superintelligence 0.1.1
Terminal LLM agent that talks to a local OpenAI-compatible server and gets its hands on your files.
what it is
A minimal Rust CLI agent that chats with any OpenAI-compatible endpoint (llama-server, LiteLLM, vLLM) from the terminal. It runs interactively or one-shot, attaches images to prompts, and gives the model tools to read, search, and write files and run shell commands — with approval prompts for anything mutating. The shell tool is cross-platform: it runs commands with cmd.exe on Windows and sh on unix, so it works out of the box on either OS. Point src/config.rs at your local model and you have a hands-on agent for your own machine. Build with `cargo build --release` and run the `si` binary; the REPL opens with no args, or pass a prompt for one-shot. To share on the hub, run `make upload`.
files
11 files — expand to browse
try it
$ cargo build --release
$ si
vibe log
model: Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp.gguf
A terminal agent that chats with a local LLM and has file/shell tools, with approval for writes.
nd download JG2HQZWB/superintelligence
nd run JG2HQZWB/superintelligence
nd fork JG2HQZWB/superintelligence
/api/projects/superintelligence/context
/api/projects/superintelligence/context