Add detailed logging for new/updated/deleted pages

Author: mishig25

src/doc_builder/build_embeddings.py

@@ -913,6 +913,23 @@ def add_gradio_docs(
         print(f"New chunks to embed: {len(chunks_to_process)}")
         print(f"Stale entries to remove: {len(stale_ids)}")
 
+        # Log which pages are being updated
+        if chunks_to_process:
+            new_pages = {c.source_page_url for c in chunks_to_process}
+            print(f"\nNew/updated pages ({len(new_pages)}):")
+            for page in sorted(new_pages)[:20]:  # Show first 20
+                print(f"  + {page}")
+            if len(new_pages) > 20:
+                print(f"  ... and {len(new_pages) - 20} more")
+
+        # Log which IDs are being removed
+        if stale_ids:
+            print(f"\nStale IDs to remove ({len(stale_ids)}):")
+            for stale_id in sorted(stale_ids)[:20]:  # Show first 20
+                print(f"  - {stale_id}")
+            if len(stale_ids) > 20:
+                print(f"  ... and {len(stale_ids) - 20} more")
+
         if len(chunks_to_process) == 0 and len(stale_ids) == 0:
             print("No new Gradio documents to process. All documents are up to date.")
             return

src/doc_builder/commands/embeddings.py

@@ -102,6 +102,23 @@ def process_hf_docs_command(args):
             print(f"New chunks to embed: {len(chunks_to_process)}")
             print(f"Stale entries to remove: {len(stale_ids)}")
 
+            # Log which pages are being updated
+            if chunks_to_process:
+                new_pages = {f"{c.package_name}/{c.page}" for c in chunks_to_process}
+                print(f"\nNew/updated pages ({len(new_pages)}):")
+                for page in sorted(new_pages)[:20]:  # Show first 20
+                    print(f"  + {page}")
+                if len(new_pages) > 20:
+                    print(f"  ... and {len(new_pages) - 20} more")
+
+            # Log which pages are being removed
+            if stale_ids:
+                print(f"\nStale IDs to remove ({len(stale_ids)}):")
+                for stale_id in sorted(stale_ids)[:20]:  # Show first 20
+                    print(f"  - {stale_id}")
+                if len(stale_ids) > 20:
+                    print(f"  ... and {len(stale_ids) - 20} more")
+
             if len(chunks_to_process) == 0 and len(stale_ids) == 0:
                 print("\nNo new documents to process. All documents are up to date.")
                 print("\n" + "=" * 80)

testdir/hub-docs

@@ -0,0 +1 @@
+Subproject commit c81752a5cfe74c1d0d09b9d11aeddbb582d80c3b