aiconfig-variations

AI Config variations let you create and manage multiple model, prompt, and parameter combinations within a single config. Add variations to test different setups—such as agent-mode versus completion-mode—run A/B comparisons, and evaluate responses without modifying your original configuration.

Yes. Each variation supports optional tool attachments and evaluation settings, enabling you to test targeted configurations. This allows comprehensive experimentation across models, prompts, parameters, and tools simultaneously within the same AI Config.

The variations API supports add, update, retrieve, and delete operations on variations within an AI Config. Use these endpoints to programmatically create test configurations, compare model responses, and manage parameter experiments without manual config edits.

No. Variations enable A/B-style comparisons of models, prompts, and parameters as separate entries within your existing config. You preserve your base setup while testing different combinations for effective experimentation.

Use variations to test multiple setups within a single config for faster iteration and easier comparison. This approach keeps related experiments organized, supports batch evaluation, and simplifies rollback compared to managing separate configs for each test.