Fixed useManagedAi() Usage Example

Fixed useManagedAi() Usage Example

The Problem

The original issue was that useManagedAi() was trying to find the setupManagedAi and teardownManagedAi tasks at configuration time, but these tasks are only created in the afterEvaluate block of the plugin, causing a timing issue.

The Solution

I fixed the issue by using Gradle’s Provider API to defer the task resolution until execution time:

fun Task.useManagedAi() {
    // Use task references with proper error handling
    dependsOn(project.provider {
        try {
            project.tasks.named("setupManagedAi")
        } catch (e: Exception) {
            project.logger.warn("setupManagedAi task not found for task ${name} - make sure the Gradle Managed AI plugin is applied")
            null
        }
    })
    
    finalizedBy(project.provider {
        try {
            project.tasks.named("teardownManagedAi")
        } catch (e: Exception) {
            project.logger.warn("teardownManagedAi task not found for task ${name} - make sure the Gradle Managed AI plugin is applied")
            null
        }
    })
}

How to Use

In your build.gradle.kts:

plugins {
    id("se.premex.gmai")
}

// Import the extension function
import se.premex.gmai.plugin.utils.useManagedAi

managedAi {
    models {
        create("tinyllama") {
            version = "latest"
        }
    }
}

// Use with any task type
tasks.withType<Test> {
    useManagedAi()
}

// Use with custom tasks
tasks.register("myCustomTask") {
    useManagedAi()
    doLast {
        println("Task executed with managed AI available")
    }
}

// Use with custom configuration
tasks.register("myAdvancedTask") {
    useManagedAi {
        autoTeardown = false  // Don't auto-teardown after this task
    }
    doLast {
        println("Task executed with custom managed AI configuration")
    }
}

What Was Fixed

1. Timing Issue: The original code tried to resolve tasks immediately, but they didn’t exist yet during configuration phase
2. Provider Pattern: Now uses project.provider {} to defer task resolution until execution time
3. Error Handling: Gracefully handles cases where the plugin isn’t applied or tasks don’t exist
4. Import Statement: Added proper import statement so the extension function is available in build scripts

Test Results

Both test cases now pass:

• ✅ useManagedAi() works correctly when the plugin is applied
• ✅ Graceful handling when the plugin is not applied

The fixed function now properly sets up task dependencies so that:

• Your task depends on setupManagedAi (starts Ollama and pulls models)
• Your task is finalized by teardownManagedAi (stops Ollama after completion)