Wayfinder Router Sends Simple Tasks to Local AI and Hard Ones to Cloud
Wayfinder Router addresses the cost and latency overhead of AI model routers that themselves call an LLM to decide routing, by using a deterministic offline scoring system instead.
Reporting from 1 sources: GIGAZINE.
An open-source tool called Wayfinder Router has been released on GitHub. It automatically routes simple AI tasks like summarization to a local model and complex tasks like code analysis to a cloud model, using structural features of the prompt rather than an AI judge to decide.
The tool, called Wayfinder Router, is now available on GitHub. It reads prompt features such as length, headings, lists, and code presence to calculate a complexity score from 0.0 to 1.0. Low-scoring tasks go to a local model like Ollama or vLLM, while high-scoring tasks go to a cloud model. The developer says the approach avoids paying top-tier fees for simple prompts like summaries and typo corrections.
Wayfinder supports OpenAI-compatible APIs. Users point their chat UI or IDE tool's base_url to Wayfinder, and it distributes tasks automatically. The response includes headers showing which model handled the request and the complexity score. Slash commands like /local, /cloud, or /auto let users override routing per request.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.