@qvac/sdk 0.7.1 → 0.8.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@qvac/sdk",
-  "version": "0.7.1",
+  "version": "0.8.0",
   "license": "Apache-2.0",
   "type": "module",
   "main": "./dist/index.js",
@@ -28,6 +28,10 @@
       "types": "./dist/server/bare/plugins/whispercpp-transcription/plugin.d.ts",
       "import": "./dist/server/bare/plugins/whispercpp-transcription/plugin.js"
     },
+    "./parakeet-transcription/plugin": {
+      "types": "./dist/server/bare/plugins/parakeet-transcription/plugin.d.ts",
+      "import": "./dist/server/bare/plugins/parakeet-transcription/plugin.js"
+    },
     "./nmtcpp-translation/plugin": {
       "types": "./dist/server/bare/plugins/nmtcpp-translation/plugin.d.ts",
       "import": "./dist/server/bare/plugins/nmtcpp-translation/plugin.js"
@@ -100,30 +104,26 @@
     "bare:ensureprebuilds": "bun run bare:ensureprebuilds:llamacpp && bun run bare:ensureprebuilds:whispercpp && bun run bare:ensureprebuilds:embeddings",
     "prepare:husky": "bun run husky",
     "bare:example": "bare ./scripts/bare-bootstrap.js",
-    "docs_api:website": "typedoc --out ./docs/api/website --entryPoints index.ts --entryPointStrategy expand --readme none",
-    "docs_api:markdown": "typedoc --plugin typedoc-plugin-markdown --out ./docs/api/markdown --entryPoints index.ts --entryPointStrategy expand --readme none && sed -i '1,4d' ./docs/api/markdown/README.md",
-    "docs_api": "bun run docs_api:website && bun run docs_api:markdown",
     "check-models": "tsx ./models/update-models/index.ts --check",
     "check-models:hook": "tsx ./models/update-models/index.ts --check --non-blocking",
     "update-models": "tsx ./models/update-models/index.ts",
-    "changelog:generate": "node ../../scripts/sdk/generate-changelog-sdk-pod.cjs --package=qvac-sdk && prettier --write changelog",
-    "docs:gen-pages": "remark \"docs/*.mdx\" --output .docs-build",
-    "docs:gen-api": "typedoc --options docs/typedoc.jsonc"
+    "changelog:generate": "node ../../scripts/sdk/generate-changelog-sdk-pod.cjs --package=sdk && prettier --write changelog"
   },
   "dependencies": {
     "@qvac/decoder-audio": "^0.3.3",
     "@qvac/dl-filesystem": "^0.2.0",
-    "@qvac/embed-llamacpp": "^0.10.7",
+    "@qvac/embed-llamacpp": "^0.12.0",
     "@qvac/error": "^0.1.1",
     "@qvac/langdetect-text": "^0.1.0",
-    "@qvac/llm-llamacpp": "^0.8.9",
+    "@qvac/llm-llamacpp": "^0.12.1",
     "@qvac/logging": "^0.1.0",
-    "@qvac/ocr-onnx": "^0.1.5",
+    "@qvac/ocr-onnx": "^0.2.0",
     "@qvac/rag": "^0.4.2",
-    "@qvac/transcription-whispercpp": "^0.3.17",
-    "@qvac/translation-nmtcpp": "^0.3.9",
-    "@qvac/tts-onnx": "^0.5.4",
     "@qvac/registry-client": "^0.2.1",
+    "@qvac/transcription-parakeet": "^0.2.0",
+    "@qvac/transcription-whispercpp": "^0.5.0",
+    "@qvac/translation-nmtcpp": "^0.6.1",
+    "@qvac/tts-onnx": "^0.6.1",
     "fast-safe-stringify": "2.1.1",
     "which-runtime": "^1.3.2",
     "zod": "^4.0.17"
@@ -159,9 +159,6 @@
     "remark-frontmatter": "^5.0.0",
     "remark-mdx": "^3.1.1",
     "tsx": "^4.21.0",
-    "typedoc": "^0.28.11",
-    "typedoc-plugin-frontmatter": "^1.3.1",
-    "typedoc-plugin-markdown": "^4.8.1",
     "typescript": "^5.9.2",
     "typescript-eslint": "^8.39.0"
   },

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

@@ -1,70 +0,0 @@
-import { z } from "zod";
-export declare const audioFormatSchema: z.ZodEnum<{
-    f32le: "f32le";
-    s16le: "s16le";
-}>;
-export type AudioFormat = z.infer<typeof audioFormatSchema>;
-export declare const whisperConfigSchema: z.ZodObject<{
-    strategy: z.ZodOptional<z.ZodEnum<{
-        greedy: "greedy";
-        beam_search: "beam_search";
-    }>>;
-    n_threads: z.ZodOptional<z.ZodNumber>;
-    n_max_text_ctx: z.ZodOptional<z.ZodNumber>;
-    offset_ms: z.ZodOptional<z.ZodNumber>;
-    duration_ms: z.ZodOptional<z.ZodNumber>;
-    audio_ctx: z.ZodOptional<z.ZodNumber>;
-    translate: z.ZodOptional<z.ZodBoolean>;
-    no_context: z.ZodOptional<z.ZodBoolean>;
-    no_timestamps: z.ZodOptional<z.ZodBoolean>;
-    single_segment: z.ZodOptional<z.ZodBoolean>;
-    print_special: z.ZodOptional<z.ZodBoolean>;
-    print_progress: z.ZodOptional<z.ZodBoolean>;
-    print_realtime: z.ZodOptional<z.ZodBoolean>;
-    print_timestamps: z.ZodOptional<z.ZodBoolean>;
-    token_timestamps: z.ZodOptional<z.ZodBoolean>;
-    thold_pt: z.ZodOptional<z.ZodNumber>;
-    thold_ptsum: z.ZodOptional<z.ZodNumber>;
-    max_len: z.ZodOptional<z.ZodNumber>;
-    split_on_word: z.ZodOptional<z.ZodBoolean>;
-    max_tokens: z.ZodOptional<z.ZodNumber>;
-    debug_mode: z.ZodOptional<z.ZodBoolean>;
-    tdrz_enable: z.ZodOptional<z.ZodBoolean>;
-    suppress_regex: z.ZodOptional<z.ZodString>;
-    initial_prompt: z.ZodOptional<z.ZodString>;
-    language: z.ZodOptional<z.ZodString>;
-    detect_language: z.ZodOptional<z.ZodBoolean>;
-    suppress_blank: z.ZodOptional<z.ZodBoolean>;
-    suppress_nst: z.ZodOptional<z.ZodBoolean>;
-    temperature: z.ZodOptional<z.ZodNumber>;
-    length_penalty: z.ZodOptional<z.ZodNumber>;
-    temperature_inc: z.ZodOptional<z.ZodNumber>;
-    entropy_thold: z.ZodOptional<z.ZodNumber>;
-    logprob_thold: z.ZodOptional<z.ZodNumber>;
-    greedy_best_of: z.ZodOptional<z.ZodNumber>;
-    beam_search_beam_size: z.ZodOptional<z.ZodNumber>;
-    vad_model_path: z.ZodOptional<z.ZodString>;
-    vad_params: z.ZodOptional<z.ZodObject<{
-        threshold: z.ZodOptional<z.ZodNumber>;
-        min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
-        min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
-        max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
-        speech_pad_ms: z.ZodOptional<z.ZodNumber>;
-        samples_overlap: z.ZodOptional<z.ZodNumber>;
-    }, z.core.$strip>>;
-    audio_format: z.ZodOptional<z.ZodEnum<{
-        f32le: "f32le";
-        s16le: "s16le";
-    }>>;
-    contextParams: z.ZodOptional<z.ZodObject<{
-        model: z.ZodOptional<z.ZodString>;
-        use_gpu: z.ZodOptional<z.ZodBoolean>;
-        flash_attn: z.ZodOptional<z.ZodBoolean>;
-        gpu_device: z.ZodOptional<z.ZodNumber>;
-    }, z.core.$strip>>;
-    miscConfig: z.ZodOptional<z.ZodObject<{
-        caption_enabled: z.ZodOptional<z.ZodBoolean>;
-    }, z.core.$strip>>;
-}, z.core.$strip>;
-export type WhisperConfig = z.infer<typeof whisperConfigSchema>;
-//# sourceMappingURL=whispercpp-config.d.ts.map

package/dist/server/bare/plugins/whispercpp-transcription/ops/transcribe-stream.d.ts

@@ -1,3 +0,0 @@
-import type { TranscribeParams } from "../../../../../schemas/index";
-export declare function transcribe(params: TranscribeParams): AsyncGenerator<string, void, void>;
-//# sourceMappingURL=transcribe-stream.d.ts.map

package/dist/server/bare/plugins/whispercpp-transcription/ops/transcribe-stream.js

@@ -1,81 +0,0 @@
-import { getModel, getModelConfig, } from "../../../registry/model-registry.js";
-import { Readable } from "bare-stream";
-import fs from "bare-fs";
-import { needsDecoding, decodeAudioToStream, } from "../../../../utils/audio/decoder.js";
-import { AudioFileNotFoundError, InvalidAudioChunkError, } from "../../../../../utils/errors-server.js";
-import { getServerLogger } from "../../../../../logging/index.js";
-const logger = getServerLogger();
-export async function* transcribe(params) {
-    const model = getModel(params.modelId);
-    const modelConfig = getModelConfig(params.modelId);
-    let originalConfig = null;
-    const audioFormat = modelConfig.audio_format || "s16le";
-    if (params.prompt && typeof model.reload === "function") {
-        originalConfig = modelConfig;
-        const updatedConfig = {
-            ...originalConfig,
-            initial_prompt: params.prompt,
-        };
-        // eslint-disable-next-line @typescript-eslint/no-unused-vars
-        const { contextParams: _, miscConfig, ...whisperParams } = updatedConfig;
-        await model.reload({
-            whisperConfig: whisperParams,
-            ...(miscConfig && { miscConfig }),
-        });
-    }
-    try {
-        let audioStream;
-        const audioChunk = params.audioChunk;
-        switch (audioChunk.type) {
-            case "base64": {
-                const audioBuffer = Buffer.from(audioChunk.value, "base64");
-                audioStream = Readable.from([audioBuffer]);
-                break;
-            }
-            case "filePath": {
-                const filePath = audioChunk.value;
-                try {
-                    fs.accessSync(filePath);
-                }
-                catch (error) {
-                    throw new AudioFileNotFoundError(filePath, error);
-                }
-                if (needsDecoding(filePath)) {
-                    audioStream = await decodeAudioToStream(filePath, audioFormat);
-                }
-                else {
-                    audioStream = fs.createReadStream(filePath);
-                }
-                break;
-            }
-            default:
-                throw new InvalidAudioChunkError();
-        }
-        // Run transcription with streaming enabled
-        const response = await model.run(audioStream);
-        for await (const output of response.iterate()) {
-            logger.debug("Streaming Transcription Update:", output);
-            // Filter out blank audio chunks and process the text
-            const text = output
-                .filter((chunk) => !chunk.text.includes("[BLANK_AUDIO]"))
-                .map((chunk) => chunk.text)
-                .join("");
-            if (text.trim()) {
-                yield text;
-            }
-        }
-    }
-    finally {
-        if (originalConfig && typeof model.reload === "function") {
-            // eslint-disable-next-line @typescript-eslint/no-unused-vars
-            const { contextParams: _, miscConfig, ...whisperParams } = originalConfig;
-            await model.reload({
-                whisperConfig: {
-                    ...whisperParams,
-                    initial_prompt: "",
-                },
-                ...(miscConfig && { miscConfig }),
-            });
-        }
-    }
-}