Local Model Toolkit v0.2.0

Add recipient-agnostic local model operations and benchmark guidance.

Recipients need a bounded CLI-first method to measure local model usefulness without bundling weights or machine-specific setup.

Model weights, credentials, private receipts and host paths remain external.

Compatibility: Adds a new toolkit; existing packages are unaffected.

Upgrade: Install the toolkit and configure a supported local runner with recipient-selected models.

Validation: Catalog, source inventory, package build, privacy audit and full release tests.

Adds a new toolkit; existing packages are unaffected.

Install the toolkit and configure a supported local runner with recipient-selected models.