refactor: migrate kvbm python bindings to the python directory (#4318)

Signed-off-by: Ryan Olson <rolson@nvidia.com>

.github/workflows/pre-merge-rust.yml

@@ -39,6 +39,7 @@ jobs:
     runs-on:
       group: Fastchecker
     strategy:
+      # removing kvbm from here - it will fail to test with nixl dep enabled
       matrix: { dir: ['.', 'lib/bindings/python', 'lib/runtime/examples', 'launch/dynamo-run'] }
     permissions:
       contents: read

Cargo.toml

@@ -14,7 +14,6 @@ members = [
     "lib/bindings/c",
     "lib/bindings/python/codegen",
     "lib/engines/*",
-    "lib/kvbm",
     "lib/config",
 ]
 # Exclude certain packages that are slow to build and we don't ship as flagship
@@ -107,7 +106,9 @@ tokio = { version = "1", features = ["full"] }
 tokio-stream = { version = "0.1" }
 tokio-util = { version = "0.7", features = ["codec", "net", "rt"] }
 tower-http = { version = "0.6", features = ["trace"] }
-axum = { version = "0.8", features = ["macros"] }
+axum = { version = "=0.8.4", features = ["macros"] }
+hyper = { version = "=1.7.0" }
+hyper-util = { version = "=0.1.17" }
 tracing = { version = "0.1" }
 tracing-subscriber = { version = "0.3", features = [
     "env-filter",
@@ -134,3 +135,4 @@ insta.opt-level = 3
 # These make the build much slower but shrink the binary, and could help performance
 codegen-units = 1
 lto = "thin"
+

container/Dockerfile

@@ -317,7 +317,7 @@ RUN --mount=type=secret,id=aws-key-id,env=AWS_ACCESS_KEY_ID \
     cd /opt/dynamo/lib/bindings/python && \
     maturin build --release --out /opt/dynamo/dist && \
     if [ "$ENABLE_KVBM" = "true" ]; then \
-        cd /opt/dynamo/lib/kvbm && \
+        cd /opt/dynamo/lib/bindings/kvbm && \
         maturin build --release --out /opt/dynamo/dist; \
     fi && \
     /tmp/use-sccache.sh show-stats "Dynamo"

lib/kvbm/Cargo.toml → lib/bindings/kvbm/Cargo.toml

 # SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
+[workspace]
 
 [package]
 name = "kvbm-py3"
@@ -23,8 +24,8 @@ default = ["block-manager"]
 block-manager = ["dynamo-llm/block-manager", "dep:dlpark", "dep:cudarc"]
 
 [dependencies]
-dynamo-llm = { path = "../llm" }
-dynamo-runtime = { path = "../runtime" }
+dynamo-llm = { path = "../../llm" }
+dynamo-runtime = { path = "../../runtime" }
 
 anyhow = { version = "1" }
 async-stream = { version = "0.3" }

lib/kvbm/README.md → lib/bindings/kvbm/README.md

@@ -19,7 +19,7 @@ limitations under the License.
 
 The Dynamo KVBM is a distributed KV-cache block management system designed for scalable LLM inference. It cleanly separates memory management from inference runtimes (vLLM, TensorRT-LLM, and SGLang), enabling GPU↔CPU↔Disk/Remote tiering, asynchronous block offload/onboard, and efficient block reuse.
 
-![A block diagram showing a layered architecture view of Dynamo KV Block manager.](../../docs/images/kvbm-architecture.png)
+![A block diagram showing a layered architecture view of Dynamo KV Block manager.](../../../docs/images/kvbm-architecture.png)
 
 
 ## Feature Highlights
@@ -84,7 +84,7 @@ DYN_KVBM_CPU_CACHE_GB=100 vllm serve \
   Qwen/Qwen3-8B
 ```
 
-For more detailed integration with dynamo, disaggregated serving support and benchmarking, please check [vllm-setup](../../docs/kvbm/vllm-setup.md)
+For more detailed integration with dynamo, disaggregated serving support and benchmarking, please check [vllm-setup](../../../docs/kvbm/vllm-setup.md)
 
 ### TensorRT-LLM
 
@@ -106,12 +106,12 @@ DYN_KVBM_CPU_CACHE_GB=100 trtllm-serve Qwen/Qwen3-8B \
   --extra_llm_api_options /tmp/kvbm_llm_api_config.yaml
 ```
 
-For more detailed integration with dynamo and benchmarking, please check [trtllm-setup](../../docs/kvbm/trtllm-setup.md)
+For more detailed integration with dynamo and benchmarking, please check [trtllm-setup](../../../docs/kvbm/trtllm-setup.md)
 
 
 ## 📚 Docs
 
-- [Architecture](../../docs/kvbm/kvbm_architecture.md)
-- [Motivation](../../docs/kvbm/kvbm_motivation.md)
-- [Design Deepdive](../../docs/kvbm/kvbm_design_deepdive.md)
+- [Architecture](../../../docs/kvbm/kvbm_architecture.md)
+- [Motivation](../../../docs/kvbm/kvbm_motivation.md)
+- [Design Deepdive](../../../docs/kvbm/kvbm_design_deepdive.md)
 - [NIXL Overview](https://github.com/ai-dynamo/nixl/blob/main/docs/nixl.md)