// the find
yomorun/yomo
🦖 Serverless AI Agent Framework with Geo-distributed Edge AI Infra.
YoMo is a Rust-based server that sits between your LLM provider and your tools, handling function calling dispatch and exposing an OpenAI-compatible API. The geo-distributed angle means you can run multiple zipper nodes closer to users, routing tool calls to nearby serverless functions. It's aimed at teams building latency-sensitive agent systems who don't want to manage the tool invocation plumbing themselves.
The OpenAI-compatible API surface means you can drop it in front of existing tooling without changing client code. Serverless tools written in TypeScript or Go are type-safe and isolated — each tool is its own deployable unit, not a function registered in a monolith. TLS 1.3 is on by default rather than an afterthought. The Rust core means the proxy layer itself won't be your bottleneck.
The README still has banned marketing phrases baked into the code comments and feature table, which makes it hard to tell what's real and what's aspirational. The geo-distributed story is thin on specifics — there's no documentation on how zipper nodes discover each other or how tool routing actually works across regions. Serverless tools are TypeScript or Go only; there's no first-class Python support, which will turn away a large chunk of the AI tooling ecosystem. The project recently rewrote from Go to Rust, so the git history and ecosystem maturity don't reflect the current codebase age.