diff --git a/.github/dependabot.yml b/.github/dependabot.yml
index 97eab795..f9cf9d96 100644
--- a/.github/dependabot.yml
+++ b/.github/dependabot.yml
@@ -6,7 +6,7 @@
 version: 2
 updates:
   - package-ecosystem: "pip" # See documentation for possible values
-    directory: "/application" # Location of package manifests
+    directory: "/docsgpt" # Location of package manifests
     schedule:
       interval: "daily"
   - package-ecosystem: "npm" # See documentation for possible values
diff --git a/.github/workflows/docker-image-verify.yml b/.github/workflows/docker-image-verify.yml
index a8ea0d1a..911581fe 100644
--- a/.github/workflows/docker-image-verify.yml
+++ b/.github/workflows/docker-image-verify.yml
@@ -11,7 +11,7 @@ on:
       - 'docsgpt/Dockerfile'
       - '.dockerignore'
       - 'application/**'
-      - 'application/requirements*.txt'
+      - 'docsgpt/requirements*.txt'
       - 'docsgpt/scripts/prefetch_models.py'
       - 'docsgpt/scripts/verify_offline.py'
       - 'docsgpt/vectorstore/model_registry.py'
diff --git a/.gitignore b/.gitignore
index 12e001a0..d2e4019f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -180,6 +180,7 @@ frontend/*.sln
 frontend/*.sw?
 
 docsgpt/vectors/
+application/vectors/
 
 **/inputs
 
diff --git a/deployment/docker-compose-azure.yaml b/deployment/docker-compose-azure.yaml
index 97b19295..2d5a7866 100644
--- a/deployment/docker-compose-azure.yaml
+++ b/deployment/docker-compose-azure.yaml
@@ -44,12 +44,14 @@ services:
     ports:
       - "7091:7091"
     volumes:
-      # Host data stays under application/ (the old package directory) for this
-      # release so an upgraded checkout keeps its indexes; it moves with the
-      # packaging work, together with an upgrade note.
-      - ../application/indexes:/app/docsgpt/indexes
-      - ../application/inputs:/app/docsgpt/inputs
-      - ../application/vectors:/app/docsgpt/vectors
+      # Host data lives under application/ (the old package directory) for this
+      # release, like the other compose files; it moves with the packaging work,
+      # together with an upgrade note. The container side is /app: the process
+      # runs there and resolves indexes/inputs/vectors relative to it (earlier
+      # versions of this file mounted /app/application/..., which nothing wrote to).
+      - ../application/indexes:/app/indexes
+      - ../application/inputs:/app/inputs
+      - ../application/vectors:/app/vectors
     depends_on:
       redis:
         condition: service_started
diff --git a/docs/content/upgrading.mdx b/docs/content/upgrading.mdx
index 4e634525..abf71d2c 100644
--- a/docs/content/upgrading.mdx
+++ b/docs/content/upgrading.mdx
@@ -106,7 +106,9 @@ alias, so nothing breaks on upgrade, but update these before the alias goes:
 - Celery task names changed with the package (`docsgpt.api.user.tasks.ingest`
   and so on). A worker on this release also accepts the old names, so tasks
   queued before the upgrade still run, and beat rewrites the periodic
-  schedule in Redis on start-up. Nothing to do.
+  schedule in Redis on start-up. The daily, weekly and monthly source-sync
+  timers restart from the upgrade, so the first sync after it can land later
+  than it would have (a monthly sync by up to a month). Nothing to do.
 - Data directories do not move: the compose files keep your indexes, inputs
   and vectors under `application/` in the checkout, where they already are.
 
diff --git a/docsgpt/celery_init.py b/docsgpt/celery_init.py
index 02da0711..4272eac1 100644
--- a/docsgpt/celery_init.py
+++ b/docsgpt/celery_init.py
@@ -121,6 +121,7 @@ def _trim_native_heap() -> None:
 # measured at ~86 ms for the collect against ~8 ms for the embed itself on a
 # worker holding the ONNX model, i.e. a 9x slowdown of the whole round trip.
 _NO_RECLAIM_TASKS = frozenset({"docsgpt.vectorstore.embeddings_tasks.embed_texts"})
+LEGACY_TASK_PREFIX = "application."
 
 
 @task_postrun.connect
@@ -132,7 +133,10 @@ def _reclaim_memory_after_task(task=None, **kwargs):
     generational collect after a task that allocated a few kilobytes just
     charges the next task for walking the whole heap.
     """
-    if getattr(task, "name", None) in _NO_RECLAIM_TASKS:
+    name = getattr(task, "name", None)
+    if isinstance(name, str) and name.startswith(LEGACY_TASK_PREFIX):
+        name = "docsgpt." + name[len(LEGACY_TASK_PREFIX):]
+    if name in _NO_RECLAIM_TASKS:
         return
     gc.collect()
     torch = sys.modules.get("torch")
@@ -169,7 +173,6 @@ celery = make_celery()
 celery.config_from_object("docsgpt.celeryconfig")
 
 #: Task-name prefix the package carried before the rename to ``docsgpt``.
-LEGACY_TASK_PREFIX = "application."
 
 
 def register_legacy_task_names(app: Celery) -> int:
diff --git a/docsgpt/core/optional_deps.py b/docsgpt/core/optional_deps.py
index 4321df78..1f1ab595 100644
--- a/docsgpt/core/optional_deps.py
+++ b/docsgpt/core/optional_deps.py
@@ -1,7 +1,7 @@
 """Optional dependency extras and the one place their install hints come from.
 
 Heavy or niche packages are not installed by default. Each extra maps to a
-pyproject extra and to an exported ``application/requirements-<extra>.txt``,
+pyproject extra and to an exported ``docsgpt/requirements-<extra>.txt``,
 so a missing module can always be explained with the exact command to run.
 """
 
@@ -28,7 +28,7 @@ _MODULE_TO_EXTRA: Dict[str, str] = {
 def install_hint(extra: str) -> str:
     """Install command for ``extra``, for error messages and logs."""
     return (
-        f"pip install -r application/requirements-{extra}.txt "
+        f"pip install -r docsgpt/requirements-{extra}.txt "
         f"(or: uv sync --extra {extra}; Docker: --build-arg EXTRAS={extra})"
     )
 
diff --git a/scripts/qa/durability_e2e.py b/scripts/qa/durability_e2e.py
index 1d645de6..c5513a48 100755
--- a/scripts/qa/durability_e2e.py
+++ b/scripts/qa/durability_e2e.py
@@ -83,12 +83,7 @@ UNIQUE = f"qa_e2e_{int(time.time())}_{uuid.uuid4().hex[:6]}"
 
 
 # Quiet noisy library loggers so the script's own output stays readable.
-for name in (
-    "application", "docsgpt.api", "docsgpt.storage",
-    "docsgpt.usage", "docsgpt.parser",
-    "docsgpt.api.user.reconciliation",
-):
-    logging.getLogger(name).setLevel(logging.ERROR)
+logging.getLogger("docsgpt").setLevel(logging.ERROR)
 
 
 # ---------------------------------------------------------------------------
diff --git a/tests/core/test_optional_deps.py b/tests/core/test_optional_deps.py
index 34f8cc6e..b5bdc01d 100644
--- a/tests/core/test_optional_deps.py
+++ b/tests/core/test_optional_deps.py
@@ -23,7 +23,7 @@ class TestExtras:
 
     def test_install_hint_names_every_install_route(self):
         hint = optional_deps.install_hint("milvus")
-        assert "requirements-milvus.txt" in hint
+        assert "pip install -r docsgpt/requirements-milvus.txt" in hint
         assert "--extra milvus" in hint
         assert "EXTRAS=milvus" in hint
 
diff --git a/tests/test_celery.py b/tests/test_celery.py
index 025e6857..c5b692df 100644
--- a/tests/test_celery.py
+++ b/tests/test_celery.py
@@ -250,6 +250,10 @@ class TestReclaimIsSkippedForEmbeds:
     def test_the_embed_task_is_skipped(self):
         assert not self._collects("docsgpt.vectorstore.embeddings_tasks.embed_texts")
 
+    def test_the_legacy_embed_name_is_skipped_too(self):
+        """Messages from the previous release carry the application.* name."""
+        assert not self._collects("application.vectorstore.embeddings_tasks.embed_texts")
+
     def test_parsing_still_reclaims(self):
         assert self._collects("docsgpt.api.user.tasks.parse_document")