fix(ml): load SigLIP image-only processor to avoid SentencePiece dep — Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

backend/app/services/ml/embedder.py

@@ -34,10 +34,21 @@ class Embedder:
         if self._model is not None:
             return
         import torch
-        from transformers import AutoModel, AutoProcessor
+        from transformers import AutoModel, SiglipImageProcessor
 
         self._torch = torch
-        self._processor = AutoProcessor.from_pretrained(str(self._model_dir))
+        # FC's embedder only does IMAGE inference — never text. AutoProcessor
+        # loads the full processor including SiglipTokenizer, which requires
+        # the sentencepiece library at import time even if we never call it.
+        # SiglipImageProcessor loads ONLY preprocessor_config.json (image
+        # side) and skips the tokenizer config entirely. Operator hit the
+        # ImportError 2026-05-25 once the ml-worker started actually running
+        # tag_and_embed; switching to the image-only loader avoids the
+        # tokenizer dep without adding ~30 MB of unused C++ build to the
+        # lean ml-worker image.
+        self._processor = SiglipImageProcessor.from_pretrained(
+            str(self._model_dir)
+        )
         self._model = AutoModel.from_pretrained(str(self._model_dir))
         self._model.eval()