A Python wrapper for inference.sh that handles the messy parts of running AI models and building agents. Automatically uploads files, manages streaming responses, and gives you both sync and async interfaces. The agent builder is solid - you can use pre-built templates or construct custom agents with tools, memory, and web search. Tool definitions are clean with proper type checking. Sessions keep workers warm between calls, which actually matters for expensive models. Good for RAG pipelines and automation where you need reliable model access without dealing with rate limits and infrastructure.
npx skills add https://github.com/inferen-sh/skills --skill python-sdk