fix: media thumbnails, product dimensions, inline 3D viewer, GLB export

Bug A: Media Library thumbnails were gray because <img src> cannot send
JWT auth headers. Added useAuthBlob() hook (fetch + createObjectURL) in
MediaBrowser.tsx. Also fixed publish_asset Celery task to populate
product_id + cad_file_id on MediaAsset for thumbnail fallback resolution.

Bug B: Product dimensions now shown in Product Details card with Ruler
icon and "from CAD" label when cad_mesh_attributes.dimensions_mm exists.

Bug C: Replaced 128×128 CAD thumbnail with InlineCadViewer component.
Queries gltf_geometry MediaAssets, fetches GLB via auth fetch → blob URL
→ Three.js Canvas with OrbitControls. Falls back to thumbnail + "Load 3D
Model" button. Polling when GLB generation is in progress.

Bug D: trimesh was in [cad] optional extra but Dockerfile only installed
[dev]. Changed to pip install -e ".[dev,cad]" — trimesh now available in
backend container, GLB + Colors export works.

Also added bbox extraction (STL-first numpy parsing) in render_step_thumbnail
and admin "Re-extract CAD Metadata" bulk endpoint.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/domains/rendering/tasks.py

@@ -15,6 +15,36 @@ from app.core.task_logs import log_task_event
 logger = logging.getLogger(__name__)
 
 
+def _update_workflow_run_status(order_line_id: str, status: str, error: str | None = None) -> None:
+    """Update the most recent WorkflowRun for an order_line after task completion."""
+    try:
+        import asyncio
+        from datetime import datetime as _dt
+
+        async def _run():
+            from app.database import AsyncSessionLocal
+            from app.domains.rendering.models import WorkflowRun
+            from sqlalchemy import select as _sel
+            async with AsyncSessionLocal() as db:
+                res = await db.execute(
+                    _sel(WorkflowRun)
+                    .where(WorkflowRun.order_line_id == order_line_id)
+                    .order_by(WorkflowRun.created_at.desc())
+                    .limit(1)
+                )
+                run = res.scalar_one_or_none()
+                if run and run.status == "pending":
+                    run.status = status
+                    run.completed_at = _dt.utcnow()
+                    if error:
+                        run.error_message = error[:2000]
+                    await db.commit()
+
+        asyncio.get_event_loop().run_until_complete(_run())
+    except Exception as _exc:
+        logger.warning("Failed to update WorkflowRun status for line %s: %s", order_line_id, _exc)
+
+
 @celery_app.task(
     bind=True,
     name="app.domains.rendering.tasks.render_still_task",
@@ -291,6 +321,7 @@ def publish_asset(
         from app.database import AsyncSessionLocal
         from app.domains.media.models import MediaAsset, MediaAssetType
         from app.domains.orders.models import OrderLine
+        from app.domains.products.models import Product
         from sqlalchemy import select
 
         async with AsyncSessionLocal() as db:
@@ -298,9 +329,20 @@ def publish_asset(
             line = res.scalar_one_or_none()
             if not line:
                 return None
+
+            # Resolve cad_file_id from the linked product
+            cad_file_id = None
+            if line.product_id:
+                prod_res = await db.execute(select(Product).where(Product.id == line.product_id))
+                product = prod_res.scalar_one_or_none()
+                if product:
+                    cad_file_id = product.cad_file_id
+
             asset = MediaAsset(
                 tenant_id=getattr(line, "tenant_id", None),
                 order_line_id=line.id,
+                product_id=line.product_id,
+                cad_file_id=cad_file_id,
                 asset_type=MediaAssetType(asset_type),
                 storage_key=storage_key,
                 render_config=render_config,
@@ -396,6 +438,7 @@ def render_order_line_still_task(self, order_line_id: str, **params) -> dict:
             })
         except Exception:
             pass
+        _update_workflow_run_status(order_line_id, "completed")
         return result
     except Exception as exc:
         log_task_event(self.request.id, f"Failed: {exc}", "error")
@@ -409,6 +452,7 @@ def render_order_line_still_task(self, order_line_id: str, **params) -> dict:
             })
         except Exception:
             pass
+        _update_workflow_run_status(order_line_id, "failed", str(exc))
         raise self.retry(exc=exc, countdown=30)
 
 
@@ -448,6 +492,29 @@ def export_gltf_for_order_line_task(self, order_line_id: str) -> dict:
 
     asset_type = "gltf_geometry"
 
+    # Load sharp edge hints from mesh_attributes for UV seam marking
+    sharp_edges_json = "[]"
+    if cad_file_id:
+        try:
+            import asyncio as _asyncio
+
+            async def _load_mesh_attrs() -> list:
+                from app.database import AsyncSessionLocal
+                from app.models.cad_file import CadFile as _CF
+                from sqlalchemy import select as _sel
+                async with AsyncSessionLocal() as _db:
+                    _res = await _db.execute(_sel(_CF).where(_CF.id == cad_file_id))
+                    _cad = _res.scalar_one_or_none()
+                    if _cad and _cad.mesh_attributes:
+                        return _cad.mesh_attributes.get("sharp_edge_midpoints") or []
+                    return []
+
+            _midpoints = _asyncio.get_event_loop().run_until_complete(_load_mesh_attrs())
+            if _midpoints:
+                sharp_edges_json = json.dumps(_midpoints)
+        except Exception as _exc:
+            logger.warning("Could not load sharp_edge_midpoints for %s: %s", cad_file_id, _exc)
+
     if is_blender_available() and export_script.exists():
         blender_bin = find_blender()
         cmd = [
@@ -458,6 +525,7 @@ def export_gltf_for_order_line_task(self, order_line_id: str) -> dict:
             "--output_path", str(output_path),
             "--asset_library_blend", "",
             "--material_map", json.dumps({}),
+            "--sharp_edges_json", sharp_edges_json,
         ]
         try:
             result = subprocess.run(cmd, capture_output=True, text=True, timeout=300)