Skip to content

[FEATURE] support application inference profiles #105

@lepfhty

Description

@lepfhty

Feature description
Instead of system-defined cross-region inference profiles (e.g. "us.anthropic..."), support application inference profiles.

Problem it solves (optional)
My company uses application inference profiles for (a) cost tracking and (b) per-user per-model access controls. For example, Bob can only use the app-inference-profile assigned to him.

Proposed solution
Use list-inference-profiles to populate a dropdown of existing model ARNs, both system and application types.
Allow the user to select or type in a model ARN.
Determine dynamically (using get-inference-profile) what the base model is (e.g. sonnet 3, opus 4, etc) to determine model capabilities.

Roocode and Cline have an interface that allows the user to select a "base model" (the system defined cross-region profile) and enter a custom model id (the application inference profile model ARN).

Alternatives considered (optional)
use a different LLM front end

Additional context (optional)
n/a

Environment (optional)

  • macOS version: 15.5
  • Amazon Bedrock Client for Mac version: 1.3.10

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions