Agent Skills: Instructions

Instructions

1. Define the Model File: A model file generally corresponds to a single database connection and includes Explores.
2. Required Parameters:
   • connection: "connection_name": Must match a connection defined in Looker Admin.
   • include: "pattern": Specifies which view and dashboard files are available to the model.
3. Best Practices:
   • Includes: Avoid include: "*.view" if possible to prevent performance issues and namespace clutter. Use specific paths or wildcards like include: "/views/users.view" or include: "/views/marketing/*.view".
   • Label: Use label: to provide a user-friendly name for the model in the UI.
   • Week Start Day: Set week_start_day: if the business logic requires a specific start day (e.g., monday).
   • Datagroups & Caching: ALWAYS use datagroups for caching policies to align Looker with your ETL/ELT processes.

4. Datagroups & Caching

Datagroups are the preferred mechanism for managing caching policies.

• Definition: Define in the model file.
• sql_trigger: A query that returns a single value (e.g., max timestamp). If the value changes, the cache is invalidated.
• max_cache_age: A fallback duration if the trigger doesn't change.
• persist_with: Apply the datagroup to Explores or the entire model.

Datagroups vs persist_for

| Feature | Datagroups (Recommended) | persist_for | | :------------- | :----------------------------------------- | :------------------------------------ | | Trigger | SQL Query (Smart) | Fixed Time (Dumb) | | Alignment | Aligns with ETL/ELT completion | Misaligned (guesswork) | | Management | Centralized in Model file | Scattered in Explores/Models | | Use Case | Production dashboards, ETL synchronization | Ad-hoc queries, Real-time (<1h) needs |

[!TIP] Use persist_for ONLY for real-time dashboards where you need to force a cache refresh every X minutes (e.g., stock tickers, fast-moving inventory). For everything else, use Datagroups.

5. Include Patterns

Use strict patterns to control scope and performance.

| Pattern | Description | Use Case | | :--- | :--- | :--- | | include: "/views/*.view" | All views in specific folder | Standard modularity | | include: "/views/marketing/users.view" | Specific file | Precise control, avoids conflicts | | include: "/**/*.view" | Recursive (all views in project) | Avoid unless small project | | include: "/dashboards/*.dashboard" | All dashboards in folder | Importing dashboards |

Examples

Basic Model

connection: "thelook"

# Include all views in the views/ folder
include: "/views/*.view"

# Include all dashboards
include: "/*.dashboard"

# Define an Explore (usually better to define in separate files for large projects, but acceptable here for small ones)
explore: orders {
  join: users {
    type: left_outer
    sql_on: ${orders.user_id} = ${users.id} ;;
    relationship: many_to_one
  }
}

Model with Specific Settings

connection: "snowlooker"

label: "eCommerce Analytics"

# Set valid week start day
week_start_day: monday

# Include specific folders
include: "/views/finance/*.view"
include: "/views/marketing/*.view"

## Model with Datagroup (Best Practice)

```lookml
connection: "thelook"

# Define the caching policy
datagroup: ecommerce_etl {
  description: "Triggers when the max created_at date changes in the events table."
  sql_trigger: SELECT MAX(created_at) FROM `project.dataset.events` ;;
  max_cache_age: "24 hours" # Fallback
}

# Apply default caching to all Explores in this model
persist_with: ecommerce_etl

include: "/views/*.view"

explore: orders {
  # This explore inherits 'persist_with: ecommerce_etl' from the model default
}

explore: real_time_dashboard {
  # Override with a different policy if needed
  persist_for: "5 minutes"
}
```