feat(container): add standalone dynamo-planner image [DYN-2533] (#7696)

.github/filters.yaml

@@ -116,6 +116,7 @@ deploy:
   - 'tests/deploy/**'
 
 planner:
+  - 'container/templates/planner.Dockerfile'
   - 'components/src/dynamo/planner/**'
   - 'components/src/dynamo/global_planner/**'
   - 'tests/planner/**'

components/src/dynamo/planner/__main__.py

@@ -84,13 +84,13 @@ def _parse_config() -> PlannerConfig:
 
 
 @dynamo_worker()
-async def worker(runtime: DistributedRuntime):
-    config = _parse_config()
+async def worker(runtime: DistributedRuntime, config: PlannerConfig):
     await init_planner(runtime, config)
 
 
 def main():
-    asyncio.run(worker())  # type: ignore[call-arg]
+    config = _parse_config()
+    asyncio.run(worker(config))  # type: ignore[call-arg]
 
 
 if __name__ == "__main__":

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

@@ -102,7 +102,7 @@ class DgdPlannerServiceConfig(BaseModel):
     replicas: int = 1
     extraPodSpec: PodSpec = PodSpec(
         mainContainer=Container(
-            image="my-registry/dynamo-runtime:my-tag",  # placeholder
+            image="my-registry/dynamo-planner:my-tag",  # placeholder
             workingDir=f"{get_workspace_dir()}/components/src/dynamo/planner",
             command=["python3", "-m", "dynamo.planner"],
             args=[],
@@ -397,7 +397,7 @@ def set_argument_value(args: list[str], arg_name: str, value: str) -> list[str]:
 
 
 def update_image(config: dict, image: str) -> dict:
-    """Update container image for all DGD services (frontend, planner, workers).
+    """Update container image for non-planner DGD services.
 
     This is a shared utility function used by all backend config modifiers.
 
@@ -410,8 +410,9 @@ def update_image(config: dict, image: str) -> dict:
     """
     cfg = Config.model_validate(config)
 
-    # Update image for all services
     for service_name, service_config in cfg.spec.services.items():
+        if getattr(service_config, "componentType", None) == "planner":
+            continue
         if service_config.extraPodSpec and service_config.extraPodSpec.mainContainer:
             service_config.extraPodSpec.mainContainer.image = image
             logger.debug(f"Updated image for {service_name} to {image}")

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

@@ -28,6 +28,7 @@ from dynamo.planner.config.planner_config import PlannerConfig
 from dynamo.profiler.utils.config import DgdPlannerServiceConfig, set_argument_value
 from dynamo.profiler.utils.profile_common import (
     ProfilerOperationalConfig,
+    derive_planner_image,
     is_mocker_enabled,
     is_planner_enabled,
     needs_profile_data,
@@ -180,8 +181,10 @@ def add_planner_to_config(
     planner_cfg.profile_results_dir = PROFILE_DATA_MOUNT
 
     planner_service = DgdPlannerServiceConfig()
-    if planner_service.extraPodSpec.mainContainer:
-        planner_service.extraPodSpec.mainContainer.image = dgdr.image
+    if planner_service.extraPodSpec.mainContainer and dgdr.image:
+        planner_service.extraPodSpec.mainContainer.image = derive_planner_image(
+            dgdr.image
+        )
 
     planner_dict = planner_service.model_dump(exclude_unset=False)
 

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

@@ -45,6 +45,29 @@ BACKEND_IMAGE_NAMES: dict[str, str] = {
     "trtllm": "tensorrtllm-runtime",
 }
 
+PLANNER_IMAGE_NAME = "dynamo-planner"
+
+
+def _replace_image_name(image_ref: str, new_name: str) -> str:
+    """Replace the image name component in a Docker image reference.
+
+    Preserves the registry path prefix and tag suffix, only replacing the
+    last ``/``-delimited component (before any ``:tag``).
+    """
+    slash_idx = image_ref.rfind("/")
+    prefix = image_ref[: slash_idx + 1] if slash_idx >= 0 else ""
+    suffix = image_ref[slash_idx + 1 :]
+    name_and_tag, has_digest, digest = suffix.partition("@")
+    colon_idx = name_and_tag.rfind(":")
+    tag = name_and_tag[colon_idx:] if colon_idx >= 0 else ""
+    digest_suffix = f"@{digest}" if has_digest else ""
+    return f"{prefix}{new_name}{tag}{digest_suffix}"
+
+
+def derive_planner_image(profiler_image: str) -> str:
+    """Derive the planner service image from the profiler image reference."""
+    return _replace_image_name(profiler_image, PLANNER_IMAGE_NAME)
+
 
 def derive_backend_image(profiler_image: str, backend: str) -> str:
     """Derive the backend worker image from the profiler image.
@@ -82,14 +105,7 @@ def derive_backend_image(profiler_image: str, backend: str) -> str:
             f"Supported backends: {list(BACKEND_IMAGE_NAMES.keys())}"
         )
 
-    # Split off the last path component: "registry/path/name:tag" → "name:tag"
-    slash_idx = profiler_image.rfind("/")
-    prefix = profiler_image[: slash_idx + 1] if slash_idx >= 0 else ""
-    suffix = profiler_image[slash_idx + 1 :]
-    colon_idx = suffix.find(":")
-    tag = suffix[colon_idx:] if colon_idx >= 0 else ""
-
-    return f"{prefix}{backend_image_name}{tag}"
+    return _replace_image_name(profiler_image, backend_image_name)
 
 
 # ---------------------------------------------------------------------------

container/Dockerfile.template

@@ -13,6 +13,10 @@
         {% include "templates/dynamo_base.Dockerfile" %}
         {% include "templates/wheel_builder.Dockerfile" %}
         {% include "templates/frontend.Dockerfile" %}
+    {% elif target == "planner" %}
+        {% include "templates/dynamo_base.Dockerfile" %}
+        {% include "templates/wheel_builder.Dockerfile" %}
+        {% include "templates/planner.Dockerfile" %}
     {% elif target == "runtime" or target == "dev" or target == "local-dev" %}
         {% include "templates/dynamo_base.Dockerfile" %}
         {% include "templates/wheel_builder.Dockerfile" %}

container/context.yaml

@@ -18,6 +18,10 @@ dynamo:
     base_image_tag: 25.11-cuda13.0-devel-ubuntu24.04
   epp_image: us-central1-docker.pkg.dev/k8s-staging-images/gateway-api-inference-extension/epp:v0.5.1
   frontend_image: nvcr.io/nvidia/base/ubuntu:noble-20250619
+  planner_build_image: python
+  planner_build_image_tag: 3.12-slim
+  planner_runtime_image: nvcr.io/nvidia/distroless/python
+  planner_runtime_image_tag: 3.12-v4.0.3
   python_version: "3.12"
 
   nats_version: v2.10.28

container/deps/README.md

@@ -8,7 +8,7 @@ so each image only installs what it needs.
 | File | Purpose |
 |------|---------|
 | `requirements.common.txt` | Core deps shared by all containers |
-| `requirements.planner.txt` | Planner, profiler, global_planner, deploy utils |
+| `requirements.planner.txt` | Planner, profiler, and global_planner deps |
 | `requirements.frontend.txt` | Frontend deps |
 | `requirements.vllm.txt` | vLLM-specific deps |
 | `requirements.benchmark.txt` | Benchmark and profiling tools |

container/deps/requirements.planner.txt

 # SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-# Dependencies for planner, profiler, global_planner, and deploy utils.
+# Dependencies required by the planner, profiler, and global_planner services.
 
 aiconfigurator[webapp] @ git+https://github.com/ai-dynamo/aiconfigurator.git@5d419f99d60fdae0d3911cba06a9b571f3b2965c
 aiofiles<=25.1.0

container/render.py

@@ -140,6 +140,7 @@ def validate_args(args):
                 "dev",
                 "local-dev",
                 "frontend",
+                "planner",
                 "wheel_builder",
                 "base",
             ],

container/templates/args.Dockerfile

@@ -81,6 +81,13 @@ ARG EPP_IMAGE={{ context.dynamo.epp_image }}
 ARG FRONTEND_IMAGE={{ context.dynamo.frontend_image }}
 {% endif %}
 
+{% if target == "planner" %}
+ARG PLANNER_BUILD_IMAGE={{ context.dynamo.planner_build_image }}
+ARG PLANNER_BUILD_IMAGE_TAG={{ context.dynamo.planner_build_image_tag }}
+ARG PLANNER_RUNTIME_IMAGE={{ context.dynamo.planner_runtime_image }}
+ARG PLANNER_RUNTIME_IMAGE_TAG={{ context.dynamo.planner_runtime_image_tag }}
+{% endif %}
+
 {% if framework == "vllm" -%}
 # Make sure to update the dependency version in pyproject.toml when updating this
 ARG VLLM_REF={{ context[framework][device_key].vllm_ref }}

container/templates/planner.Dockerfile

+{#
+# SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#}
+# === BEGIN templates/planner.Dockerfile ===
+##############################################
+########## Planner / Profiler image ##########
+##############################################
+# Standalone planner/profiler image:
+# - install deps in a slim builder stage that has git/git-lfs available
+# - ship only the runtime artifacts in a distroless final stage
+
+FROM ${PLANNER_BUILD_IMAGE}:${PLANNER_BUILD_IMAGE_TAG} AS planner_builder
+
+ARG PYTHON_VERSION
+
+# Install only the packages needed to resolve and install the planner runtime
+# dependencies in the builder stage. git/git-lfs are only needed because
+# aiconfigurator is currently installed from a Git URL with LFS-backed assets.
+RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
+    apt-get update -y && \
+    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
+        ca-certificates \
+        git \
+        git-lfs \
+        libgomp1 && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Create dynamo user with group 0 for OpenShift compatibility.
+RUN useradd -m -s /bin/bash -g 0 dynamo \
+    && [ `id -u dynamo` -eq 1000 ] \
+    && mkdir -p /home/dynamo/.cache /opt/dynamo /workspace \
+    && chown -R dynamo:0 /home/dynamo /opt/dynamo /workspace \
+    && chmod -R g+w /home/dynamo/.cache /opt/dynamo /workspace
+
+ENV HOME=/home/dynamo \
+    VIRTUAL_ENV=/opt/dynamo/venv \
+    PATH="/opt/dynamo/venv/bin:/usr/local/bin/etcd:/usr/local/bin:/bin" \
+    PYTHONPATH="/workspace"
+
+WORKDIR /workspace
+
+COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
+COPY --from=dynamo_base /usr/local/bin/nats-server /usr/local/bin/nats-server
+COPY --from=dynamo_base /usr/local/bin/etcd /usr/local/bin/etcd
+COPY --chown=dynamo:0 --from=wheel_builder /opt/dynamo/dist/*.whl /opt/dynamo/wheelhouse/
+
+USER dynamo
+
+RUN --mount=type=cache,target=/home/dynamo/.cache/uv,uid=1000,gid=0,mode=0775 \
+    export UV_CACHE_DIR=/home/dynamo/.cache/uv && \
+    uv venv ${VIRTUAL_ENV} --python ${PYTHON_VERSION}
+
+# Install the local wheels and planner/profiler runtime dependencies before the
+# repo copies so changes in tests/configs don't invalidate the dependency layer.
+RUN --mount=type=bind,source=./container/deps/requirements.planner.txt,target=/tmp/requirements.planner.txt \
+    --mount=type=cache,target=/home/dynamo/.cache/uv,uid=1000,gid=0,mode=0775 \
+    export UV_CACHE_DIR=/home/dynamo/.cache/uv UV_GIT_LFS=1 UV_HTTP_TIMEOUT=300 UV_HTTP_RETRIES=5 && \
+    uv pip install \
+        --requirement /tmp/requirements.planner.txt \
+        /opt/dynamo/wheelhouse/ai_dynamo_runtime*.whl \
+        /opt/dynamo/wheelhouse/ai_dynamo*any.whl
+
+# Copy only the subset of the repository needed for planner/profiler service
+# startup and targeted planner/profiler unit tests.
+COPY --chmod=664 --chown=dynamo:0 pyproject.toml /workspace/pyproject.toml
+COPY --chmod=775 --chown=dynamo:0 tests /workspace/tests
+COPY --chmod=775 --chown=dynamo:0 components/src/dynamo/planner /workspace/components/src/dynamo/planner
+COPY --chmod=775 --chown=dynamo:0 components/src/dynamo/profiler /workspace/components/src/dynamo/profiler
+COPY --chmod=775 --chown=dynamo:0 components/src/dynamo/global_planner /workspace/components/src/dynamo/global_planner
+COPY --chmod=775 --chown=dynamo:0 deploy /workspace/deploy
+COPY --chmod=775 --chown=dynamo:0 examples /workspace/examples
+
+FROM ${PLANNER_RUNTIME_IMAGE}:${PLANNER_RUNTIME_IMAGE_TAG} AS planner
+
+COPY --from=planner_builder /etc/group /etc/passwd /etc/
+COPY --from=planner_builder /bin/dash /usr/bin/sh
+COPY --from=planner_builder /bin/uv /bin/uvx /usr/local/bin/
+COPY --chown=1000:0 --from=planner_builder /home/dynamo /home/dynamo
+COPY --chown=1000:0 --from=planner_builder /opt/dynamo/venv /opt/dynamo/venv
+COPY --from=planner_builder /usr/lib/*-linux-gnu/libgomp.so.1* /opt/dynamo/lib/
+COPY --from=planner_builder /usr/local/bin/etcd /usr/local/bin/etcd
+COPY --from=planner_builder /usr/local/bin/nats-server /usr/local/bin/nats-server
+COPY --chown=1000:0 --from=planner_builder /workspace /workspace
+
+ARG DYNAMO_COMMIT_SHA
+ENV DYNAMO_COMMIT_SHA=${DYNAMO_COMMIT_SHA} \
+    HOME=/home/dynamo \
+    VIRTUAL_ENV=/opt/dynamo/venv \
+    LD_LIBRARY_PATH="/opt/dynamo/lib" \
+    PATH="/opt/dynamo/venv/bin:/usr/local/bin/etcd:/usr/local/bin:/bin" \
+    PYTHONPATH="/workspace"
+
+WORKDIR /workspace
+USER dynamo
+
+CMD []

tests/profiler/test_planner_image_selection.py

+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+import sys
+from pathlib import Path
+
+import pytest
+
+project_root = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(project_root))
+sys.path.insert(0, str(project_root / "components" / "src"))
+
+pytestmark = [pytest.mark.pre_merge, pytest.mark.gpu_0, pytest.mark.unit]
+
+try:
+    from dynamo.profiler.utils.config import update_image
+    from dynamo.profiler.utils.dgd_generation import add_planner_to_config
+    from dynamo.profiler.utils.dgdr_v1beta1_types import (
+        DynamoGraphDeploymentRequestSpec,
+        HardwareSpec,
+        SLASpec,
+        WorkloadSpec,
+    )
+    from dynamo.profiler.utils.profile_common import (
+        derive_backend_image,
+        derive_planner_image,
+    )
+except ImportError as e:
+    pytest.skip(f"Skip (missing dependency): {e}", allow_module_level=True)
+
+
+def _make_dgdr(image: str) -> DynamoGraphDeploymentRequestSpec:
+    return DynamoGraphDeploymentRequestSpec(
+        model="Qwen/Qwen3-32B",
+        backend="trtllm",
+        image=image,
+        hardware=HardwareSpec(gpuSku="h200_sxm", totalGpus=8, numGpusPerNode=8),
+        workload=WorkloadSpec(isl=4000, osl=1000),
+        sla=SLASpec(ttft=2000.0, itl=50.0),
+    )
+
+
+def _base_dgd_config(image: str) -> dict:
+    return {
+        "metadata": {"name": "test-dgd"},
+        "spec": {
+            "services": {
+                "Frontend": {
+                    "replicas": 1,
+                    "extraPodSpec": {
+                        "mainContainer": {
+                            "image": image,
+                            "args": ["serve"],
+                        }
+                    },
+                }
+            }
+        },
+    }
+
+
+@pytest.mark.parametrize(
+    ("image", "expected"),
+    [
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3",
+            "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.3",
+        ),
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend@sha256:deadbeef",
+            "nvcr.io/nvidia/ai-dynamo/dynamo-planner@sha256:deadbeef",
+        ),
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3@sha256:deadbeef",
+            "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.3@sha256:deadbeef",
+        ),
+    ],
+)
+def test_derive_planner_image_preserves_registry_tag_and_digest(
+    image: str, expected: str
+):
+    assert derive_planner_image(image) == expected
+
+
+@pytest.mark.parametrize(
+    ("image", "backend", "expected"),
+    [
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3",
+            "vllm",
+            "nvcr.io/nvidia/ai-dynamo/vllm-runtime:1.2.3",
+        ),
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend@sha256:deadbeef",
+            "sglang",
+            "nvcr.io/nvidia/ai-dynamo/sglang-runtime@sha256:deadbeef",
+        ),
+        (
+            "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3@sha256:deadbeef",
+            "trtllm",
+            "nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.2.3@sha256:deadbeef",
+        ),
+    ],
+)
+def test_derive_backend_image_preserves_registry_tag_and_digest(
+    image: str, backend: str, expected: str
+):
+    assert derive_backend_image(image, backend) == expected
+
+
+def test_add_planner_to_config_uses_dynamo_planner_image():
+    image = "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3"
+    dgdr = _make_dgdr(image)
+    config = _base_dgd_config(image)
+
+    add_planner_to_config(dgdr, config)
+
+    planner_image = config["spec"]["services"]["Planner"]["extraPodSpec"][
+        "mainContainer"
+    ]["image"]
+    assert planner_image == "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.3"
+
+
+def test_update_image_does_not_overwrite_planner_service_image():
+    profiler_image = "nvcr.io/nvidia/ai-dynamo/dynamo-frontend:1.2.3"
+    worker_image = "nvcr.io/nvidia/ai-dynamo/vllm-runtime:1.2.3"
+    dgdr = _make_dgdr(profiler_image)
+    config = _base_dgd_config(profiler_image)
+
+    add_planner_to_config(dgdr, config)
+
+    updated = update_image(config, worker_image)
+
+    assert (
+        updated["spec"]["services"]["Frontend"]["extraPodSpec"]["mainContainer"][
+            "image"
+        ]
+        == worker_image
+    )
+    assert (
+        updated["spec"]["services"]["Planner"]["extraPodSpec"]["mainContainer"]["image"]
+        == "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.3"
+    )