chore: version bump for 0.2.1 release (#111)

Signed-off-by: Harrison King Saturley-Hall <hsaturleyhal@nvidia.com>
Signed-off-by: Harrison Saturley-Hall <hsaturleyhal@nvidia.com>

Author: saturley-hall

Cargo.lock

@@ -12,7 +12,7 @@ members = [
 resolver = "3"
 
 [workspace.package]
-version = "0.2.0"
+version = "0.2.1"
 edition = "2024"
 description = "Model Express: High-performance model serving and management"
 authors = ["NVIDIA Inc. <sw-dl-dynamo@nvidia.com>"]
@@ -34,9 +34,9 @@ hf-hub = { version = "0.4.3", default-features = false, features = [
     "rustls-tls",
 ] }
 jiff = { version = "0.2.15", features = ["serde"] }
-modelexpress-common = { path = "modelexpress_common", version = "0.2.0" }
-modelexpress-client = { path = "modelexpress_client", version = "0.2.0" }
-modelexpress-server = { path = "modelexpress_server", version = "0.2.0" }
+modelexpress-common = { path = "modelexpress_common", version = "0.2.1" }
+modelexpress-client = { path = "modelexpress_client", version = "0.2.1" }
+modelexpress-server = { path = "modelexpress_server", version = "0.2.1" }
 once_cell = "1.21.3"
 prost = "0.13"
 rusqlite = { version = "0.37", features = ["bundled", "chrono"] }

Cargo.toml

@@ -20,7 +20,7 @@ RUN cargo build --release --bin modelexpress-server && \
     cargo build --release --bin fallback_test
 
 # Create a minimal runtime image
-FROM nvcr.io/nvidia/base/ubuntu:noble-20250619
+FROM nvcr.io/nvidia/base/ubuntu:noble-20250619 AS runtime
 
 WORKDIR /app
 

Dockerfile

@@ -5,8 +5,8 @@ apiVersion: v2
 name: modelexpress
 description: A Helm chart for ModelExpress - a AI model cache management service
 type: application
-version: 0.2.0
-appVersion: "0.2.0"
+version: 0.2.1
+appVersion: "0.2.1"
 keywords:
   - model-serving
   - ai

helm/Chart.yaml

@@ -101,7 +101,7 @@ The following table lists the configurable parameters of the ModelExpress chart
 | `replicaCount`                               | Number of ModelExpress replicas                | `1`     |
 | `image.repository`                           | ModelExpress image repository                  | `nvcr.io/nvidia/ai-dynamo/modelexpress-server` |
 | `image.pullPolicy`                           | Image pull policy                              | `IfNotPresent` |
-| `image.tag`                                  | ModelExpress image tag                         | `0.2.0` |
+| `image.tag`                                  | ModelExpress image tag                         | `0.2.1` |
 | `imagePullSecrets`                           | Image pull secrets for nvcr.io access          | `[]`     |
 | `nameOverride`                               | Override the chart name                        | `""`     |
 | `fullnameOverride`                           | Override the full app name                     | `""`     |

helm/README.md

@@ -7,7 +7,7 @@ replicaCount: 1
 image:
   repository: nvcr.io/nvidia/ai-dynamo/modelexpress-server
   pullPolicy: IfNotPresent
-  tag: "0.2.0"
+  tag: "0.2.1"
 
 serviceAccount:
   create: true

helm/test-values.yaml

@@ -7,7 +7,7 @@ replicaCount: 1
 image:
   repository: nvcr.io/nvidia/ai-dynamo/modelexpress-server
   pullPolicy: IfNotPresent
-  tag: "0.2.0"
+  tag: "0.2.1"
 
 serviceAccount:
   create: true

helm/values-development.yaml

@@ -7,7 +7,7 @@ replicaCount: 3
 image:
   repository: nvcr.io/nvidia/ai-dynamo/modelexpress-server
   pullPolicy: Always
-  tag: "0.2.0"
+  tag: "0.2.1"
 
 serviceAccount:
   create: true

helm/values-production.yaml

@@ -10,7 +10,7 @@ replicaCount: 1
 image:
   repository: nvcr.io/nvidia/ai-dynamo/modelexpress-server
   pullPolicy: IfNotPresent
-  tag: "0.2.0"
+  tag: "0.2.1"
 
 imagePullSecrets:
   - name: nvcr-secret

helm/values.yaml

@@ -21,7 +21,7 @@ spec:
         - name: ngc-secret
       containers:
         - name: model-express-server
-          image: nvcr.io/nvidia/ai-dynamo/modelexpress-server:0.2.0
+          image: nvcr.io/nvidia/ai-dynamo/modelexpress-server:0.2.1
           imagePullPolicy: IfNotPresent
           ports:
             - containerPort: 8001