Create the Go project

Create a normal Go module for the coding assistant. Looprig does not require a generator, monorepo checkout, or a specific application framework.

Prerequisites

• Go 1.26 or newer
• A terminal
• An API key for a hosted model, or a local model server such as Ollama

You can complete the structural steps with the deterministic example client before choosing a hosted provider.

Create the module

Run these commands in a new directory of your choice:

mkdir looprig-coding-assistant
cd looprig-coding-assistant
go mod init example.com/looprig-coding-assistant

# Install only the boundaries used by the first model call.
go get github.com/looprig/core github.com/looprig/inference github.com/looprig/llm

As the tutorial adds capabilities, install their modules from the same project directory:

# Harness supplies the runtime. Tools, Fsstore, and Sandbox remain optional.
go get github.com/looprig/harness github.com/looprig/tools
go get github.com/looprig/fsstore github.com/looprig/storage
go get github.com/looprig/sandbox

go get resolves compatible module versions into your go.mod. Applications can apply their own upgrade and dependency-review policy.

Project layout

Use a small layout while learning the boundaries:

looprig-coding-assistant/
├── go.mod
├── go.sum
├── main.go          # CLI entry point and sandbox.Init
├── model.go         # Model selection and inference.Client construction
├── agent.go         # Loop, Rig, Session, and event handling
└── workspace/       # Files the tutorial assistant may inspect

The file split is for readability, not a framework requirement. All files use package main and compile into one application binary.

Runnable checkpoint

Open the complete first Inference example. It uses a deterministic client so the request and response path runs without credentials. The next page replaces that client boundary with OpenAI, Anthropic, or a local model.

Continue to connect a model with Inference.