Picovoice · laves · Mar 17, 2025 · Mar 12, 2025 · Mar 12, 2025 · Mar 12, 2025
diff --git a/.github/workflows/android-tests.yml b/.github/workflows/android-tests.yml
@@ -36,7 +36,7 @@ jobs:
       run: ~/Android/Sdk/platform-tools/adb shell pm uninstall ai.picovoice.picollm.testapp.test || true
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Make Android dir
       run: ~/Android/Sdk/platform-tools/adb shell mkdir -p /sdcard/Android/data/ai.picovoice.picollm.testapp/files/external
@@ -89,7 +89,7 @@ jobs:
       run: ~/Android/Sdk/platform-tools/adb shell pm uninstall ai.picovoice.picollm.testapp.test || true
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Make Android dir
       run: ~/Android/Sdk/platform-tools/adb shell mkdir -p /sdcard/Android/data/ai.picovoice.picollm.testapp/files/external

diff --git a/.github/workflows/c-demos.yml b/.github/workflows/c-demos.yml
@@ -76,7 +76,7 @@ jobs:
           echo PATH=$PATH >> $GITHUB_ENV
 
       - name: Download resource files
-        run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+        run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
       - name: Create build directory
         run: cmake -G "${{ matrix.make_file }}" -B ./build

diff --git a/.github/workflows/dotnet-demos.yml b/.github/workflows/dotnet-demos.yml
@@ -40,7 +40,7 @@ jobs:
     - uses: actions/checkout@v3
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Package restore
       run: dotnet restore

diff --git a/.github/workflows/dotnet.yml b/.github/workflows/dotnet.yml
@@ -72,7 +72,7 @@ jobs:
     - uses: actions/checkout@v3
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
       working-directory: binding/dotnet
 
     - name: Build binding

diff --git a/.github/workflows/ios-tests.yml b/.github/workflows/ios-tests.yml
@@ -33,7 +33,7 @@ jobs:
           PicoLLMAppTestUITests/BaseTest.swift
 
       - name: Inject Resource URL
-        run: sed -i '.bak' 's?{TESTING_MODEL_URL_HERE}?http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm?'
+        run: sed -i '.bak' 's?{TESTING_MODEL_URL_HERE}?http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm?'
           PicoLLMAppTestUITests/BaseTest.swift
 
       - name: XCode Build

diff --git a/.github/workflows/nodejs-demos.yml b/.github/workflows/nodejs-demos.yml
@@ -51,7 +51,7 @@ jobs:
       run: yarn install
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Test
       run: yarn completion --access_key ${{secrets.PV_VALID_ACCESS_KEY}} --model_path phi2-290.pllm --completion_token_limit 10 --prompt "Hello my name is" --device ${{matrix.xpu}}
diff --git a/.github/workflows/nodejs.yml b/.github/workflows/nodejs.yml
@@ -48,7 +48,7 @@ jobs:
       run: yarn install
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o test/phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o test/phi2-290.pllm
 
     - name: Test
       run: yarn test test/index.test.ts --access_key=${{secrets.PV_VALID_ACCESS_KEY}} --device=${{matrix.xpu}}
diff --git a/.github/workflows/python-demos.yml b/.github/workflows/python-demos.yml
@@ -56,7 +56,7 @@ jobs:
       run: python3 -m pip install -r requirements.txt --force-reinstall
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Test
       run: python3 picollm_demo_completion.py --access_key ${{secrets.PV_VALID_ACCESS_KEY}} --model_path phi2-290.pllm --completion_token_limit 10 --prompt "Hello my name is" --device ${{matrix.xpu}}
diff --git a/.github/workflows/python.yml b/.github/workflows/python.yml
@@ -64,7 +64,7 @@ jobs:
       run: python3 -m pip install -r binding/python/requirements.txt
 
     - name: Download resource files
-      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm -o phi2-290.pllm
+      run: curl http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm -o phi2-290.pllm
 
     - name: Test
       run: python3 -m binding.python.test_picollm ${{secrets.PV_VALID_ACCESS_KEY}} ./phi2-290.pllm ${{matrix.xpu}}
diff --git a/.github/workflows/web.yml b/.github/workflows/web.yml
@@ -42,7 +42,7 @@ jobs:
       run: yarn build
 
     - name: Prepare Test
-      run: yarn setup-test http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/04-d5f2aa0/phi2-290.pllm phi2-290.pllm
+      run: yarn setup-test http://${{secrets.PV_CICD_RES_SERVER_AUTHORITY}}/github/picollm/res/phi2-290.pllm/05-fae87bc/phi2-290.pllm phi2-290.pllm
 
     - name: Test Web Worker
       run: yarn test --env ACCESS_KEY=${{secrets.PV_VALID_ACCESS_KEY}}
diff --git a/README.md b/README.md
@@ -467,6 +467,11 @@ pv_picollm_delete(pllm);
 
 ## Releases
 
+### v1.3.0 - March 14th, 2025
+
+- Performance improvements
+- Significant performance improvements for iOS and Web
+
 ### v1.2.0 - November 26th, 2024
 
 - Performance improvements

diff --git a/binding/android/PicoLLM/picollm/build.gradle b/binding/android/PicoLLM/picollm/build.gradle
@@ -2,7 +2,7 @@ apply plugin: 'com.android.library'
 
 ext {
     PUBLISH_GROUP_ID = 'ai.picovoice'
-    PUBLISH_VERSION = '1.2.3'
+    PUBLISH_VERSION = '1.3.0'
     PUBLISH_ARTIFACT_ID = 'picollm-android'
 }
 

diff --git a/binding/android/PicoLLMTestApp/picollm-test-app/build.gradle b/binding/android/PicoLLMTestApp/picollm-test-app/build.gradle
@@ -80,7 +80,7 @@ dependencies {
     implementation 'androidx.constraintlayout:constraintlayout:2.1.4'
     implementation 'com.google.code.gson:gson:2.10'
 
-    implementation 'ai.picovoice:picollm-android:1.2.3'
+    implementation 'ai.picovoice:picollm-android:1.3.0'
 
     // Espresso UI Testing
     androidTestImplementation 'androidx.test.ext:junit:1.1.5'

diff --git a/binding/dotnet/PicoLLM/PicoLLM.csproj b/binding/dotnet/PicoLLM/PicoLLM.csproj
@@ -1,7 +1,7 @@
 <Project Sdk="Microsoft.NET.Sdk">
 	<PropertyGroup>
 		<TargetFrameworks>net8.0;net6.0;netcoreapp3.0;netstandard2.0</TargetFrameworks>
-		<Version>1.2.1</Version>
+		<Version>1.3.0</Version>
 		<Authors>Picovoice</Authors>
 		<Company />
 		<Product>picoLLM Inference Engine</Product>

diff --git a/binding/ios/PicoLLMAppTest/PicoLLMAppTest.xcodeproj/project.pbxproj b/binding/ios/PicoLLMAppTest/PicoLLMAppTest.xcodeproj/project.pbxproj
@@ -563,7 +563,7 @@
 			repositoryURL = "https://github.com/picovoice/picollm";
 			requirement = {
 				kind = exactVersion;
-				version = 1.2.4;
+				version = 1.3.0;
 			};
 		};
 /* End XCRemoteSwiftPackageReference section */

diff --git a/binding/ios/picoLLM-iOS.podspec b/binding/ios/picoLLM-iOS.podspec
@@ -1,7 +1,7 @@
 Pod::Spec.new do |s|
   s.name = 'picoLLM-iOS'
   s.module_name = 'PicoLLM'
-  s.version = '1.2.4'
+  s.version = '1.3.0'
   s.license = {:type => 'Apache 2.0'}
   s.summary = 'picoLLM Inference Engine'
   s.description =

diff --git a/binding/nodejs/package.json b/binding/nodejs/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@picovoice/picollm-node",
-  "version": "1.2.5",
+  "version": "1.3.0",
   "description": "Picovoice picoLLM Node.js binding",
   "main": "dist/index.js",
   "types": "dist/types/index.d.ts",

diff --git a/binding/python/setup.py b/binding/python/setup.py
@@ -54,7 +54,7 @@
 
 setuptools.setup(
     name="picollm",
-    version="1.2.5",
+    version="1.3.0",
     author="Picovoice",
     author_email="hello@picovoice.ai",
     description="picoLLM Inference Engine",

diff --git a/binding/python/test_picollm.py b/binding/python/test_picollm.py
@@ -410,7 +410,7 @@ def test_reset(self) -> None:
 
         if 'gpu' in self._device:
             for x, y in zip(logits, self._picollm.forward(79)):
-                self.assertAlmostEqual(x, y, delta=0.01)
+                self.assertAlmostEqual(x, y, delta=0.02)
         else:
             self.assertListEqual(list(logits), list(self._picollm.forward(79)))
 

diff --git a/binding/web/.gitignore b/binding/web/.gitignore
@@ -1,7 +1,7 @@
 node_modules
 dist
-lib/*
-!lib/.gitkeep
+src/lib/*
+src/!lib/.gitkeep
 test/*.bin
 .idea/
 test_data.json

diff --git a/binding/web/README.md b/binding/web/README.md
@@ -13,13 +13,26 @@ models. picoLLM Inference Engine is:
 - Runs on CPU and GPU
 - Free for open-weight models
 
+## Requirements
+
+PicoLLM Web Binding uses [SharedArrayBuffer](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/SharedArrayBuffer) to generate
+text. Modern browsers require the following response headers to allow the usage of `SharedArrayBuffers`:
+
+```
+Cross-Origin-Opener-Policy: same-origin
+Cross-Origin-Embedder-Policy: require-corp
+```
+
+Refer to our [Web demos](https://github.com/Picovoice/picollm/tree/master/demo/web) for example on creating a server
+with the corresponding response headers.
+
 ## Compatibility
 
 - Chrome / Edge
 - Firefox
 - Safari
 
-**NOTE**: IndexedDB and SIMD are required to use `picoLLM`.
+**NOTE**: IndexedDB, SIMD and SharedArrayBuffers are required to use `picoLLM`.
 
 ## Installation
 

diff --git a/binding/web/cypress.config.ts b/binding/web/cypress.config.ts
@@ -8,5 +8,14 @@
     specPattern: 'test/*.test.{js,jsx,ts,tsx}',
     video: false,
     screenshotOnRunFailure: false,
+    setupNodeEvents(on, config) {
+      on('before:browser:launch', (browser, launchOptions) => {
+        if (browser.name === 'chrome') {
+          launchOptions.args.push('--enable-features=SharedArrayBuffer');
+        }
+
+        return launchOptions;
+      });
+    },
   },
 });
diff --git a/binding/web/lib/.gitkeep b/binding/web/lib/.gitkeep
diff --git a/binding/web/module.d.ts b/binding/web/module.d.ts
@@ -3,6 +3,11 @@ declare module "*.wasm" {
   export default content;
 }
 
+declare module "*.txt" {
+  const content: string;
+  export default content;
+}
+
 declare module 'web-worker:*' {
   const WorkerFactory: new () => Worker;
   export default WorkerFactory;

diff --git a/binding/web/package.json b/binding/web/package.json
@@ -3,7 +3,7 @@
   "description": "picoLLM Inference Engine is a highly accurate and cross-platform SDK optimized for running compressed large language models.",
   "author": "Picovoice Inc",
   "license": "Apache-2.0",
-  "version": "1.2.3",
+  "version": "1.3.0",
   "keywords": [
     "web",
     "ai",
@@ -22,6 +22,7 @@
     "lint": "eslint . --ext .js,.ts",
     "prepack": "npm-run-all build",
     "start": "cross-env TARGET='debug' rollup --config --watch",
+    "serve": "node server.js",
     "watch": "rollup --config --watch",
     "format": "prettier --write \"**/*.{js,ts,json}\"",
     "copywasm": "node scripts/copy_wasm.js",
@@ -41,20 +42,18 @@
     "@rollup/plugin-node-resolve": "^15.0.1",
     "@rollup/plugin-terser": "^0.4.0",
     "@rollup/pluginutils": "^5.0.2",
+    "@types/emscripten": "^1.39.13",
     "@types/node": "^18.11.9",
-    "@types/pv-xpu-web-worker": "../../lib/wasm/dist/types/xpu_web_worker",
-    "@types/pv-xpu-webgpu": "../../lib/wasm/dist/types/xpu_webgpu",
     "@typescript-eslint/eslint-plugin": "^5.51.0",
     "@typescript-eslint/parser": "^5.51.0",
     "async-mutex": "^0.4.0",
     "cross-env": "^7.0.3",
     "cypress": "^12.8.1",
     "eslint": "^8.22.0",
     "eslint-plugin-cypress": "^2.12.1",
+    "mime-types": "^2.1.35",
     "npm-run-all": "^4.1.5",
     "prettier": "^2.8.3",
-    "pv-xpu-web-worker": "../../lib/wasm/dist/xpu_web_worker/esm",
-    "pv-xpu-webgpu": "../../lib/wasm/dist/xpu_webgpu/esm",
     "rollup": "^2.79.1",
     "rollup-plugin-typescript2": "^0.34.1",
     "rollup-plugin-web-worker-loader": "^1.6.1",

diff --git a/binding/web/rollup.config.js b/binding/web/rollup.config.js
@@ -69,7 +69,7 @@ export default {
       exclude: '**/node_modules/**',
     }),
     base64({
-      include: ['./lib/**/*.wasm'],
+      include: ['./**/*.wasm', './**/*.txt'],
     })
   ],
 };
diff --git a/binding/web/scripts/copy_wasm.js b/binding/web/scripts/copy_wasm.js
@@ -1,11 +1,11 @@
 const fs = require("fs");
-const { join } = require("path");
+const { join, extname } = require("path");
 
 const wasmFiles = [
   "pv_picollm.wasm",
+  "pv_picollm.js",
   "pv_picollm_simd.wasm",
-  "pv_picollm_web_worker_helper.wasm",
-  "pv_picollm_web_worker_helper_simd.wasm",
+  "pv_picollm_simd.js",
 ]
 
 console.log("Copying the WASM model...");
@@ -19,12 +19,16 @@ const sourceDirectory = join(
   "wasm"
 );
 
-const outputDirectory = join(__dirname, "..", "lib");
+const outputDirectory = join(__dirname, "..", "src", "lib");
 
 try {
   fs.mkdirSync(outputDirectory, { recursive: true });
   wasmFiles.forEach(file => {
     fs.copyFileSync(join(sourceDirectory, file), join(outputDirectory, file))
+    const ext = extname(file);
+    if (ext === ".js") {
+      fs.copyFileSync(join(sourceDirectory, file), join(outputDirectory, file.replace(ext, ".txt")));
+    }
   })
 } catch (error) {
   console.error(error);

diff --git a/binding/web/src/index.ts b/binding/web/src/index.ts
@@ -43,12 +43,15 @@ import {
   PicoLLMWorkerResponse,
 } from './types';
 
-import picoLLMWasmSimd from '../lib/pv_picollm_simd.wasm';
+import picoLLMWasmSimd from './lib/pv_picollm_simd.wasm';
+import picoLLMWasmLib from './lib/pv_picollm_simd.txt';
 
 import * as PicoLLMErrors from './picollm_errors';
 
 PicoLLM.setWasmSimd(picoLLMWasmSimd);
 PicoLLMWorker.setWasmSimd(picoLLMWasmSimd);
+PicoLLM.setWasmLib(picoLLMWasmLib);
+PicoLLMWorker.setWasmLib(picoLLMWasmLib);
 
 export {
   Dialog,