Repository navigation
Releases
Orca on AWS v0.2.0
New workstations now default to a g5.2xlarge with one NVIDIA A10G, a 30 GiB root disk and a 150 GiB encrypted workspace. Select --instance-type t3.large for a lower-cost CPU workstation. The GPU supports local development SDKs. Codex and OpenCode use Amazon Bedrock for inference. Review instance pricing and GPU driver requirements before creating a stack.
The seven-day field report records the reasons for the instance change, the independent desktop/mobile routes, and separate cloud and operating-system permissions. The quickstart and architecture diagram show how to reproduce the configuration.
Validation included 24 offline tests, ShellCheck, GitHub Actions, and read-only AWS CloudFormation plans for all three G5 agent profiles and the T3 alternative. The release changes fresh-stack defaults. It did not resize or redeploy the existing workstation.
Orca on AWS v0.1.0
Deploy an Orca workstation on Amazon EC2 with an encrypted, retained workspace, daily backups, Session Manager administration and independent desktop and Android access over Tailscale.
Choose Codex on Amazon Bedrock (GPT-6.2 Sol by default), OpenCode on Bedrock, or Kiro CLI with Kiro-managed sign-in. The quickstart covers a real inference check, pairing, costs and cleanup.
This tag passed the 21-test offline suite, ShellCheck, GitHub Actions, and AWS CloudFormation plan validation for all three agent profiles in us-east-1. Release checks did not provision a new EC2 instance. The deployer resolves a current OS image unless you supply an AMI ID, and package repositories can change.