@qvac/transcription-whispercpp 0.6.1 → 0.6.3

package/README.md

@@ -128,16 +128,16 @@ $env:VCPKG_ROOT = "C:\path\to\vcpkg"
 
 **Linux:**
 ```bash
-# Ubuntu/Debian
+# Ubuntu/Debian — includes Clang and libc++ required by the native addon
 sudo apt update
-sudo apt install build-essential cmake git pkg-config
+sudo apt install clang libc++-dev libc++abi-dev build-essential cmake git pkg-config
 
 # CentOS/RHEL/Fedora
 sudo yum groupinstall "Development Tools"
-sudo yum install cmake git pkgconfig
+sudo yum install cmake git pkgconfig clang
 # or for newer versions:
 sudo dnf groupinstall "Development Tools"
-sudo dnf install cmake git pkgconfig
+sudo dnf install cmake git pkgconfig clang
 ```
 
 **macOS:**
@@ -206,6 +206,18 @@ This command runs the complete build sequence:
 2. `bare-make build` - Compiles the native C++ addon
 3. `bare-make install` - Installs the built addon
 
+#### Building with Vulkan GPU Acceleration
+
+On Linux, Android, and Windows, Vulkan support can be enabled at build time. Ensure the [Vulkan SDK](#gpu-acceleration-optional) is installed, then pass `-D ENABLE_VULKAN=ON` during the generate step:
+
+```bash
+bare-make generate -D ENABLE_VULKAN=ON
+bare-make build
+bare-make install
+```
+
+CI prebuilds for Linux, Android, and Windows include Vulkan by default. macOS and iOS use Metal instead.
+
 #### Running Tests
 
 After building, you can run the test suite:
@@ -261,6 +273,20 @@ Most users interact with the addon exclusively through `index.js`. From that ent
 | `whisperConfig` | *(any `whisper_full_params` key)* | Forwarded untouched. We surface convenience defaults in `index.js`, but every whisper.cpp flag is accepted—see [Advanced configuration](#advanced-configuration). |
 | `miscConfig` | `caption_enabled` | Formats segments with `<\|start\|>..<\|end\|>` markers |
 
+#### GPU acceleration is opt-in
+
+`use_gpu` defaults to `false`. To enable Vulkan (Linux/Windows/Android) or Metal (macOS/iOS) acceleration, set `use_gpu: true` explicitly in `contextParams`:
+
+```javascript
+const config = {
+  contextParams: {
+    model: './models/ggml-tiny.bin',
+    use_gpu: true,   // opt-in to GPU
+    gpu_device: 0
+  }
+}
+```
+
 #### Context keys that force a full reload
 
 Internally `WhisperModel::configContextIsChanged()` watches `model`, `use_gpu`, `flash_attn` and `gpu_device`. If any of these change we must:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@qvac/transcription-whispercpp",
-  "version": "0.6.1",
+  "version": "0.6.3",
   "description": "transcription addon for qvac",
   "addon": true,
   "engines": {
@@ -12,6 +12,7 @@
     "build": "bare-make generate && bare-make build && bare-make install",
     "test:unit": "brittle-bare test/unit/**/*.test.js",
     "test:integration": "brittle-bare test/integration/addon.test.js",
+    "test:benchmark:rtf": "brittle-bare test/benchmark/rtf-benchmark.test.js",
     "test:integration:long": "brittle-bare test/integration/**/longES.test.js",
     "test:cpp:build": "bare-make generate -D BUILD_TESTING=ON && bare-make build --target test-whisper-core && bare-make install",
     "test:cpp:run": "cd build/addon/tests/ && LLVM_PROFILE_FILE=default.profraw ./test-whisper-core --gtest_output=xml:cpp-test-results.xml",

package/prebuilds/darwin-arm64/qvac__transcription-whispercpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.6.1"
+        "version": "0.6.3"
       }
     ],
     "exported_symbols": [

package/prebuilds/darwin-x64/qvac__transcription-whispercpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.6.1"
+        "version": "0.6.3"
       }
     ],
     "exported_symbols": [
@@ -58,6 +58,7 @@
             "__ZTSNSt3__120__shared_ptr_emplaceIN32qvac_lib_inference_addon_whisper25JsTranscriptOutputHandlerENS_9allocatorIS2_EEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp16OutputCallBackJsE",
             "__ZTINSt3__111__end_stateIcEE",
+            "__ZTSNSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS2_10TranscriptEE_NS_9allocatorISA_EEFvS9_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp6Output5ErrorE",
             "__ZTSNSt3__120__shared_ptr_emplaceIN28qvac_lib_inference_addon_cpp6logger8JsLogger5StateENS_9allocatorIS4_EEEE",
             "__ZTIN32qvac_lib_inference_addon_whisper10TranscriptE",
@@ -142,6 +143,7 @@
             "__ZTSNSt3__111__end_stateIcEE",
             "__ZTSN8js_blobs15FinalizedStreamIcEE",
             "__ZTSNSt3__120__shared_ptr_emplaceIN28qvac_lib_inference_addon_cpp9out_handl21JsLogMsgOutputHandlerENS_9allocatorIS3_EEEE",
+            "__ZTSZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS_10TranscriptEE_",
             "__ZTINSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper30JsTranscriptArrayOutputHandlerC1EvEUlRKNS_6vectorINS2_10TranscriptENS_9allocatorIS5_EEEEE_NS6_ISB_EEFP10js_value_sSA_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp5model12IModelCancelE",
             "__ZTSN28qvac_lib_inference_addon_cpp9out_handl17BaseOutputHandlerIP10js_value_sN32qvac_lib_inference_addon_whisper10TranscriptEEE",
@@ -155,6 +157,8 @@
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl17BaseOutputHandlerIP10js_value_sNS_6Output6LogMsgEEE",
             "__ZTSNSt3__111__match_anyIcEE",
             "__ZTINSt3__121__empty_non_own_stateIcEE",
+            "__ZTSNSt3__110__function6__baseIFvRKN32qvac_lib_inference_addon_whisper10TranscriptEEEE",
+            "__ZTIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS_10TranscriptEE_",
             "__ZTSNSt3__120__shared_ptr_pointerIPNS_13__empty_stateIcEENS_10shared_ptrIS2_E27__shared_ptr_default_deleteIS2_S2_EENS_9allocatorIS2_EEEE",
             "__ZN32qvac_lib_inference_addon_whisper19g_streamingSessionsE",
             "__ZTSN4ggml3cpu6repack13tensor_traitsI10block_q8_0Lx8ELx4EL9ggml_type8EEE",
@@ -268,11 +272,13 @@
             "__ZTSNSt3__110__function6__baseIFP10js_value_sRKNS_6vectorINS_4pairINS_12basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEEENS_7variantIJdxEEEEENS9_ISE_EEEEEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp9out_handl19JsBaseOutputHandlerINSt3__16vectorIN32qvac_lib_inference_addon_whisper10TranscriptENS2_9allocatorIS5_EEEEEE",
             "__ZTSNSt3__16vectorINS_4pairINS_12basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEEENS_7variantIJdxEEEEENS5_ISA_EEEE",
+            "__ZTINSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS2_10TranscriptEE_NS_9allocatorISA_EEFvS9_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl27JsRuntimeStatsOutputHandlerE",
             "__ZTSN4ggml3cpu6repack18tensor_traits_baseE",
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl22OutputHandlerInterfaceIP10js_value_sEE",
             "__ZTINSt3__120__match_char_collateIcNS_12regex_traitsIcEEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp2js16UniqueRefDeleterE",
+            "__ZTINSt3__110__function6__baseIFvRKN32qvac_lib_inference_addon_whisper10TranscriptEEEE",
             "__ZTIN28qvac_lib_inference_addon_cpp16OutputCallBackJsE",
             "__ZTIN4ggml3cpu6repack13tensor_traitsI10block_q8_0Lx4ELx4EL9ggml_type8EEE",
             "__ZTINSt3__110__function6__baseIFP11ggml_cgraphvEEE",
@@ -912,16 +918,16 @@
             "_ggml_metal_pipeline_init",
             "__ZN4ggml3cpu6repack4gemvI10block_q6_KLx4ELx8EL9ggml_type15EEEviPfmPKvS7_ii",
             "__ZN4ggml3cpu6repack4gemvI12block_iq4_nlLx4ELx4EL9ggml_type8EEEviPfmPKvS7_ii",
-            "_ggml_new_tensor_1d",
             "_ggml_op_name",
+            "_ggml_vec_dot_iq2_s_q8_K",
             "_dequantize_row_iq3_s",
             "__ZN4ggml3cpu6repack4gemmI10block_q2_KLx8ELx8EL9ggml_type15EEEviPfmPKvS7_ii",
             "_ggml_add_cast",
             "_ggml_metal_graph_compute",
             "_ggml_backend_buffer_get_alignment",
-            "_whisper_vad_free_segments",
+            "_ggml_new_tensor_1d",
             "_ggml_sqr",
-            "_ggml_vec_dot_iq2_s_q8_K",
+            "_whisper_vad_free_segments",
             "_ggml_compute_forward_timestep_embedding",
             "_ggml_metal_cv_set_int16",
             "_ggml_set_i32_nd",

package/prebuilds/ios-arm64/qvac__transcription-whispercpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.6.1"
+        "version": "0.6.3"
       }
     ],
     "exported_symbols": [

package/prebuilds/ios-arm64-simulator/qvac__transcription-whispercpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.6.1"
+        "version": "0.6.3"
       }
     ],
     "exported_symbols": [

package/prebuilds/ios-x64-simulator/qvac__transcription-whispercpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.6.1"
+        "version": "0.6.3"
       }
     ],
     "exported_symbols": [
@@ -59,6 +59,7 @@
             "__ZTSNSt3__120__shared_ptr_emplaceIN32qvac_lib_inference_addon_whisper25JsTranscriptOutputHandlerENS_9allocatorIS2_EEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp16OutputCallBackJsE",
             "__ZTINSt3__111__end_stateIcEE",
+            "__ZTSNSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS2_10TranscriptEE_NS_9allocatorISA_EEFvS9_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp6Output5ErrorE",
             "__ZTSNSt3__120__shared_ptr_emplaceIN28qvac_lib_inference_addon_cpp6logger8JsLogger5StateENS_9allocatorIS4_EEEE",
             "__ZTIN32qvac_lib_inference_addon_whisper10TranscriptE",
@@ -145,6 +146,7 @@
             "__ZTSNSt3__111__end_stateIcEE",
             "__ZTSN8js_blobs15FinalizedStreamIcEE",
             "__ZTSNSt3__120__shared_ptr_emplaceIN28qvac_lib_inference_addon_cpp9out_handl21JsLogMsgOutputHandlerENS_9allocatorIS3_EEEE",
+            "__ZTSZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS_10TranscriptEE_",
             "__ZTINSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper30JsTranscriptArrayOutputHandlerC1EvEUlRKNS_6vectorINS2_10TranscriptENS_9allocatorIS5_EEEEE_NS6_ISB_EEFP10js_value_sSA_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp5model12IModelCancelE",
             "__ZTSN28qvac_lib_inference_addon_cpp9out_handl17BaseOutputHandlerIP10js_value_sN32qvac_lib_inference_addon_whisper10TranscriptEEE",
@@ -158,6 +160,8 @@
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl17BaseOutputHandlerIP10js_value_sNS_6Output6LogMsgEEE",
             "__ZTSNSt3__111__match_anyIcEE",
             "__ZTINSt3__121__empty_non_own_stateIcEE",
+            "__ZTSNSt3__110__function6__baseIFvRKN32qvac_lib_inference_addon_whisper10TranscriptEEEE",
+            "__ZTIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS_10TranscriptEE_",
             "__ZTSNSt3__120__shared_ptr_pointerIPNS_13__empty_stateIcEENS_10shared_ptrIS2_E27__shared_ptr_default_deleteIS2_S2_EENS_9allocatorIS2_EEEE",
             "__ZN32qvac_lib_inference_addon_whisper19g_streamingSessionsE",
             "__ZTSN4ggml3cpu6repack13tensor_traitsI10block_q8_0Lx8ELx4EL9ggml_type8EEE",
@@ -272,11 +276,13 @@
             "__ZTSNSt3__110__function6__baseIFP10js_value_sRKNS_6vectorINS_4pairINS_12basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEEENS_7variantIJdxEEEEENS9_ISE_EEEEEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp9out_handl19JsBaseOutputHandlerINSt3__16vectorIN32qvac_lib_inference_addon_whisper10TranscriptENS2_9allocatorIS5_EEEEEE",
             "__ZTSNSt3__16vectorINS_4pairINS_12basic_stringIcNS_11char_traitsIcEENS_9allocatorIcEEEENS_7variantIJdxEEEEENS5_ISA_EEEE",
+            "__ZTINSt3__110__function6__funcIZN32qvac_lib_inference_addon_whisper6runJobEP8js_env_sP18js_callback_info_sEUlRKNS2_10TranscriptEE_NS_9allocatorISA_EEFvS9_EEE",
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl27JsRuntimeStatsOutputHandlerE",
             "__ZTSN4ggml3cpu6repack18tensor_traits_baseE",
             "__ZTIN28qvac_lib_inference_addon_cpp9out_handl22OutputHandlerInterfaceIP10js_value_sEE",
             "__ZTINSt3__120__match_char_collateIcNS_12regex_traitsIcEEEE",
             "__ZTSN28qvac_lib_inference_addon_cpp2js16UniqueRefDeleterE",
+            "__ZTINSt3__110__function6__baseIFvRKN32qvac_lib_inference_addon_whisper10TranscriptEEEE",
             "__ZTIN28qvac_lib_inference_addon_cpp16OutputCallBackJsE",
             "__ZTIN4ggml3cpu6repack13tensor_traitsI10block_q8_0Lx4ELx4EL9ggml_type8EEE",
             "__ZTINSt3__110__function6__baseIFP11ggml_cgraphvEEE",
@@ -916,16 +922,16 @@
             "_ggml_metal_pipeline_init",
             "__ZN4ggml3cpu6repack4gemvI10block_q6_KLx4ELx8EL9ggml_type15EEEviPfmPKvS7_ii",
             "__ZN4ggml3cpu6repack4gemvI12block_iq4_nlLx4ELx4EL9ggml_type8EEEviPfmPKvS7_ii",
-            "_ggml_new_tensor_1d",
             "_ggml_op_name",
+            "_ggml_vec_dot_iq2_s_q8_K",
             "_dequantize_row_iq3_s",
             "__ZN4ggml3cpu6repack4gemmI10block_q2_KLx8ELx8EL9ggml_type15EEEviPfmPKvS7_ii",
             "_ggml_add_cast",
             "_ggml_metal_graph_compute",
             "_ggml_backend_buffer_get_alignment",
-            "_whisper_vad_free_segments",
+            "_ggml_new_tensor_1d",
             "_ggml_sqr",
-            "_ggml_vec_dot_iq2_s_q8_K",
+            "_whisper_vad_free_segments",
             "_ggml_compute_forward_timestep_embedding",
             "_ggml_metal_cv_set_int16",
             "_ggml_set_i32_nd",

package/whisper.js

@@ -90,6 +90,22 @@ class WhisperInterface {
 
     if (mappedEvent === 'Output') {
       this._setState(state.PROCESSING)
+      if (this._outputCb != null) {
+        const isTranscriptArray = Array.isArray(data) && data.length > 0 &&
+          typeof data[0]?.text === 'string'
+        const isSingleTranscript = !Array.isArray(data) &&
+          data && typeof data === 'object' && typeof data.text === 'string'
+        if (isTranscriptArray) {
+          for (const segment of data) {
+            this._outputCb(addon, 'Output', jobId, [segment], null)
+          }
+        } else if (isSingleTranscript) {
+          this._outputCb(addon, 'Output', jobId, [data], null)
+        } else {
+          this._outputCb(addon, 'Output', jobId, data, null)
+        }
+      }
+      return
     }
 
     if (this._outputCb != null) {