@qvac/sdk 0.9.2 → 0.10.0

package/dist/schemas/transcription-config.d.ts

@@ -76,6 +76,7 @@ export declare const whisperConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -86,7 +87,6 @@ export declare const whisperConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -139,6 +139,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -149,7 +150,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -169,6 +169,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -179,7 +180,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -199,6 +199,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -209,7 +210,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -229,6 +229,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -239,7 +240,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -259,6 +259,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -269,7 +270,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -289,6 +289,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -299,7 +300,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";
@@ -319,6 +319,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
         expectedSize: z.ZodOptional<z.ZodNumber>;
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -329,7 +330,6 @@ export declare const parakeetConfigSchema: z.ZodObject<{
             "sdcpp-generation": "sdcpp-generation";
             llm: "llm";
             whisper: "whisper";
-            embeddings: "embeddings";
             nmt: "nmt";
             parakeet: "parakeet";
             tts: "tts";

package/dist/schemas/transcription.d.ts

@@ -9,6 +9,7 @@ export declare const audioInputSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
 export declare const transcribeParamsSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodOptional<z.ZodString>;
+    metadata: z.ZodOptional<z.ZodBoolean>;
     audioChunk: z.ZodDiscriminatedUnion<[z.ZodObject<{
         type: z.ZodLiteral<"base64">;
         value: z.ZodString;
@@ -29,9 +30,17 @@ export declare const transcribeStatsSchema: z.ZodObject<{
     decoderTime: z.ZodOptional<z.ZodNumber>;
     melSpecTime: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
+export declare const transcribeSegmentSchema: z.ZodObject<{
+    text: z.ZodString;
+    startMs: z.ZodNumber;
+    endMs: z.ZodNumber;
+    append: z.ZodBoolean;
+    id: z.ZodNumber;
+}, z.core.$strip>;
 export declare const transcribeRequestSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodOptional<z.ZodString>;
+    metadata: z.ZodOptional<z.ZodBoolean>;
     audioChunk: z.ZodDiscriminatedUnion<[z.ZodObject<{
         type: z.ZodLiteral<"base64">;
         value: z.ZodString;
@@ -57,20 +66,30 @@ export declare const transcribeResponseSchema: z.ZodObject<{
         melSpecTime: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     error: z.ZodOptional<z.ZodString>;
+    segment: z.ZodOptional<z.ZodObject<{
+        text: z.ZodString;
+        startMs: z.ZodNumber;
+        endMs: z.ZodNumber;
+        append: z.ZodBoolean;
+        id: z.ZodNumber;
+    }, z.core.$strip>>;
     type: z.ZodLiteral<"transcribe">;
 }, z.core.$strip>;
 export type AudioInput = z.infer<typeof audioInputSchema>;
 export type TranscribeParams = z.infer<typeof transcribeParamsSchema>;
+export type TranscribeSegment = z.infer<typeof transcribeSegmentSchema>;
 export type TranscribeClientParams = {
     modelId: string;
     audioChunk: string | Buffer;
     prompt?: string;
+    metadata?: boolean;
 };
 export type TranscribeRequest = z.infer<typeof transcribeRequestSchema>;
 export type TranscribeResponse = z.infer<typeof transcribeResponseSchema>;
 export declare const transcribeStreamRequestSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodOptional<z.ZodString>;
+    metadata: z.ZodOptional<z.ZodBoolean>;
     type: z.ZodLiteral<"transcribeStream">;
 }, z.core.$strip>;
 export declare const transcribeStreamResponseSchema: z.ZodObject<{
@@ -89,6 +108,13 @@ export declare const transcribeStreamResponseSchema: z.ZodObject<{
         melSpecTime: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     error: z.ZodOptional<z.ZodString>;
+    segment: z.ZodOptional<z.ZodObject<{
+        text: z.ZodString;
+        startMs: z.ZodNumber;
+        endMs: z.ZodNumber;
+        append: z.ZodBoolean;
+        id: z.ZodNumber;
+    }, z.core.$strip>>;
     type: z.ZodLiteral<"transcribeStream">;
 }, z.core.$strip>;
 export type TranscribeStreamRequest = z.infer<typeof transcribeStreamRequestSchema>;
@@ -96,6 +122,7 @@ export type TranscribeStreamResponse = z.infer<typeof transcribeStreamResponseSc
 export type TranscribeStreamClientParams = {
     modelId: string;
     prompt?: string;
+    metadata?: boolean;
 };
 export interface TranscribeStreamSession {
     write(audioChunk: Buffer): void;
@@ -103,5 +130,11 @@ export interface TranscribeStreamSession {
     destroy(): void;
     [Symbol.asyncIterator](): AsyncIterator<string>;
 }
+export interface TranscribeStreamMetadataSession {
+    write(audioChunk: Buffer): void;
+    end(): void;
+    destroy(): void;
+    [Symbol.asyncIterator](): AsyncIterator<TranscribeSegment>;
+}
 export type TranscribeStats = z.infer<typeof transcribeStatsSchema>;
 //# sourceMappingURL=transcription.d.ts.map

package/dist/schemas/transcription.js

@@ -12,6 +12,7 @@ export const audioInputSchema = z.discriminatedUnion("type", [
 const transcribeBaseSchema = z.object({
     modelId: z.string(),
     prompt: z.string().optional(),
+    metadata: z.boolean().optional(),
 });
 export const transcribeParamsSchema = transcribeBaseSchema.extend({
     audioChunk: audioInputSchema,
@@ -28,6 +29,13 @@ export const transcribeStatsSchema = z.object({
     decoderTime: z.number().optional(),
     melSpecTime: z.number().optional(),
 });
+export const transcribeSegmentSchema = z.object({
+    text: z.string(),
+    startMs: z.number(),
+    endMs: z.number(),
+    append: z.boolean(),
+    id: z.number(),
+});
 export const transcribeRequestSchema = transcribeParamsSchema.extend({
     type: z.literal("transcribe"),
 });
@@ -36,6 +44,7 @@ const transcriptionResultBase = z.object({
     done: z.boolean().optional(),
     stats: transcribeStatsSchema.optional(),
     error: z.string().optional(),
+    segment: transcribeSegmentSchema.optional(),
 });
 export const transcribeResponseSchema = transcriptionResultBase.extend({
     type: z.literal("transcribe"),

package/dist/schemas/translate.js

@@ -1,24 +1,43 @@
 import { z } from "zod";
 import { nmtModelTypeSchema, llmModelTypeSchema, normalizeModelType, } from "./model-types.js";
 const translateParamsNmtSchema = z.object({
-    modelId: z.string(),
-    text: z.union([
+    modelId: z
+        .string()
+        .describe("The identifier of the NMT translation model to use."),
+    text: z
+        .union([
         z.string().min(1, "Text cannot be empty"),
         z
             .array(z.string().min(1, "Text cannot be empty"))
             .min(1, "Array cannot be empty"),
-    ]),
-    stream: z.boolean(),
-    modelType: nmtModelTypeSchema,
+    ])
+        .describe("The input text(s) to translate. A single string returns a single translation; an array returns one translation per input."),
+    stream: z
+        .boolean()
+        .describe("Whether to stream tokens (`true`) or resolve the complete translation once (`false`)."),
+    modelType: nmtModelTypeSchema.describe("NMT model-type variant identifier."),
 });
 const translateParamsLlmSchema = z.object({
-    modelId: z.string(),
-    text: z.string().min(1, "Text cannot be empty"),
-    stream: z.boolean(),
-    modelType: llmModelTypeSchema,
-    from: z.string().optional(),
-    to: z.string(),
-    context: z.string().optional(),
+    modelId: z
+        .string()
+        .describe("The identifier of the LLM model used for translation."),
+    text: z
+        .string()
+        .min(1, "Text cannot be empty")
+        .describe("The input text to translate."),
+    stream: z
+        .boolean()
+        .describe("Whether to stream tokens (`true`) or resolve the complete translation once (`false`)."),
+    modelType: llmModelTypeSchema.describe("LLM model-type variant identifier."),
+    from: z
+        .string()
+        .optional()
+        .describe("Source language code. When omitted, the SDK attempts to auto-detect the source language."),
+    to: z.string().describe("Target language code."),
+    context: z
+        .string()
+        .optional()
+        .describe("Optional translation context passed to the LLM as a system hint."),
 });
 // Using z.union since each modelType accepts multiple values
 const translateParamsSchema = z.union([
@@ -27,15 +46,36 @@ const translateParamsSchema = z.union([
 ]);
 export const translationStatsSchema = z.object({
     // Common stats
-    totalTime: z.number().optional(),
-    totalTokens: z.number().optional(),
-    tokensPerSecond: z.number().optional(),
-    timeToFirstToken: z.number().optional(),
+    totalTime: z
+        .number()
+        .optional()
+        .describe("Total translation time in milliseconds."),
+    totalTokens: z
+        .number()
+        .optional()
+        .describe("Total tokens produced by the translation."),
+    tokensPerSecond: z
+        .number()
+        .optional()
+        .describe("Tokens generated per second."),
+    timeToFirstToken: z
+        .number()
+        .optional()
+        .describe("Time to first token in milliseconds (LLM translation only)."),
     // NMT-specific
-    decodeTime: z.number().optional(),
-    encodeTime: z.number().optional(),
+    decodeTime: z
+        .number()
+        .optional()
+        .describe("Time spent in the NMT decoder in milliseconds."),
+    encodeTime: z
+        .number()
+        .optional()
+        .describe("Time spent in the NMT encoder in milliseconds."),
     // LLM-specific
-    cacheTokens: z.number().optional(),
+    cacheTokens: z
+        .number()
+        .optional()
+        .describe("Tokens served from the KV cache during LLM translation."),
 });
 export const translateRequestSchema = z.union([
     translateParamsNmtSchema.extend({ type: z.literal("translate") }),