context blueprint

The blueprint command exports your data products as AI Blueprints — self-contained YAML files that capture the complete semantic specification following the Open Semantic Interchange (OSI) v1.0 standard.

An AI Blueprint is the Gold-tier deliverable. It contains every dataset, field, relationship, metric, business rule, guardrail filter, golden query, and glossary term for a data product in a single portable file.

Usage

context blueprint [model]

Arguments

Argument	Description
model	Model name to export (omit for all models)

Options

Option	Description	Default
--out <path>	Output directory for blueprint files	./blueprints
--stdout	Print to stdout instead of writing files	—
--context-dir <path>	Path to context directory	context/

Examples

Export all data products

context blueprint

Writes one .data-product.osi.yaml file per model to ./blueprints/.

Export a single data product

context blueprint sales-analytics

Print to stdout (for piping)

context blueprint sales-analytics --stdout

Custom output directory

context blueprint --out ./dist/blueprints

What’s in an AI Blueprint

sales-analytics.data-product.osi.yaml

# AI Blueprint — Open Semantic Interchange (OSI) v1.0

osi_version: "1.0"

semantic_model:
  name: sales-analytics
  description: Revenue and order analytics
  owner: analytics-team
  tier: gold
  trust_status: verified

  glossary:
    - term: Revenue
      definition: Total completed transaction amount
      related_fields:
        - orders.total_amount

  datasets:
    - name: orders
      description: Customer orders
      primary_key: [order_id]
      fields:
        - name: total_amount
          semantic_role: measure
          aggregation: SUM
          # ... all fields with full metadata

  relationships:
    - name: orders_to_customers
      from: { dataset: orders, columns: [customer_id] }
      to: { dataset: customers, columns: [id] }
      cardinality: many_to_one

  metrics:
    - name: total_revenue
      expression: SUM(total_amount)

  business_rules:
    - name: completed_orders_only
      definition: Only include completed orders in revenue calculations
      applied_always: true

  guardrail_filters:
    - name: exclude_test_orders
      filter: "status != 'test'"
      tables: [orders]

  golden_queries:
    - question: "Top customers by revenue last quarter"
      sql: |
        SELECT customer_id, SUM(total_amount) ...

How to use AI Blueprints

• Share them — hand a blueprint to any team or tool that needs to understand your data
• Serve via MCP — the context://data-product/{name} resource exposes blueprints to AI agents
• Import them — any OSI-compliant tool can consume the YAML directly
• Version them — commit blueprints to git alongside your code

Next steps

• Serve blueprints to agents: context serve
• Check what’s missing for Gold: context tier
• Create more data products: context new