Switch the home-search tutorial to place_blocks now that the endpoint is fixed.

Author: Nathaniel Henry

.claude/CLAUDE.md

@@ -44,9 +44,10 @@ OIDC trusted publishing) is the eventual target. Public repo:
   `_blocks_gdf` drops them (`geometry.notna()`) so plotting and spatial joins work. The
   block join renames the key to `geoid` before merging (the R SDK had a bug here; do
   not regress).
-- **`place_blocks()` (`/v1/places/{geoid}/blocks`) is BROKEN server-side:** it times
-  out (15s Lambda) for every real place. Tutorials use `blocks_query(center, radius_m)`
-  instead. Leave the wrapper, but do not build examples on it until the API is fixed.
+- **`place_blocks()` works since 2026-07-22** (server-side query fix in wtm.api). The
+  home-search tutorial now uses it for whole-city pulls; `blocks_query` remains the
+  tool for a radius or polygon. A rarely-queried big place can take a few seconds on
+  the first call while the database cache warms; it is not broken again.
 
 ## Local live doc build (this EFS host)
 

docs/tutorials/home_search.md

@@ -63,13 +63,13 @@ stops.plot(ax = ax, color = "#f36e21", marker = "^")
 
 ## Find the blocks that qualify
 
-Pull the per-block walk times for the blocks around the city centre. The result is a
-GeoDataFrame with one row per (block, category); the boundaries come from `pygris`,
-downloaded once.
+Somerville is a census place, so one call by place GEOID pulls the per-block walk
+times for every block in the city. The result is a GeoDataFrame with one row per
+(block, category); the boundaries come from `pygris`, downloaded once. To search an
+arbitrary area instead, use `blocks_query` with a centre and radius or a polygon.
 
 ```{code-cell} python
-blocks = close.blocks_query(center = {"lon": city["lon"], "lat": city["lat"]},
-                            radius_m = 3000, mode = "walk",
+blocks = close.place_blocks(city["geoid"], mode = "walk",
                             type = [grocery, restaurant, transit])
 ```