Skip to content

Latest commit

 

History

History
193 lines (147 loc) · 10.3 KB

README.md

File metadata and controls

193 lines (147 loc) · 10.3 KB

Deployment README.md

The application leverages Terraform to define and provision the underlying infrastructure. The application uses CloudBuild to build and deploy the containers for the application.

Setup

Prerequisites:

  1. A Google Cloud project.
  2. Terraform installed on your local machine.
  3. The account running the terraform must have the necessary permissions. Recommended roles listed below:
  • roles/serviceusage.serviceUsageAdmin
  • roles/resourcemanager.projectIamAdmin
  • roles/iam.serviceAccountAdmin
  • roles/iam.serviceAccountUser
  • roles/storage.admin
  • roles/artifactregistry.admin
  • roles/run.admin
  • roles/cloudbuild.builds.editor

Step-by-Step Guide (Cloud Build)

  1. Configure Env Variables

    • Edit deployment/config/dev.yaml with your Google Cloud settings. There are certain variables that must be set manually and some variables that will be auto-populated if you run build.py. These variables that must be set manually are marked below as required.
    Variable Description Required
    PROJECT_ID Google Cloud Project ID for resource deployment. Yes
    VERTEX_AI_LOCATION The region to use for the various resources. Yes
    AGENT_BUILDER_LOCATION The region to use for Agent Builder (options shown below). Yes
    AGENT_INDUSTRY_TYPE The selected industry type (options shown below). Yes
    AGENT_ORCHESTRATION_FRAMEWORK The selected agent framework (options shown below). Yes
    AGENT_FOUNDATION_MODEL The selected foundation model (options shown below). Yes
    USER_AGENT The name of the agent. Yes
    AGENT_DESCRIPTION The description of the agent. Yes
    FRONTEND_URL The URL of frontend Cloud Run service. Used for CORS policy setting. No
    GCS_STAGING_BUCKET The name of the Staging bucket for Vertex AI resources. No
    DATA_STORE_ID The ID of the Vertex AI Search Datastore No
    GOOGLE_GENAI_USE_VERTEXAI Boolean value (set to "TRUE" if using ADK) No
    GOOGLE_CLOUD_PROJECT Google Cloud Project ID for resource deployment. (Used by ADK) No
    GOOGLE_CLOUD_LOCATION The region to use for the various resources. (Used by ADK) No

    • Options:

      AGENT_BUILDER_LOCATION:

      • "us"
      • "global"

      AGENT_INDUSTRY_TYPE:

      • "finance"
      • "healthcare"
      • "retail"

      AGENT_ORCHESTRATION_FRAMEWORK:

      • "langchain_prebuilt_agent"
      • "langchain_vertex_ai_agent_engine_agent" # use if using Agent Engine deployment
      • "langgraph_prebuilt_agent"
      • "langgraph_vertex_ai_agent_engine_agent" # use if using Agent Engine deployment
      • "llamaindex_agent" # can use with either Agent Engine or Cloud Run deployment
      • "agent_development_kit_agent" # can use with either Agent Engine or Cloud Run deployment"

      AGENT_FOUNDATION_MODEL:

      • "gemini-2.5-pro"
      • "gemini-2.5-flash"
      • "gemini-2.5-flash-lite"
      • "gemini-2.0-flash"
      • "gemini-1.5-pro"
      • "gemini-1.5-flash"
      • "claude-3-7-sonnet" # requires permission / signing T&C; must be us-east5
      • "claude-3-5-sonnet-v2" # requires permission / signing T&C; must be us-east5
      • "llama-4-maverick-17b-128e-instruct-maas" # requires the model to be enabled
      • "llama-4-scout-17b-16e-instruct-maas" # requires the model to be enabled
      • "llama-3.3-70b-instruct-maas" # requires the model to be enabled
      • "llama-3.1-405b-instruct-maas" # requires the model to be enabled

    • Example Configuration:

    PROJECT_ID: next-2024-industry-demos
VERTEX_AI_LOCATION: us-central1
AGENT_BUILDER_LOCATION: us
AGENT_INDUSTRY_TYPE: finance
AGENT_ORCHESTRATION_FRAMEWORK: langgraph_prebuilt_agent
AGENT_FOUNDATION_MODEL: gemini-2.0-flash
USER_AGENT: agentsmithy-starter-agent
AGENT_DESCRIPTION: "This is a test agent"

    • If you want your agent runtime to use an Agent Engine deployment, specify this using the AGENT_ENGINE_RESOURCE_ID variable shown below. Do not define this variable if you do not want to use Agent Engine. Example: AGENT_ENGINE_RESOURCE_ID: projects/599247973214/locations/us-central1/reasoningEngines/5008011581729013760

    • If you want to use Agent Development Kit as your orchestration framework, specify this using the GOOGLE_GENAI_USE_VERTEXAI, GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION variables shown below. Example:

    GOOGLE_GENAI_USE_VERTEXAI: 'TRUE' GOOGLE_CLOUD_PROJECT: next-2024-industry-demos GOOGLE_CLOUD_LOCATION: us-central1

  2. Run Cloud Build Job

    • The file deployment/cd/dev.yaml contains a Cloud Build Manifest that will build and push your agent backend container and deploy it to Cloud Run. There are a couple of substitution variables to be set when using this file.
    Variable Description Required
    _CONTAINER_NAME The name of the backend container. Defaults to "agent_runtime" No
    _PROJECT_ID Google Cloud Project ID for resource deployment. Yes
    _ARTIFACT_REGISTRY_REPO_NAME Artifact registry for containers. Yes
    _SERVICE_NAME The name of the backend Cloud Run service. Defaults to "agent-runtime-service" No
    _REGION The region to use for the container and Cloud Run service. Defaults to "us-central1" No
    • Example run: gcloud builds submit --config deployment/cd/dev.yaml --substitutions _CONTAINER_NAME=agent_runtime,_PROJECT_ID=next-2024-industry-demos,_ARTIFACT_REGISTRY_REPO_NAME=my_artifact_registry,_SERVICE_NAME=agent-runtime-service,_REGION=us-central1 .

Step-by-Step Guide (Terraform)

  1. Configure Terraform Variables

    • Edit deployment/terraform/vars/env.tfvars with your Google Cloud settings. These will be auto-populated if you run build.py, which will source these variables from deployment/config/dev.yaml. Terraform will use prod_project_id by default.
    Variable Description Required
    prod_project_id Google Cloud Project ID for resource deployment. [prod] Yes
    stage_project_id Google Cloud Project ID for resource deployment. [stage] No
    dev_project_id Google Cloud Project ID for resource deployment. [dev] No
    region Google Cloud region for resource deployment. Yes
    agent_name Name of the Agent, used for creating resources. Yes
    default_agents_prefix Default naming prefix for resources. Yes
    vertex_ai_staging_bucket Staging bucket for Vertex AI resources. Yes
    artifact_registry_repo_name Artifact registry for containers. Yes
    backend_cloud_run_service_name The name of the backend Cloud Run service. Yes
    frontend_cloud_run_service_name The name of the frontend Cloud Run service. Yes

  2. Deploy Infrastructure with Terraform

    • First, open a terminal and enable required Google Cloud APIs:
    gcloud config set project $YOUR_DEV_PROJECT
gcloud services enable serviceusage.googleapis.com cloudresourcemanager.googleapis.com
    • Then, navigate to the Terraform directory:
    cd deployment/terraform
    • Initialize Terraform:
    terraform init
    • Apply the Terraform configuration:
    terraform apply --var-file vars/env.tfvars
    • Type 'yes' when prompted to confirm

After completing these steps, your infrastructure will be set up and ready for deployment!

  1. De-Provision Infrastructure with Terraform

    • Open a terminal and navigate to the Terraform directory:
    cd deployment/terraform
    • Initialize Terraform:
    terraform init
    • Apply the Terraform configuration:
    terraform destroy --var-file vars/env.tfvars
    • Type 'yes' when prompted to confirm

  2. There is an optional counter that sents a request to our AgentSmithy server which increments when this terraform script is run. This allows us to track number of runs. The post request is empty and no information is sent over to this server other than an empty request. If you wish to opt-out of this counter, you can do so by setting the following terraform/variables.tf block to false:

variable "increment_runs" {
  type        = bool
  description = "Whether to increment installs counter."
  default     = false
}