Tailored AI
A lightweight, modular AI agent framework designed for local LLMs. Supports hot-reloadable tools, providers, and config — the agent can update its own capabilities without restarting. Runs as an interactive CLI, Discord bot, or HTTP service with a React web UI, and supports sub-agent delegation, cron scheduling, and a built-in knowledge base.
The Story
Most AI agent frameworks are designed for fast cloud APIs. I built this one for a different constraint: local 30B parameter models on consumer hardware. That meant keeping system prompts concise, tool counts low, and context tight. The agent supports hot-reloading its tools and config at runtime, can run as a Discord bot or cron scheduler, spawn async sub-agents for parallel work, and modify its own capabilities mid-conversation — all without an API bill.