Verify packaged workers from Unicode extraction directories

tools/embeddinggemma2/package_smoke.py

@@ -31,7 +31,7 @@ def main():
         assert hashlib.file_digest(stream, "sha256").hexdigest() == manifest["files"][0]["sha256"]
     parent = ROOT / ".searchi/packages"
     parent.mkdir(parents=True, exist_ok=True)
-    with tempfile.TemporaryDirectory(prefix="smoke-", dir=parent) as folder:
+    with tempfile.TemporaryDirectory(prefix="smoke-Gemäß_日本_😀-", dir=parent) as folder:
         target = Path(folder).resolve()
         assert target.is_relative_to(parent.resolve())
         with zipfile.ZipFile(args.archive) as archive:
@@ -82,6 +82,7 @@ def main():
             "vulkan_module_loaded": hello.get("vulkan_module_loaded"),
             "baseline_only": args.baseline,
             "worker_role": hello["role"], "context_tokens": hello["context_tokens"],
+            "unicode_extraction_directory": True,
             "archive_sha256": archive_checksum,
             "worker_requested_device": "auto", "worker_actual_device": hello["device"],
             "no_gpu_condition": "GGML_DISABLE_VULKAN=1 only in child; SEARCHI_FORCE_CPU absent",