feat(cli): temporarily disable ONNX Runtime integration

To unblock release, temporarily disable ONNX/MiniLM integration:
- Comment out extract-onnx-runtime and extract-minilm-model in build steps
- Remove from build:js and build:watch scripts in package.json
- Disable MiniLM inference in handle-ask.mts (falls back to pattern matching)

The ask command continues to work using pattern matching only.
ONNX integration will be re-enabled once build issues are resolved.

Author: jdalton

packages/cli/package.json

@@ -35,8 +35,8 @@
   "scripts": {
     "build": "node --import=./scripts/load.mjs scripts/build.mjs",
     "build:sea:internal:bootstrap": "node .config/esbuild.sea-bootstrap.build.mjs",
-    "build:js": "node scripts/extract-yoga-wasm.mjs && node scripts/extract-onnx-runtime.mjs && node scripts/extract-minilm-model.mjs && node .config/esbuild.cli.build.mjs",
-    "build:watch": "node scripts/extract-yoga-wasm.mjs && node scripts/extract-onnx-runtime.mjs && node scripts/extract-minilm-model.mjs && node .config/esbuild.cli.build.mjs --watch",
+    "build:js": "node scripts/extract-yoga-wasm.mjs && node .config/esbuild.cli.build.mjs",
+    "build:watch": "node scripts/extract-yoga-wasm.mjs && node .config/esbuild.cli.build.mjs --watch",
     "dev": "pnpm run build:watch",
     "publish:sea": "node --import=./scripts/load.mjs scripts/publish-sea.mjs",
     "check": "node --import=./scripts/load.mjs scripts/check.mjs",

packages/cli/scripts/build.mjs

@@ -84,16 +84,16 @@ async function main() {
         command: 'pnpm',
         args: ['run', 'clean:dist'],
       },
-      {
-        name: 'Extract MiniLM Model',
-        command: 'node',
-        args: ['scripts/extract-minilm-model.mjs'],
-      },
-      {
-        name: 'Extract ONNX Runtime',
-        command: 'node',
-        args: ['scripts/extract-onnx-runtime.mjs'],
-      },
+      // {
+      //   name: 'Extract MiniLM Model',
+      //   command: 'node',
+      //   args: ['scripts/extract-minilm-model.mjs'],
+      // },
+      // {
+      //   name: 'Extract ONNX Runtime',
+      //   command: 'node',
+      //   args: ['scripts/extract-onnx-runtime.mjs'],
+      // },
       {
         name: 'Extract Yoga WASM',
         command: 'node',

packages/cli/src/commands/ask/handle-ask.mts

@@ -271,13 +271,17 @@ async function getEmbeddingPipeline() {
   }
 
   try {
+    // TEMPORARILY DISABLED: ONNX Runtime build issues.
     // Load our custom MiniLM inference engine.
     // This uses direct ONNX Runtime + embedded WASM (no transformers.js).
     // Note: Model is optional - pattern matching works fine without it.
-    const { MiniLMInference } = await import('../../utils/minilm-inference.mts')
-    embeddingPipeline = await MiniLMInference.create()
+    // const { MiniLMInference } = await import('../../utils/minilm-inference.mts')
+    // embeddingPipeline = await MiniLMInference.create()
+    // return embeddingPipeline
 
-    return embeddingPipeline
+    // Temporarily fall back to pattern matching only.
+    embeddingPipelineFailure = true
+    return null
   } catch (_e) {
     // Model not available - silently fall back to pattern matching.
     embeddingPipelineFailure = true