Problem

When using Amazon Bedrock with application inference profiles, the Bedrock region is determined by AWS_REGION / AWS_DEFAULT_REGION environment variables. In enterprise environments, teams commonly set AWS_REGION to one region (e.g. us-east-1) for general AWS CLI operations, while their Bedrock inference profiles live in a different region (e.g. us-west-2).

This forces users to override AWS_REGION specifically when launching pi, which creates friction when running AWS CLI commands from within pi sessions that need the original region.

Current Behavior

Region resolution in getConfiguredBedrockRegion():

options.region || process.env.AWS_REGION || process.env.AWS_DEFAULT_REGION || undefined

The only way to set options.region is programmatically — there is no CLI flag or settings.json field exposed to users.

Proposed Solution

One or more of:

1. Parse region from ARN: If the model ID is an inference profile ARN (arn:aws:bedrock:<region>:<account>:application-inference-profile/...), extract and use the region from it automatically. This is the most ergonomic solution since the region is already embedded in the identifier.

2. Add a --region CLI flag: Allow pi --provider amazon-bedrock --region us-west-2 to set options.region directly.

3. Add a region field to settings.json: Allow persistent configuration:

   {
     "provider": "amazon-bedrock",
     "region": "us-west-2"
   }

4. Support a dedicated env var like AWS_BEDROCK_REGION that takes priority over AWS_REGION for Bedrock specifically.

Use Case

• AWS_REGION=us-east-1 set globally for all AWS CLI tooling
• Bedrock inference profiles provisioned in us-west-2
• Model configured as an inference profile ARN pointing to us-west-2

Currently requires wrapping pi launch with AWS_REGION=us-west-2 pi, which then breaks AWS CLI commands run from within the pi session that expect us-east-1.