Verify packaged workers from Unicode extraction directories

docs/WINDOWS-UNICODE.md

@@ -57,6 +57,22 @@ $env:SEARCHI_GEMMA_WORKER = 'PATH/TO/CURRENT/searchi-gemma-worker.exe'
 cargo test -p searchi-semantic --test unicode_windows_cpu --locked -- --ignored --nocapture
 ```
 
-Native package restaging and installed Unicode-profile paths require their own
-checks. This focused model-path regression does not certify every filesystem,
+## Updated development archive
+
+The normal worker was restaged and the embedded desktop Release rebuilt in
+2m 02s. Executable: 23,924,224 bytes, SHA-256
+`bce713e7d081095e650756d1b0ed497cb21e451ea2e47bead6676a22327729b9`.
+The unsigned Windows archive is 145,409,579 bytes, SHA-256
+`512be4960da197a45710d49bdbc27b84c18e0270364e41bbcc546b8791826842`.
+
+Both actual package checks extracted under a Unicode directory and verified all
+39 recorded files. Default `auto` mode used CPU with Vulkan registration disabled
+only in the child: optimized indexing context 2048 and baseline-only query
+context 512 both produced normalized 768-dimensional real Google vectors.
+The packaged FFprobe started successfully in that directory. Local reports are
+`.searchi/packages/smoke-unicode-{indexing,query-baseline}.json`.
+
+These are worker/dependency checks outside the build tree, not complete installed
+desktop interaction, Unicode user-profile, clean-machine, signing or older
+physical-CPU certification. This regression does not certify every filesystem,
 codec, launcher or installer path, or any GPU hardware.