fix Ruff / Type check

Author: terapyon

scripts/copy_metadata_to_clip_db.py

@@ -26,7 +26,8 @@ def main() -> None:
     """)
 
     row = target.execute("SELECT COUNT(*) FROM images").fetchone()
-    print(f"CLIP DB now has {row[0]} images at {CLIP_DB_PATH}")
+    count = row[0] if row else 0
+    print(f"CLIP DB now has {count} images at {CLIP_DB_PATH}")
 
     target.execute("DETACH source")
     target.close()

src/pyconjp_image_search/embedding/clip.py

@@ -31,7 +31,7 @@ def _normalize(embeddings: np.ndarray) -> np.ndarray:
     def embed_images(self, image_paths: list[Path]) -> np.ndarray:
         """Embed a batch of images. Returns L2-normalized projected vectors (768-dim)."""
         images = [Image.open(p).convert("RGB") for p in image_paths]
-        inputs = self.processor(images=images, return_tensors="pt")
+        inputs = self.processor(images=images, return_tensors="pt")  # type: ignore[operator]
         inputs = {k: v.to(self.device) for k, v in inputs.items()}
         with torch.no_grad():
             # Explicitly run vision_model + visual_projection to get 768-dim
@@ -45,12 +45,15 @@ def embed_images(self, image_paths: list[Path]) -> np.ndarray:
 
     def embed_text(self, text: str) -> np.ndarray:
         """Embed a single text query. Returns L2-normalized projected vector (1, 768)."""
-        inputs = self.processor(text=[text], return_tensors="pt", padding=True, truncation=True)
+        proc_kwargs = dict(text=[text], return_tensors="pt", padding=True, truncation=True)
+        inputs = self.processor(**proc_kwargs)  # type: ignore[operator]
         inputs = {k: v.to(self.device) for k, v in inputs.items()}
         with torch.no_grad():
             # Explicitly run text_model + text_projection to guarantee
             # 768-dim projected features matching JS CLIPTextModelWithProjection.
-            text_out = self.model.text_model(input_ids=inputs["input_ids"], attention_mask=inputs["attention_mask"])
+            text_out = self.model.text_model(
+                input_ids=inputs["input_ids"], attention_mask=inputs["attention_mask"]
+            )
             projected = self.model.text_projection(text_out.pooler_output)
         embedding = projected.cpu().numpy()
         return self._normalize(embedding).astype(np.float32)

src/pyconjp_image_search/search/app.py

@@ -572,7 +572,9 @@ def _do_search_cropped(
                 text_embedding_state = gr.State(None)
                 text_selected_index_state = gr.State(None)
 
-                def do_text_search(query: str, selected_events: list[str], model_choice: str) -> tuple:
+                def do_text_search(
+                    query: str, selected_events: list[str], model_choice: str
+                ) -> tuple:
                     if not query.strip():
                         return (
                             [],
@@ -1005,15 +1007,25 @@ def _on_tab_switch():
 
         text_find_similar_btn.click(
             fn=_do_find_similar,
-            inputs=[text_selected_index_state, text_metadata_state, text_event_filter, model_selector],
+            inputs=[
+                text_selected_index_state,
+                text_metadata_state,
+                text_event_filter,
+                model_selector,
+            ],
             outputs=_find_similar_outputs,
         ).then(
             fn=_on_close_preview,
             outputs=_text_close_outputs,
         )
         img_find_similar_btn.click(
             fn=_do_find_similar,
-            inputs=[image_selected_index_state, image_metadata_state, image_event_filter, model_selector],
+            inputs=[
+                image_selected_index_state,
+                image_metadata_state,
+                image_event_filter,
+                model_selector,
+            ],
             outputs=_find_similar_outputs,
         ).then(
             fn=_on_close_preview,