Deploy with Terraform | Sarvam API Docs

Prefer infrastructure-as-code? This Terraform module provisions everything a Sarvam endpoint needs — the IAM execution role, the model (from your Marketplace package ARN), the endpoint configuration, and the endpoint — in a single terraform apply. It works for Saaras v3, Bulbul v3, and Sarvam Vision; just point model_package_arn at the right package (all three run on ml.g6e.xlarge by default).

Prerequisites

Terraform 1.5+ and AWS credentials configured.
An active Marketplace subscription and the model package ARN for your region.

Variables

variables.tf

1 variable "region"            { type = string  default = "ap-south-1" }
2 variable "endpoint_name"     { type = string  default = "sarvam-endpoint" }
3 variable "model_package_arn" { type = string } # from your Marketplace subscription
4 variable "instance_type"     { type = string  default = "ml.g6e.xlarge" }
5 variable "instance_count"    { type = number  default = 1 }

Main configuration

main.tf

1 provider "aws" {
2   region = var.region
3 }
4 
5 # Execution role SageMaker assumes to run the endpoint
6 resource "aws_iam_role" "exec" {
7   name = "${var.endpoint_name}-exec"
8   assume_role_policy = jsonencode({
9     Version = "2012-10-17"
10     Statement = [{
11       Effect    = "Allow"
12       Principal = { Service = "sagemaker.amazonaws.com" }
13       Action    = "sts:AssumeRole"
14     }]
15   })
16 }
17 
18 resource "aws_iam_role_policy_attachment" "sagemaker" {
19   role       = aws_iam_role.exec.name
20   policy_arn = "arn:aws:iam::aws:policy/AmazonSageMakerFullAccess"
21 }
22 
23 # Model from the Marketplace package
24 resource "aws_sagemaker_model" "this" {
25   name                     = var.endpoint_name
26   execution_role_arn       = aws_iam_role.exec.arn
27   enable_network_isolation = true
28 
29   primary_container {
30     model_package_name = var.model_package_arn
31   }
32 }
33 
34 resource "aws_sagemaker_endpoint_configuration" "this" {
35   name = var.endpoint_name
36 
37   production_variants {
38     variant_name           = "AllTraffic"
39     model_name             = aws_sagemaker_model.this.name
40     instance_type          = var.instance_type
41     initial_instance_count = var.instance_count
42   }
43 }
44 
45 resource "aws_sagemaker_endpoint" "this" {
46   name                 = var.endpoint_name
47   endpoint_config_name = aws_sagemaker_endpoint_configuration.this.name
48 }
49 
50 output "endpoint_name" {
51   value = aws_sagemaker_endpoint.this.name
52 }

Deploy

$ terraform init
$ terraform plan  -var="model_package_arn=arn:aws:sagemaker:ap-south-1:<vendor>:model-package/<id>"
$ terraform apply -var="model_package_arn=arn:aws:sagemaker:ap-south-1:<vendor>:model-package/<id>"

Once apply finishes and the endpoint reaches InService, invoke it exactly as in the Deploy Saaras v3 or Deploy Vision guides.

For async endpoints, add an async_inference_config block (with an S3 output path) to the endpoint configuration and attach S3 permissions to the execution role. Add an aws_appautoscaling_target / aws_appautoscaling_policy pair to autoscale — including scale-to-zero for async.

Tear down

$ terraform destroy -var="model_package_arn=arn:aws:sagemaker:ap-south-1:<vendor>:model-package/<id>"