Skip to content

README.md

Latest commit

 

History

History
82 lines (56 loc) · 2.24 KB

README.md

File metadata and controls

82 lines (56 loc) · 2.24 KB
Raw

Models Only

models-only provisions remote ASR/TTS model endpoints without a backend host. It is intended for the workflow where users run backend and frontend locally and point them at model APIs running on EC2.

This is the current recommended deployment path for the public reference repository.

The stack creates its own EC2 IAM role and instance profile by default, so users do not need to pre-create IAM just to launch the model hosts.

Start with the main repository README if you want the short quickstart first.

What this stack creates

  • VPC networking for the model hosts
  • EC2-backed ASR runtime
  • EC2-backed TTS runtime
  • Security groups, IAM role, and instance profile for the model hosts

This stack does not create a hosted backend or hosted frontend. The expected workflow is to run those locally and point them at the EC2 model endpoints.

Deploy the stack

Apply the stack in this order:

cd iac/terraform/models-only/networking
terraform init
terraform apply -var-file="../common.tfvars"

cd ../compute
terraform init
terraform apply -var-file="../common.tfvars" -var-file="terraform.tfvars"

Useful outputs from compute/:

  • qwen3_asr_base_url
  • qwen3_tts_base_url

Configure the backend

After terraform apply, print the ready-to-paste backend environment values:

../scripts/print_model_env.sh

That prints values such as:

QWEN3_ASR_BASE_URL=http://...
QWEN3_TTS_BASE_URL=http://...

Copy those into backend/.env together with your SEALION_API_KEY.

This stack does not create any public DNS. It emits EC2-based URLs for local backend configuration, and can optionally keep the private *.internal names disabled by default with create_internal_dns = false.

Cleanup

Destroy the resources when you are done testing. The EC2 instances created by this stack continue to incur AWS charges until you tear them down.

Destroy in the reverse order of apply:

cd iac/terraform/models-only/compute
terraform destroy -var-file="../common.tfvars" -var-file="terraform.tfvars"

cd ../networking
terraform destroy -var-file="../common.tfvars"

Destroy compute first and networking second so Terraform can remove the EC2 instances and their dependent resources cleanly.