Aiconfig Targeting

This handles targeting rules for LaunchDarkly AI Configs, letting you control which model variations serve to different users through percentage rollouts, attribute-based conditions, and segment targeting. It's basically feature flag targeting but for AI model configurations. The workflow requires fetching variation UUIDs first since the API uses IDs instead of keys, and there's a gotcha: you can't use turnTargetingOn like you would with regular flags. Instead you update the fallthrough rule directly. Useful if you're running gradual rollouts of new prompts or models, or want to serve different configurations based on user tier or context attributes. Includes Python helpers that wrap the semantic patch API.

Install to Claude Code

npx -y skills add launchdarkly/agent-skills --skill aiconfig-targeting --agent claude-code