.env.example

# AWS Configuration
AWS_REGION=us-west-2
# AWS_PROFILE is optional. Leave it commented out to use your default AWS
# credentials. Set it to a named profile only if the account you want to deploy
# to is not your default one. Leave it commented (not blank) when unused — an
# empty value is treated by the AWS CLI as a profile literally named "".
# AWS_PROFILE=
#
# AWS_ACCOUNT_ID is optional but recommended. Set it to the 12-digit account you
# intend to deploy to, and every deploy first asserts the live AWS account
# matches — refusing to run (and exporting nothing) if your credentials resolve
# to a different account. It also backstops Terraform via allowed_account_ids.
# Leave it commented out to skip the guard.
# AWS_ACCOUNT_ID=123456789012

# Deployment Configuration
# AWS_REGION, ENVIRONMENT, and PROJECT_NAME are required and drive both the
# deploy scripts and Terraform (passed as -var). Deploy one environment per AWS
# account.
PYTHON_RUNTIME=PYTHON_3_13
ENVIRONMENT=dev
PROJECT_NAME=assess-workbench

# Model Configuration
# PLANNER / JUDGE / IMAGE_ANALYSIS are consumed by the deploy (passed to
# Terraform automatically) and are required.
#
# Model IDs use a cross-Region inference-profile prefix. The default `global.`
# routes to any commercial Region and works from every Region — so changing
# AWS_REGION above is all you need to deploy elsewhere. For data residency,
# swap `global.` for your geography's prefix (the request then stays within
# that geo): `us.` (US), `eu.` (Europe), `au.` (Australia), `jp.` (Japan).
# Not every model is offered in every geo/Region — check the model's detail
# page under "Regional availability" in the Bedrock docs.
#
# These three are the workflow models (planner, judge, image analysis). The
# review and chat AgentCore agents set their model per-agent in
# agents/<name>/agent.yaml (seeded into the registry by `task deploy:seed`).
PLANNER_MODEL_ID=global.anthropic.claude-sonnet-4-6
JUDGE_MODEL_ID=global.anthropic.claude-sonnet-4-6
IMAGE_ANALYSIS_MODEL_ID=global.anthropic.claude-sonnet-4-6
# GENERATOR is optional — used only by the standards-authoring helper
# (scripts/generate_standard.py). It is not part of the deploy; the script
# falls back to a built-in default if unset.
GENERATOR_MODEL_ID=global.anthropic.claude-sonnet-4-6

# Test Configuration (used by tests/live/)
# TEST_USER_EMAIL=your-cognito-user@example.com
# TEST_USER_PASSWORD=your-password