smart-models
par samstone908
Coding Agents & IDEsRésumé
Intelligent multi-model router — automatically selects the best AI model based on task type (vision, image.
Description
Ce que fait ce skill
Routes tasks to the most appropriate AI model automatically based on task type: vision, image generation, video, audio, reasoning, code, or general chat. Supports 35+ models across 7 categories via any OpenAI-compatible API endpoint. Users can also target a specific model directly with @alias shortcuts like @o3, @sora, or @imagen.
Instead of tracking which model handles which task type, the router classifies the request and picks the right model automatically, with fallback to the next model in the category if a call fails.
Quand l'utiliser
- Analyzing a screenshot or photo without manually picking a vision model
- Generating an image from a text description routed to the best available model
- Running a math proof or complex logic problem through a reasoning-optimized model
- Producing speech or music without knowing which TTS or audio model to call
- Bypassing auto-routing to call a specific model directly with an @alias prefix
Exemple de workflow
User asks: @imagen a photorealistic mountain lake at sunset
- 1
Detects @imagen alias, maps it to google/imagen-4-ultra in the image_gen category
- 2
Skips auto-classification since alias is explicit
- 3
Calls scripts/call-model.sh with --model google/imagen-4-ultra --type image and the prompt
- 4
Receives the generated image URL from the API response
- 5
Returns the URL formatted as a markdown image
Markdown image link pointing to the generated photo
Prérequis
- SMART_ROUTER_BASE_URL environment variable (OpenAI-compatible API base URL, e.g. https://api.openai.com/v1)
- SMART_ROUTER_API_KEY environment variable (API key for the configured provider)
Installation
clawhub install samstone908/smart-modelsExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.