2d ago
Per-Action Model & Provider Selection (Multi-Model / Multi-Agent Routing)
Currently, ActionClip allows choosing between Apple Intelligence, Ollama Local, and External AI Provider on a per-action basis, which is great for separating local and cloud tasks. However, External AI and Ollama Local are each locked to a single globally active provider and model in Settings.
I would love the ability to configure multiple AI providers/keys simultaneously and select a specific provider and model per action (both for built-in actions and Custom AI Prompt actions).
Use Cases & Benefits:
Cost & Speed Optimization: Use lightweight, fast, or inexpensive models (e.g., GPT-4o-mini, Gemini Flash, Claude Haiku, or a small 3B Ollama model) for simple transformations, quick translations, or formatting.
Task Specialization: Different models excel at different jobs. Having per-action model overrides makes it possible to dedicate Anthropic Claude to creative writing/email replies, OpenAI/Coding models to refactoring or prompt-improvement actions.
Proposed Implementation:
Multi-Provider Credentials in Settings: Allow users to store and verify API keys for multiple External AI providers (OpenAI, Anthropic, Gemini, OpenRouter, Custom API) at the same time, while designating one as the "Default External AI" and one local model as the "Default Ollama Model."
Per-Action Model Override: In the action settings (Actions → [Action] → Gear → Source) and in the Custom AI Action Builder, add an optional Model Override dropdown:
Use Default External AI (Settings)
Use Default Ollama Local (Settings)
Apple Intelligence
Specific Provider & Model
Pending