docs(profiler): clarify when planner-profile-data ConfigMap is emitted [DYN-2751] (cherry-pick of #8486)

components/src/dynamo/profiler/profile_sla.py

@@ -385,6 +385,16 @@ async def run_profile(
                 phase=ops.current_phase,
             )
             if not is_disagg_config:
+                # TODO: agg + throughput-scaling / agg + mocker has no
+                # profiling-data fallback today. run_interpolation is
+                # shaped around prefill + decode picks, so for agg picks
+                # the NPZ sweep is skipped entirely and no
+                # planner-profile-data ConfigMap is produced — the
+                # planner and mocker have nothing to consume on
+                # aggregated deployments. Extend run_interpolation (and
+                # the downstream _load_profiling_data / mount logic) to
+                # handle an agg pick so both paths work for aggregated
+                # deployments too.
                 logger.info(
                     "Picked config is aggregated (chosen_exp=%r) — "
                     "skipping interpolation (requires disaggregated config).",

components/src/dynamo/profiler/utils/dgd_generation.py

@@ -70,7 +70,12 @@ def assemble_final_config(
     1. **Mocker** — swap the base to the mocker DGD template if enabled.
     2. **Planner** — inject the Planner service + planner-config ConfigMap.
     3. **Profile data** — attach interpolation-data ConfigMap when mocker
-       or planner-throughput is enabled.
+       or planner-throughput is enabled. The ConfigMap is only emitted
+       when the picked config is disaggregated AND the interpolation NPZ
+       files were produced on disk; agg picks skip interpolation entirely
+       (see the gate in ``profile_sla.py``), so no ConfigMap is attached
+       in that case and the planner / mocker have no profiling-data
+       fallback for aggregated deployments today.
     """
     if not dgd_config:
         return dgd_config