feat: decouple bento dependency (#1266)

Author: biswapanda

container/Dockerfile.sglang

@@ -347,9 +347,6 @@ RUN mkdir -p /opt/dynamo/bindings/wheels && \
 RUN uv pip install /workspace/dist/ai_dynamo_runtime*cp312*.whl && \
     uv pip install /workspace/dist/ai_dynamo*any.whl
 
-# TODO: remove kantoku as a transitive dependency
-RUN uv pip uninstall kantoku circus && uv pip install circus==0.19.0
-
 # Copy launch banner
 RUN --mount=type=bind,source=./container/launch_message.txt,target=/workspace/launch_message.txt \
     sed '/^#\s/d' /workspace/launch_message.txt > ~/.launch_screen && \
@@ -399,9 +396,6 @@ RUN uv pip install ai-dynamo[vllm] --find-links wheelhouse && \
     ln -sf $VIRTUAL_ENV/bin/* /usr/local/bin/ && \
     rm -r wheelhouse
 
-# TODO: remove kantoku as a transitive dependency
-RUN uv pip uninstall kantoku circus && uv pip install circus==0.19.0
-
 # Tell vllm to use the Dynamo LLM C API for KV Cache Routing
 ENV VLLM_KV_CAPI_PATH="/opt/dynamo/bindings/lib/libdynamo_llm_capi.so"
 

container/Dockerfile.tensorrt_llm

@@ -323,10 +323,6 @@ RUN . /opt/dynamo/venv/bin/activate && \
     uv pip install /workspace/dist/ai_dynamo_runtime*cp312*.whl && \
     uv pip install /workspace/dist/ai_dynamo*any.whl
 
-# TODO: remove kantoku as a transitive dependency
-RUN . /opt/dynamo/venv/bin/activate && \
-    uv pip uninstall kantoku circus && uv pip install circus==0.19.0
-
 # Install dynamo.runtime and dynamo.llm wheels globally in container for tests
 # TODO: In future, we may use a virtualenv for everything and remove this.
 RUN pip install dist/ai_dynamo_runtime*cp312*.whl  && \

container/Dockerfile.vllm

@@ -438,9 +438,6 @@ RUN mkdir -p /opt/dynamo/bindings/wheels && \
 RUN uv pip install /workspace/dist/ai_dynamo_runtime*cp312*.whl && \
     uv pip install /workspace/dist/ai_dynamo*any.whl
 
-# TODO: remove kantoku as a transitive dependency
-RUN uv pip uninstall kantoku circus && uv pip install circus==0.19.0
-
 # Copy launch banner
 RUN --mount=type=bind,source=./container/launch_message.txt,target=/workspace/launch_message.txt \
     sed '/^#\s/d' /workspace/launch_message.txt > ~/.launch_screen && \
@@ -517,9 +514,6 @@ RUN uv pip install ai-dynamo[vllm] --find-links wheelhouse && \
     ln -sf $VIRTUAL_ENV/bin/* /usr/local/bin/ && \
     rm -r wheelhouse
 
-# TODO: remove kantoku as a transitive dependency
-RUN uv pip uninstall kantoku circus && uv pip install circus==0.19.0
-
 # Tell vllm to use the Dynamo LLM C API for KV Cache Routing
 ENV VLLM_KV_CAPI_PATH="/opt/dynamo/bindings/lib/libdynamo_llm_capi.so"
 

deploy/helm/deploy.sh

@@ -87,4 +87,4 @@ cd -
 # Install the Helm chart with the correct tag (SHA)
 echo "Installing Helm chart with image: $docker_tag_for_registry"
 HELM_RELEASE="${DYNAMO_MODULE//_/\-}"
-helm upgrade -i "$HELM_RELEASE" ./chart -f ~/bentoml/bentos/"$DYNAMO_NAME"/"$docker_sha"/bento.yaml --set image="$docker_tag_for_registry" --set dynamoIdentifier="$DYNAMO_IDENTIFIER" --set configFilePath="$DYNAMO_CONFIG_FILE" -n "$NAMESPACE"
+helm upgrade -i "$HELM_RELEASE" ./chart -f ~/.dynamo/packages/"$DYNAMO_MODULE"/"$DYNAMO_NAME"/dynamo.yaml --set image="$docker_tag_for_registry" --set dynamoIdentifier="$DYNAMO_IDENTIFIER" --set configFilePath="$DYNAMO_CONFIG_FILE" -n "$NAMESPACE"

deploy/sdk/src/dynamo/sdk/__init__.py

@@ -20,19 +20,16 @@
 # TODO: Remove this line after the bentoml import is removed from this file
 warnings.filterwarnings("ignore", category=UserWarning, message=".*pkg_resources.*")
 
-# flake8: noqa: E402
-from bentoml import on_shutdown as async_on_shutdown
-
 # flake8: noqa: E402
 from dynamo.sdk.core.decorators.endpoint import api, endpoint
 from dynamo.sdk.core.lib import DYNAMO_IMAGE, depends, liveness, readiness, service
-from dynamo.sdk.lib.decorators import async_on_start
+from dynamo.sdk.lib.decorators import async_on_start, on_shutdown
 
 dynamo_context: dict[str, Any] = {}
 
 __all__ = [
     "DYNAMO_IMAGE",
-    "async_on_shutdown",
+    "on_shutdown",
     "async_on_start",
     "depends",
     "dynamo_context",

deploy/sdk/src/dynamo/sdk/cli/allocator.py

@@ -21,8 +21,7 @@
 import os
 from typing import Any
 
-from _bentoml_sdk import Service
-from simple_di import inject
+from dynamo.sdk.core.protocol.interface import ServiceInterface
 
 # Import our own resource module
 from dynamo.sdk.lib.resource import (
@@ -149,10 +148,9 @@ def get_gpu_stats(self) -> list[dict[str, Any]]:
         """Get detailed statistics for all GPUs."""
         return self.gpu_manager.get_gpu_stats()
 
-    @inject
     def get_resource_envs(
         self,
-        service: Service[Any],
+        service: ServiceInterface[Any],
     ) -> tuple[int, list[dict[str, str]]]:
         """
         Get resource environment variables for a service.