release: v2.6.0

build_artifacts/v2/v2.6/v2.6.0-alpha-cpu/Dockerfile

@@ -0,0 +1,8 @@
+FROM 700843992353.dkr.ecr.us-east-1.amazonaws.com/sagemaker-distribution-staging-repository@sha256:f72dd98369557b3235f8982a841e4c8439af370b901cb2d19ff7f5d60de24aa3
+
+ENV MAMBA_USER="sagemaker-user"
+USER $MAMBA_USER
+
+COPY sagemaker_ui_post_startup.sh /etc/sagemaker-ui/
+RUN micromamba install -y --name base -c conda-forge "amazon-sagemaker-jupyter-ai-q-developer=1.2.4" "jupyter-ai=2.31.1"
+

build_artifacts/v2/v2.6/v2.6.0-alpha-cpu/sagemaker_ui_post_startup.sh

@@ -0,0 +1,199 @@
+#!/bin/bash
+set -eux
+
+# Only run the following commands if SAGEMAKER_APP_TYPE_LOWERCASE is jupyterlab
+if [ "${SAGEMAKER_APP_TYPE_LOWERCASE}" = "jupyterlab" ]; then
+    # Override Jupyter AI config file with specific content only for 2.6
+    # this is a potential workaround for Q chat issue where the chat does not
+    # load since it does not pick up the latest config file
+    NB_USER=sagemaker-user
+    # Check if Jupyter AI config file exists, override with specific content if so only for 2.6
+    # this is a potential workaround for Q chat issue
+    JUPYTER_AI_CONFIG_PATH=/home/${NB_USER}/.local/share/jupyter/jupyter_ai/config.json
+    JUPYTER_AI_CONFIG_CONTENT='{
+        "model_provider_id": "amazon-q:Q-Developer",
+        "embeddings_provider_id": null,
+        "send_with_shift_enter": false,
+        "fields": {},
+        "api_keys": {},
+        "completions_model_provider_id": null,
+        "completions_fields": {},
+        "embeddings_fields": {}
+    }'
+
+    # Overwrite the file if it exists (or create it if it doesn't)
+    echo "$JUPYTER_AI_CONFIG_CONTENT" > "$JUPYTER_AI_CONFIG_PATH"
+fi
+
+# Writes script status to file. This file is read by an IDE extension responsible for dispatching UI post-startup-status to the user.
+write_status_to_file() {
+    local status="$1"
+    local message="$2"
+    local file="/tmp/.post-startup-status.json"
+
+    # Check if the file exists, if not, create it
+    if [ ! -f "$file" ]; then
+        touch "$file" || {
+            echo "Failed to create $file" >&2
+            return 0
+        }
+    fi
+
+    # Ensure the file is writable
+    if [ ! -w "$file" ]; then
+        echo "Error: $file is not writable" >&2
+        return 0
+    fi
+
+    # Create the JSON object and write to file
+    jq -n --arg status "$status" --arg message "$message" '{"status":$status,"message":$message}' > "$file"
+
+}
+
+# checks if the script status is "in-progress". If so, no errors were detected and it can be marked successful.
+write_status_to_file_on_script_complete() {
+    local file="/tmp/.post-startup-status.json"
+    local check_key="status"
+    local check_value="in-progress"
+
+
+    if jq -e --arg key "$check_key" --arg value "$check_value" '.[$key] == $value' "$file" > /dev/null; then
+        write_status_to_file "success" "IDE configured successfully."
+        echo "Post-startup script completed successfully. Success status written to $file"
+    else
+        echo "Skipping writing post-startup script "success" status. An error was detected during execution and written to $file".
+    fi
+}
+
+write_status_to_file "in-progress" "IDE configuration in progress."
+
+sourceMetaData=/opt/ml/metadata/resource-metadata.json
+
+# Extract the required fields from meta data stored in opt/ml/metadata.
+dataZoneDomainId=$(jq -r '.AdditionalMetadata.DataZoneDomainId' < $sourceMetaData)
+dataZoneUserId=$(jq -r '.AdditionalMetadata.DataZoneUserId' < $sourceMetaData)
+dataZoneProjectRepositoryName=$(jq -r '.AdditionalMetadata.DataZoneProjectRepositoryName' < $sourceMetaData)
+dataZoneEndPoint=$(jq -r '.AdditionalMetadata.DataZoneEndpoint' < $sourceMetaData)
+dataZoneProjectId=$(jq -r '.AdditionalMetadata.DataZoneProjectId' < $sourceMetaData)
+dataZoneDomainRegion=$(jq -r '.AdditionalMetadata.DataZoneDomainRegion' < $sourceMetaData)
+
+set +e
+
+# Creating a directory where the repository will be cloned
+mkdir -p $HOME/src
+
+# Remove the ~/.aws/config file to start clean when space restart
+rm -f /home/sagemaker-user/.aws/config
+echo "Successfully removed the ~/.aws/config file"
+
+aws configure set credential_source EcsContainer
+echo "Successfully configured default profile"
+
+# add SparkMonitor and Connection Magic entrypoint
+NB_USER=sagemaker-user
+
+config_path=/home/${NB_USER}/.ipython/profile_default/ipython_config.py
+
+if [ ! -f "$config_path" ] || ! grep -q "sagemaker_studio_dataengineering_sessions" "$config_path"; then
+  ipython profile create && echo "c.InteractiveShellApp.extensions.extend(['sagemaker_sparkmonitor.kernelextension','sagemaker_studio_dataengineering_sessions.sagemaker_connection_magic'])" >>  $config_path
+  cat << EOT >> "$config_path"
+c.Application.logging_config = {
+    "loggers": {
+        "": {
+            "level": "INFO",
+            # console handler is required to keep the default behavior of jupyter logging.
+            # https://jupyter-server.readthedocs.io/en/latest/operators/configuring-logging.html
+            "handlers": ["console"],
+        },
+    },
+}
+EOT
+fi
+
+# Setting this to +x to not log credentials from the response of fetching credentials.
+set +x
+
+# Note: The $? check immediately follows the sagemaker-studio command to ensure we're checking its exit status.
+# Adding commands between these lines could lead to incorrect error handling.
+response=$( sagemaker-studio credentials get-domain-execution-role-credential-in-space --domain-id "$dataZoneDomainId" --profile default)
+responseStatus=$?
+
+set -x
+
+if [ $responseStatus -ne 0 ]; then
+        echo "Failed to fetch domain execution role credentials. Will skip adding new credentials profile: DomainExecutionRoleCreds."
+        write_status_to_file "error" "Network issue detected. Your domain may be using a public subnet, which affects IDE functionality. Please contact your administrator."
+else
+        aws configure set credential_process "sagemaker-studio credentials get-domain-execution-role-credential-in-space --domain-id $dataZoneDomainId --profile default" --profile DomainExecutionRoleCreds
+        echo "Successfully configured DomainExecutionRoleCreds profile"
+fi
+
+echo "Starting execution of Git Cloning script"
+bash /etc/sagemaker-ui/git_clone.sh
+
+# Run AWS CLI command to get the username from DataZone User Profile.
+if [ ! -z "$dataZoneEndPoint" ]; then
+    response=$( aws datazone get-user-profile --endpoint-url "$dataZoneEndPoint" --domain-identifier "$dataZoneDomainId" --user-identifier "$dataZoneUserId" --region "$dataZoneDomainRegion" )
+else
+    response=$( aws datazone get-user-profile --domain-identifier "$dataZoneDomainId" --user-identifier "$dataZoneUserId" --region "$dataZoneDomainRegion" )
+fi
+
+# Extract the Auth Mode from the response. Unified Studio currently supports IAM, SSO and SAML.
+auth_mode=$(echo "$response" | jq -r '.type')
+
+case "$auth_mode" in
+    "IAM")
+        # For IAM users - extract IAM ARN from response. Response does not contain username or email.
+        arn=$(echo "$response" | jq -r '.details.iam.arn')
+        # Split ARN by / and return the last field
+        username=$(echo "$arn" | awk -F'/' '{print $NF}')
+        email="$arn"
+        ;;
+    "SSO"|"SAML")
+        # For SSO and SAML user, extract username and email if present in response.
+        username=$(echo "$response" | jq -r '.details.sso.username')
+        email=$(echo "$response" | jq -r '.details.sso.email')
+        # Setting up the email as username if email is not present
+        if [ -z "$email" ] || [ "$email" = "null" ]; then
+            email="$username"
+        fi
+        ;;
+    *)
+        echo "Unknown authentication mode: $auth_mode"
+        exit 1
+        ;;
+esac
+
+# Setting up the Git identity for the user .
+git config --global user.email "$email"
+git config --global user.name "$username"
+
+# MLFlow tracking server uses the LOGNAME environment variable to track identity. Set the LOGNAME to the username of the user associated with the space
+export LOGNAME=$username
+if grep -q "^LOGNAME=" ~/.bashrc; then
+  echo "LOGNAME is defined in the env"
+else
+  echo LOGNAME=$username >> ~/.bashrc
+  echo readonly LOGNAME >> ~/.bashrc
+fi
+
+set -e
+
+# write unexpected error to file if any of the remaining scripts fail.
+trap 'write_status_to_file "error" "An unexpected error occurred. Please stop and restart your space to retry."' ERR
+
+# Generate sagemaker pysdk intelligent default config
+nohup python /etc/sagemaker/sm_pysdk_default_config.py &
+# Only run the following commands if SAGEMAKER_APP_TYPE_LOWERCASE is jupyterlab
+if [ "${SAGEMAKER_APP_TYPE_LOWERCASE}" = "jupyterlab" ]; then
+    # Start workflows local runner
+    bash /etc/sagemaker-ui/workflows/start-workflows-container.sh
+
+    # Install conda and pip dependencies if lib mgmt config existing
+    bash /etc/sagemaker-ui/libmgmt/install-lib.sh $HOME/src
+
+    # Install sm-spark-cli
+    bash /etc/sagemaker-ui/workflows/sm-spark-cli-install.sh
+fi
+
+write_status_to_file_on_script_complete

build_artifacts/v2/v2.6/v2.6.0-alpha-gpu/Dockerfile

@@ -0,0 +1,8 @@
+FROM 700843992353.dkr.ecr.us-east-1.amazonaws.com/sagemaker-distribution-staging-repository@sha256:43b4a3af19ba16df3df3298ade0be6dae95c35f9c9454ba1d37bb7b7e6b29b5e
+
+ENV MAMBA_USER="sagemaker-user"
+USER $MAMBA_USER
+
+COPY sagemaker_ui_post_startup.sh /etc/sagemaker-ui/
+RUN micromamba install -y --name base -c conda-forge "amazon-sagemaker-jupyter-ai-q-developer=1.2.4" "jupyter-ai=2.31.1"
+