yomo

YoMo is an open-source framework for LLM function calling, built for geo-distributed, low-latency deployment. You write tools as serverless functions, scaffold them with yomo init and run them with yomo run; the yomo server then exposes an OpenAI-compatible /v1/chat/completions endpoint that calls the right tool when the model asks for it. Tools can also be invoked directly over HTTP.

The pitch is edge placement: rather than keeping inference and tools in one data center far from users, YoMo is designed so you can deploy tools close to end users across regions, cutting response time. Every data packet is encrypted with TLS 1.3 by design.

Getting started is a CLI install (curl | sh), an LLM provider such as Ollama, then serve, init and run. The core and tools can be built in Rust from source, and tool functions are written in TypeScript. License is Apache-2.0. It suits teams adding function-calling tools to a chat product who care about latency and want an OpenAI-compatible surface.