fix(vision): use ARM Python venv for Metal GPU detection

The hardware detection was using system python3 which was x86_64
(Rosetta) instead of native ARM. This prevented TensorFlow from
detecting the Metal GPU on Apple Silicon.

Now uses a Python 3.11 venv with tensorflow-macos and tensorflow-metal
for proper M4 Max GPU detection.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

apps/web/.gitignore

@@ -63,3 +63,4 @@ data/uploads/
 
 # ML training data
 training-data/
+scripts/train-column-classifier/.venv/

apps/web/src/app/api/vision-training/hardware/route.ts

@@ -45,9 +45,16 @@ export async function GET(): Promise<Response> {
       let stderr = ''
       let hasError = false
 
-      const childProcess = spawn('python3', [scriptPath], {
+      // Use the venv Python with tensorflow-metal for proper GPU detection
+      const venvPython = path.join(cwd, 'scripts/train-column-classifier/.venv/bin/python')
+
+      const childProcess = spawn(venvPython, [scriptPath], {
         cwd,
-        env: { ...process.env, PYTHONUNBUFFERED: '1' },
+        env: {
+          ...process.env,
+          PYTHONUNBUFFERED: '1',
+          PYTHONWARNINGS: 'ignore::FutureWarning', // Suppress keras warning
+        },
       })
 
       childProcess.stdout?.on('data', (data: Buffer) => {