@code-yeongyu/senpi 2026.9.11 → 2026.9.12-3

package/node_modules/@google/genai/dist/tokenizer/node.cjs

@@ -1241,6 +1241,22 @@ var ServiceTier;
      */
     ServiceTier["PRIORITY"] = "priority";
 })(ServiceTier || (ServiceTier = {}));
+/** How the model processes input media for understanding. */
+var MediaProcessing;
+(function (MediaProcessing) {
+    /**
+     * Default. Uses model-specific processing
+     */
+    MediaProcessing["MEDIA_PROCESSING_UNSPECIFIED"] = "MEDIA_PROCESSING_UNSPECIFIED";
+    /**
+     * Fixed-rate frame extraction. All frames placed in context.
+     */
+    MediaProcessing["STATIC"] = "STATIC";
+    /**
+     * Model-driven dynamic navigation. Recommended for most use cases.
+     */
+    MediaProcessing["AGENTIC"] = "AGENTIC";
+})(MediaProcessing || (MediaProcessing = {}));
 /** The tokenization quality used for given media. */
 var PartMediaResolutionLevel;
 (function (PartMediaResolutionLevel) {
@@ -1292,6 +1308,10 @@ var ToolType;
      * File search tool, maps to Tool.file_search.
      */
     ToolType["FILE_SEARCH"] = "FILE_SEARCH";
+    /**
+     * Media processing tool.
+     */
+    ToolType["MEDIA_PROCESSING"] = "MEDIA_PROCESSING";
 })(ToolType || (ToolType = {}));
 /** Resource scope. */
 var ResourceScope;
@@ -1757,6 +1777,22 @@ var TurnCoverage;
      */
     TurnCoverage["TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO"] = "TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO";
 })(TurnCoverage || (TurnCoverage = {}));
+/** Transcription mode. */
+var AudioTranscriptionConfigMode;
+(function (AudioTranscriptionConfigMode) {
+    /**
+     * Unspecified transcription mode.
+     */
+    AudioTranscriptionConfigMode["MODE_UNSPECIFIED"] = "MODE_UNSPECIFIED";
+    /**
+     * Verbatim transcription mode.
+     */
+    AudioTranscriptionConfigMode["VERBATIM"] = "VERBATIM";
+    /**
+     * Smart transcription mode.
+     */
+    AudioTranscriptionConfigMode["SMART"] = "SMART";
+})(AudioTranscriptionConfigMode || (AudioTranscriptionConfigMode = {}));
 /** Scale of the generated music. */
 var Scale;
 (function (Scale) {
@@ -6153,13 +6189,20 @@ class NodeTokenizerCache {
     constructor() {
         this.cacheDir = path__namespace.join(os__namespace.tmpdir(), 'vertexai_tokenizer_model');
     }
+    // Narrower than `TokenizerCache.load`, which returns `Uint8Array | null` so
+    // that the web implementation can satisfy it too. `Buffer` is a `Uint8Array`,
+    // so this still implements the interface, and stating it here keeps the
+    // native-decode path below from looking like an accident.
     async load(cacheKey, expectedHash) {
         const filePath = path__namespace.join(this.cacheDir, cacheKey);
         try {
             const data = await fs__namespace.readFile(filePath);
             const hash = crypto__namespace.createHash('sha256').update(data).digest('hex');
             if (hash === expectedHash) {
-                return new Uint8Array(data);
+                // Returned as-is rather than copied into a plain `Uint8Array`: keeping
+                // the `Buffer` lets protobufjs select its `BufferReader`, which decodes
+                // strings with native `utf8Slice`.
+                return data;
             }
             await this.removeFile(filePath);
             return null;
@@ -6196,13 +6239,17 @@ class NodeTokenizerCache {
  * Node.js implementation of tokenizer file system operations.
  */
 class NodeTokenizerFileSystem {
+    // Narrower than `TokenizerFileSystem.fetchFromUrl` for the same reason as
+    // `NodeTokenizerCache.load` above.
     async fetchFromUrl(url) {
         const response = await fetch(url);
         if (!response.ok) {
             throw new Error(`Failed to fetch tokenizer model from ${url}: ${response.statusText}`);
         }
         const arrayBuffer = await response.arrayBuffer();
-        return new Uint8Array(arrayBuffer);
+        // `Buffer.from` rather than `new Uint8Array` for the same reason as in
+        // `NodeTokenizerCache.load`: it keeps protobufjs on its native decode path.
+        return Buffer.from(arrayBuffer);
     }
     async validateHash(data, expectedHash) {
         const hash = crypto__namespace.createHash('sha256').update(data).digest('hex');

package/node_modules/@google/genai/dist/tokenizer/node.d.ts

@@ -134,6 +134,24 @@ declare interface AudioTranscriptionConfig {
     /** Configures speaker diarization.
      */
     diarization?: boolean;
+    /** Optional. Configures transcription mode. Supported values: `VERBATIM`, `SMART`. If unspecified, defaults to `VERBATIM` transcription. In `SMART` mode, the model performs disfluency removal (eliminating filler words, repetitions, and false starts), light grammatical cleanup, automatic formatting (paragraphs, bullet points, numbered lists), and minor user edits (inline self-corrections). Timestamps and diarization are incompatible with mode `SMART`. */
+    mode?: AudioTranscriptionConfigMode;
+}
+
+/** Transcription mode. */
+declare enum AudioTranscriptionConfigMode {
+    /**
+     * Unspecified transcription mode.
+     */
+    MODE_UNSPECIFIED = "MODE_UNSPECIFIED",
+    /**
+     * Verbatim transcription mode.
+     */
+    VERBATIM = "VERBATIM",
+    /**
+     * Smart transcription mode.
+     */
+    SMART = "SMART"
 }
 
 /** The authentication config to access the API. */
@@ -845,6 +863,22 @@ declare interface McpServer {
     streamableHttpTransport?: StreamableHttpTransport;
 }
 
+/** How the model processes input media for understanding. */
+declare enum MediaProcessing {
+    /**
+     * Default. Uses model-specific processing
+     */
+    MEDIA_PROCESSING_UNSPECIFIED = "MEDIA_PROCESSING_UNSPECIFIED",
+    /**
+     * Fixed-rate frame extraction. All frames placed in context.
+     */
+    STATIC = "STATIC",
+    /**
+     * Model-driven dynamic navigation. Recommended for most use cases.
+     */
+    AGENTIC = "AGENTIC"
+}
+
 /** The media resolution to use. */
 declare enum MediaResolution {
     /**
@@ -959,6 +993,8 @@ declare interface Part {
     videoMetadata?: VideoMetadata;
     /** Custom metadata associated with the Part. Agents using genai.Part as content representation may need to keep track of the additional information. For example it can be name of a file/source from which the Part originates or a way to multiplex multiple Part streams. This field is not supported in Vertex AI. */
     partMetadata?: Record<string, unknown>;
+    /** How the model processes this part's media for understanding. */
+    mediaProcessing?: MediaProcessing;
 }
 
 /** Partial argument value of the function call. This data type is not supported in Gemini API. */
@@ -1452,7 +1488,11 @@ declare enum ToolType {
     /**
      * File search tool, maps to Tool.file_search.
      */
-    FILE_SEARCH = "FILE_SEARCH"
+    FILE_SEARCH = "FILE_SEARCH",
+    /**
+     * Media processing tool.
+     */
+    MEDIA_PROCESSING = "MEDIA_PROCESSING"
 }
 
 /** Audio transcription in Server Content. */
@@ -1473,9 +1513,9 @@ declare interface Transcription {
 
 /** Config for stream translation. */
 declare interface TranslationConfig {
-    /** Optional. If true, the model will generate audio when the target language is spoken, essentially it will parrot the input. If false, we will not produce audio for the target language. */
+    /** If true, the model will generate audio when the target language is spoken, essentially it will parrot the input. If false, we will not produce audio for the target language. */
     echoTargetLanguage?: boolean;
-    /** Required. The target language for translation. Supported values are BCP-47 language codes (e.g. "en", "es", "fr"). */
+    /** The target language for translation. Supported values are BCP-47 language codes (e.g. "en", "es", "fr"). */
     targetLanguageCode?: string;
 }
 

package/node_modules/@google/genai/dist/tokenizer/node.mjs

@@ -1217,6 +1217,22 @@ var ServiceTier;
      */
     ServiceTier["PRIORITY"] = "priority";
 })(ServiceTier || (ServiceTier = {}));
+/** How the model processes input media for understanding. */
+var MediaProcessing;
+(function (MediaProcessing) {
+    /**
+     * Default. Uses model-specific processing
+     */
+    MediaProcessing["MEDIA_PROCESSING_UNSPECIFIED"] = "MEDIA_PROCESSING_UNSPECIFIED";
+    /**
+     * Fixed-rate frame extraction. All frames placed in context.
+     */
+    MediaProcessing["STATIC"] = "STATIC";
+    /**
+     * Model-driven dynamic navigation. Recommended for most use cases.
+     */
+    MediaProcessing["AGENTIC"] = "AGENTIC";
+})(MediaProcessing || (MediaProcessing = {}));
 /** The tokenization quality used for given media. */
 var PartMediaResolutionLevel;
 (function (PartMediaResolutionLevel) {
@@ -1268,6 +1284,10 @@ var ToolType;
      * File search tool, maps to Tool.file_search.
      */
     ToolType["FILE_SEARCH"] = "FILE_SEARCH";
+    /**
+     * Media processing tool.
+     */
+    ToolType["MEDIA_PROCESSING"] = "MEDIA_PROCESSING";
 })(ToolType || (ToolType = {}));
 /** Resource scope. */
 var ResourceScope;
@@ -1733,6 +1753,22 @@ var TurnCoverage;
      */
     TurnCoverage["TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO"] = "TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO";
 })(TurnCoverage || (TurnCoverage = {}));
+/** Transcription mode. */
+var AudioTranscriptionConfigMode;
+(function (AudioTranscriptionConfigMode) {
+    /**
+     * Unspecified transcription mode.
+     */
+    AudioTranscriptionConfigMode["MODE_UNSPECIFIED"] = "MODE_UNSPECIFIED";
+    /**
+     * Verbatim transcription mode.
+     */
+    AudioTranscriptionConfigMode["VERBATIM"] = "VERBATIM";
+    /**
+     * Smart transcription mode.
+     */
+    AudioTranscriptionConfigMode["SMART"] = "SMART";
+})(AudioTranscriptionConfigMode || (AudioTranscriptionConfigMode = {}));
 /** Scale of the generated music. */
 var Scale;
 (function (Scale) {
@@ -6129,13 +6165,20 @@ class NodeTokenizerCache {
     constructor() {
         this.cacheDir = path.join(os.tmpdir(), 'vertexai_tokenizer_model');
     }
+    // Narrower than `TokenizerCache.load`, which returns `Uint8Array | null` so
+    // that the web implementation can satisfy it too. `Buffer` is a `Uint8Array`,
+    // so this still implements the interface, and stating it here keeps the
+    // native-decode path below from looking like an accident.
     async load(cacheKey, expectedHash) {
         const filePath = path.join(this.cacheDir, cacheKey);
         try {
             const data = await fs.readFile(filePath);
             const hash = crypto.createHash('sha256').update(data).digest('hex');
             if (hash === expectedHash) {
-                return new Uint8Array(data);
+                // Returned as-is rather than copied into a plain `Uint8Array`: keeping
+                // the `Buffer` lets protobufjs select its `BufferReader`, which decodes
+                // strings with native `utf8Slice`.
+                return data;
             }
             await this.removeFile(filePath);
             return null;
@@ -6172,13 +6215,17 @@ class NodeTokenizerCache {
  * Node.js implementation of tokenizer file system operations.
  */
 class NodeTokenizerFileSystem {
+    // Narrower than `TokenizerFileSystem.fetchFromUrl` for the same reason as
+    // `NodeTokenizerCache.load` above.
     async fetchFromUrl(url) {
         const response = await fetch(url);
         if (!response.ok) {
             throw new Error(`Failed to fetch tokenizer model from ${url}: ${response.statusText}`);
         }
         const arrayBuffer = await response.arrayBuffer();
-        return new Uint8Array(arrayBuffer);
+        // `Buffer.from` rather than `new Uint8Array` for the same reason as in
+        // `NodeTokenizerCache.load`: it keeps protobufjs on its native decode path.
+        return Buffer.from(arrayBuffer);
     }
     async validateHash(data, expectedHash) {
         const hash = crypto.createHash('sha256').update(data).digest('hex');