AA6FD;;;;;;;;;;;;;;;;;;;GAmBG;AACH,iEAAiE;AACjE,MAAM,WAAW,kBAAkB;IACjC,UAAU,CAAC,OAAO,EAAE,iBAAiB,GAAG,OAAO,CAAC,yBAAyB,CAAC,CAAA;CAC3E;AAED,wBAAgB,wBAAwB,CAAC,MAAM,EAAE,wBAAwB,GAAG,kBAAkB,CA4K7F;AAED,+BAA+B;AAC/B,eAAO,MAAM,UAAU,iCAA2B,CAAA"}
+{"version":3,"file":"transcribe.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/transcribe.ts"],"names":[],"mappings":"AAAA;;GAEG;AASH,OAAO,EAA6B,KAAK,yBAAyB,EAAE,MAAM,2BAA2B,CAAA;AAErG,kDAAkD;AAClD,MAAM,WAAW,gBAAgB;IAC/B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,KAAK,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;CAC7B;AAQD,MAAM,WAAW,iBAAiB;IAChC,yBAAyB;IACzB,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yEAAyE;IACzE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,mFAAmF;IACnF,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,kDAAkD;IAClD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,iDAAiD;IACjD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iHAAiH;IACjH,UAAU,CAAC,EAAE,MAAM,GAAG,SAAS,CAAA;IAC/B;;;OAGG;IACH,WAAW,CAAC,EAAE,MAAM,EAAE,CAAA;IACtB,+EAA+E;IAC/E,SAAS,CAAC,EAAE,OAAO,CAAA;IACnB,iEAAiE;IACjE,cAAc,CAAC,EAAE,MAAM,CAAA;IACvB,uCAAuC;IACvC,MAAM,CAAC,EAAE,gBAAgB,CAAA;IACzB,mCAAmC;IACnC,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAED,MAAM,WAAW,wBAAwB;IACvC,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAA;IACd,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AA6FD;;;;;;;;;;;;;;;;;;;GAmBG;AACH,iEAAiE;AACjE,MAAM,WAAW,kBAAkB;IACjC,UAAU,CAAC,OAAO,EAAE,iBAAiB,GAAG,OAAO,CAAC,yBAAyB,CAAC,CAAA;CAC3E;AAED,wBAAgB,wBAAwB,CAAC,MAAM,EAAE,wBAAwB,GAAG,kBAAkB,CAmL7F;AAED,+BAA+B;AAC/B,eAAO,MAAM,UAAU,iCAA2B,CAAA"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wovin/tranz",
-  "version": "0.1.36",
+  "version": "0.2.0",
   "type": "module",
   "description": "Audio transcription library with provider support and auto-splitting",
   "author": "gotjoshua @gotjoshua",
@@ -12,29 +12,30 @@
     "directory": "packages/@wovin/tranz"
   },
   "bugs": "https://gitlab.com/onezoomin/ztax/tranz/-/issues",
-  "main": "./dist/index.min.js",
-  "module": "./dist/index.min.js",
+  "main": "./dist/index.js",
+  "module": "./dist/index.js",
   "types": "./dist/index.d.ts",
   "exports": {
     ".": {
-      "import": "./dist/index.min.js",
+      "import": "./dist/index.js",
       "types": "./dist/index.d.ts"
     },
     "./providers": {
-      "import": "./dist/providers.min.js",
+      "import": "./dist/providers.js",
       "types": "./dist/providers.d.ts"
     },
     "./audio": {
-      "import": "./dist/audio.min.js",
+      "import": "./dist/audio.js",
       "types": "./dist/audio.d.ts"
     },
     "./realtime": {
-      "import": "./dist/realtime.min.js",
+      "import": "./dist/realtime.js",
       "types": "./dist/realtime.d.ts"
     }
   },
   "files": [
-    "./dist/"
+    "./dist/",
+    "./src/"
   ],
   "publishConfig": {
     "access": "public"
@@ -72,6 +73,7 @@
     "dev:code": "tsup --watch",
     "dev:types": "tsc --emitDeclarationOnly --declaration --watch",
     "clean": "rm -rf .turbo && rm -rf node_modules && rm -rf dist",
+    "test": "tsx --test test/*.test.ts",
     "test:realtime": "tsx test/realtime-transcription.ts",
     "test:realtime-api": "tsx test/realtime-api-test.ts"
   }

package/src/audio.ts

@@ -0,0 +1,25 @@
+/**
+ * @wovin/tranz/audio - Audio utilities for splitting and merging
+ */
+
+export {
+  autoSplitAudio,
+  analyzeSplitPoints,
+  detectSilenceRegions,
+  getAudioDuration,
+  findOptimalSplitPoints,
+  splitAudioAtPoints,
+  DEFAULT_SPLIT_CONFIG,
+  type SplitConfig,
+  type SilenceRegion,
+  type SplitPoint,
+  type AudioSegment,
+  type SplitAnalysis,
+} from './utils/audio/split.ts'
+
+export {
+  mergeTranscriptionResults,
+  formatMergedText,
+  type MergedTranscriptionResult,
+  type WordData,
+} from './utils/audio/merge-results.ts'

package/src/index.ts

@@ -0,0 +1,61 @@
+/**
+ * @wovin/tranz - Audio transcription library
+ */
+
+// Transcription providers
+export {
+  createProvider,
+  MistralProvider,
+  WhisperProvider,
+  GreenPTProvider,
+  VOXTRAL_LIMITS,
+  type ProviderName,
+  type TranscribeParams,
+  type TranscriptionResult,
+  type TranscriptionProvider,
+  type TranscriptSegment,
+} from './utils/transcription/providers.ts'
+
+// Audio utilities
+export {
+  autoSplitAudio,
+  analyzeSplitPoints,
+  detectSilenceRegions,
+  getAudioDuration,
+  findOptimalSplitPoints,
+  splitAudioAtPoints,
+  DEFAULT_SPLIT_CONFIG,
+  type SplitConfig,
+  type SilenceRegion,
+  type SplitPoint,
+  type AudioSegment,
+  type SplitAnalysis,
+} from './utils/audio/split.ts'
+
+// Result merging
+export {
+  mergeTranscriptionResults,
+  formatMergedText,
+  type MergedTranscriptionResult,
+  type WordData,
+  type AudioChunk,
+} from './utils/audio/merge-results.ts'
+
+// Transcription formatting
+export {
+  formatTranscriptWithPauses,
+  formatTranscriptAsMarkdown,
+  type FormatMarkdownOptions,
+} from './utils/transcription/format.ts'
+
+// MIME type detection
+export { detectAudioMimeType } from './utils/transcription/mime-detection.ts'
+
+// Simple high-level API
+export {
+  createMistralTranscriber,
+  transcribe,
+  type TranscribeOptions,
+  type MistralTranscriberConfig,
+  type MistralTranscriber,
+} from './utils/transcription/transcribe.ts'

package/src/providers.ts

@@ -0,0 +1,23 @@
+/**
+ * @wovin/tranz/providers - Transcription provider implementations
+ */
+
+export {
+  createProvider,
+  MistralProvider,
+  WhisperProvider,
+  GreenPTProvider,
+  VOXTRAL_LIMITS,
+  type ProviderName,
+  type TranscribeParams,
+  type TranscriptionResult,
+  type TranscriptionProvider,
+  type TranscriptSegment,
+} from './utils/transcription/providers.ts'
+
+export {
+  createMistralTranscriber,
+  transcribe,
+  type TranscribeOptions,
+  type MistralTranscriberConfig,
+} from './utils/transcription/transcribe.ts'

package/src/realtime.ts

@@ -0,0 +1,58 @@
+/**
+ * Realtime transcription API
+ *
+ * This module provides a simple, event-driven interface for streaming audio
+ * transcription using Mistral's realtime WebSocket API.
+ *
+ * **Node.js only** - Browser support is currently disabled due to WebSocket
+ * authentication limitations with Mistral API.
+ *
+ * @example Node.js
+ * ```typescript
+ * import {
+ *   createRealtimeTranscriber,
+ *   captureAudioFromMicrophone,
+ * } from '@wovin/tranz/realtime'
+ *
+ * const transcriber = createRealtimeTranscriber({
+ *   apiKey: process.env.MISTRAL_API_KEY,
+ * })
+ *
+ * const { stream, stop } = await captureAudioFromMicrophone(16000)
+ *
+ * try {
+ *   for await (const event of transcriber.transcribe(stream)) {
+ *     if (event.type === 'transcription.text.delta') {
+ *       process.stdout.write(event.text)
+ *     } else if (event.type === 'transcription.done') {
+ *       console.log('\nComplete!')
+ *       break
+ *     }
+ *   }
+ * } finally {
+ *   stop()
+ * }
+ * ```
+ *
+ * @module @wovin/tranz/realtime
+ */
+
+export {
+  createRealtimeTranscriber,
+  captureAudioFromMicrophone,
+  captureAudioFromBrowser,
+  AudioEncoding,
+  type RealtimeEvent,
+  type RealtimeConfig,
+  type RealtimeTranscriber,
+  type TranscribeOptions,
+  type AudioFormat,
+  type AudioCaptureResult,
+  type SessionCreatedEvent,
+  type SessionUpdatedEvent,
+  type TranscriptionTextDeltaEvent,
+  type TranscriptionLanguageEvent,
+  type TranscriptionSegmentEvent,
+  type TranscriptionDoneEvent,
+  type ErrorEvent,
+} from "./utils/transcription/realtime.js";

package/src/utils/audio/index.ts

@@ -0,0 +1,6 @@
+/**
+ * Audio utilities for tranz-cli
+ */
+
+export * from './split.ts'
+export * from './merge-results.ts'