SentinelLocal
Codeberg

Getting started

Installation

SentinelLocal builds from source as a standard Rust workspace.

Prerequisites

  • Rust 1.75+ — install via rustup.
  • Ollama — only required for the agent, chat, and serve subcommands. Install it from ollama.com/download; the app starts and stops ollama serve for you. You still need a tool-call-capable model pulled (the app will offer to pull a missing one): 
ollama pull llama3.2          # or any tool-call-capable model

Build & install

cargo build --release
./target/release/sentinel --help

The deterministic scanners run with no LLM required — only the agent-driven subcommands need Ollama. See the quick start for running them.

Automatic Ollama management

You do not have to start Ollama yourself. When you run agent, chat, or serve, SentinelLocal:

  1. probes the configured host for a running Ollama;
  2. if none is found, locates the ollama binary and starts ollama serve in the background, then waits for it to become healthy;
  3. checks the requested model is pulled, prompting before downloading a missing one (Pull now? [y/N]);
  4. on exit, stops only a server it started — a pre-existing ollama serve is left running.

The binary is located via OLLAMA_BIN, then PATH, then the macOS app bundle. If Ollama is not installed, you get a clear install hint and a clean exit.

Supported platforms

The current prototype runs on Linux, macOS, and Windows. Scanner parity and CI across all three platforms is on the roadmap.

View the source on Codeberg →