sofianhamiti

Component	Purpose
Fargate	Serveless compute for LiteLLM
Aplication Load Balancer (internal)	Route traffic to LiteLLM, perform health checks
Aurora PostgreSQL	Serverless database for LiteLLM config, API keys, and usage metrics
Bedrock	Provide access to foundation models
S3 Bucket	Capture LiteLLM logs for monitoring and troubleshooting
Private VPC Subnets	Isolate your LiteLLM stack from direct internet exposure
Client VPN Endpoint	Provide secure, certificate-based access to LiteLLM in private subnet
OpenVPN Connect	Client to connect to VPN Endpoint

sofianhamiti

# LiteLLM Configuration (see https://docs.litellm.ai/docs/proxy/config_settings)

# -----------------------------------------------------
# General application settings - Using Environment Variables from ECS Fargate
# -----------------------------------------------------
general_settings:
  store_prompts_in_spend_logs: true
  master_key: os.environ/LITELLM_MASTER_KEY
  salt_key: os.environ/LITELLM_SALT_KEY
  database_url: os.environ/DATABASE_URL

sofianhamiti

# -----------------------------------------------------
# Global Model Configuration
# -----------------------------------------------------
model_defaults: &model_defaults
  model: "bedrock/us.anthropic.claude-3-7-sonnet-20250219-v1:0" # Bedrock inference profile (see https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-support.html)
  tpm: 20000
  rpm: 5
  aws_region_name: ${AWS_REGION}  

# -----------------------------------------------------

sofianhamiti

FROM public.ecr.aws/sagemaker/sagemaker-distribution:latest-cpu

# Set environment variables
ARG NB_USER="sagemaker-user"
ARG NB_UID=1000
ARG NB_GID=100
ENV MAMBA_USER=$NB_USER

USER root

sofianhamiti

{
  "nameShort": "code-server",
  "nameLong": "code-server",
  "applicationName": "code-server",
  "dataFolderName": ".code-server",
  "win32MutexName": "codeserver",
  "licenseName": "MIT",
  "licenseUrl": "https://github.com/coder/code-server/blob/main/LICENSE",
  "serverLicenseUrl": "https://github.com/microsoft/vscode/blob/main/LICENSE.txt",
  "serverGreeting": [],

sofianhamiti

cd /home

apt-get update && sudo apt-get install -y gnupg software-properties-common

# Install the HashiCorp GPG key.
wget -O- https://apt.releases.hashicorp.com/gpg | \
gpg --dearmor | \
tee /usr/share/keyrings/hashicorp-archive-keyring.gpg

# Verify the key's fingerprint.

sofianhamiti

#!/bin/bash

set -eux

source activate studio

pip install jupyterlab_s3_browser

jupyter serverextension enable --py jupyterlab_s3_browser

sofianhamiti

{
    "data-root": "/home/ec2-user/SageMaker/docker",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []        }
    }
}

sofianhamiti

FROM python:3.10.9


# INSTALL CODE-SERVER IN CONTAINER
RUN curl -fsSL https://code-server.dev/install.sh | sh -s -- --version=4.9.1

# POINT VS CODE (IN CONTAINER) TO THE CONFIGURATION (ON NOTEBOOK INSTANCE)
# ENV XDG_DATA_HOME=/home/SageMaker/vscode-config

sofianhamiti

{
  "nameShort": "code-server",
  "nameLong": "code-server",
  "applicationName": "code-server",
  "dataFolderName": ".code-server",
  "win32MutexName": "codeserver",
  "licenseName": "MIT",
  "licenseUrl": "https://github.com/coder/code-server/blob/main/LICENSE",
  "serverGreeting": [],
  "serverLicense": [],