VDK Blueprints Overview

Blueprints are the standardized context artifacts used to configure AI coding tools across platforms.

A blueprint describes behavioral intent first, then gets adapted by CLI into a tool-native carrier format.

Canonical Blueprint Kinds

Each blueprint belongs to exactly one canonical kind:

Vendor names are carriers, not canonical categories.

Curated Library Model

Blueprints are indexed along a second axis: specificity.

Why this matters

• L0-L3 gives users broad baseline + project-relevant depth.
• L4 stays opt-in to prevent recommendation noise.
• Curated growth can continue without polluting default retrieval quality.

CLI Deployment Contract

Blueprint deployment is executed by CLI:

• vdk search retrieves ranked candidates from curated inventory.
• vdk deploy <blueprint-id> installs deterministically into tool-native target paths.
• Adaptation outcomes are explicit: lossless, lossy, unsupported.

This is the operational heart of the ecosystem. Hub and Wiki mirror these semantics.

Blueprint Metadata Expectations

A production-ready blueprint should include:

• canonical kind
• schema version compatibility
• target platform component mappings
• specificity placement (L0..L4)
• adaptation/equivalence metadata when conversion is involved

Authoring Guardrails

1. Keep content schema-valid and deterministic.
2. Avoid duplicate top-level recommendations for provenance-only variants.
3. Place new blueprints deliberately in generic vs specific layers.
4. Preserve behavioral intent across adapters.

Practical Workflow

1. Discover candidates with vdk search.
2. Review compatibility and adaptation outcome.
3. Deploy selected artifact with vdk deploy <blueprint-id>.
4. Verify generated carrier files in target tool directories.

That workflow keeps the curated library effective while ensuring consistent runtime behavior across the ecosystem.