This page is synced from app/docs/快速入门/onboarding.md in the GitHub repository. If it differs from the source, the repository wins.

Onboarding Guide

This guide helps new users understand Liri’s core concepts and workflow. If this is your first time using Liri, read it in order.

---

What is Liri?

Liri is an AI assistant that runs in the command-line terminal. You can ask it questions as if you were chatting, and it can:

• Answer technical questions and explain code
• Read, write, and search files
• Execute shell commands
• Search the web and fetch web pages
• Manage tasks and to-dos
• Write and debug code

---

Core concepts

Agent

The Agent is Liri’s core AI entity, responsible for understanding user intent, calling tools, and generating replies. Each session is managed by one Agent instance.

Tool

Tools are capability units the Agent can invoke, including file operations, web search, and code execution. The Agent automatically picks the right tool for each request.

Skill

Skills are predefined workflow templates that combine multiple tool calls into a complete workflow. For example, the “debug analysis” skill automatically collects logs, analyzes errors, and suggests fixes.

Session

A session represents one continuous interaction, including message history, context state, and configuration.

---

First-time workflow

Step 1: Start up

cd app
bun run dev

After starting, you will see:

═══════════════════════════════════════════════════════════════
  Liri - Interactive REPL
═══════════════════════════════════════════════════════════════

Welcome to Liri - AI Agent

ℹ Type a command to start, type exit to quit

Step 2: Explore

Enter /help to see all available commands. Enter /skill list to see the available skills.

Step 3: Ask questions

Just ask in natural language; the Agent handles the rest:

# Ask AI to read a file
Read the contents of src/main.ts

# Ask AI to search code
Search for the keyword "launch" in all TypeScript files

# Ask AI to search the web
Search the web for the latest TypeScript news

# Ask AI to run a command
Show me the files in the current directory

Step 4: Use skills

/skill debug "Error: Connection refused"
/skill review src/index.ts

---

Common features at a glance

📁 File operations

# Ask AI to read a file
Read config.json

# Ask AI to edit a file
Replace "old" with "new" in README.md

You can also use commands directly:

# Search file contents (like grep)
/grep "function" --type ts

# Edit a file (replace text)
/edit README.md old-text new-text

🌐 Web features

# Search the web
/websearch Python async programming tutorial

# Fetch web page content
/fetch https://example.com

📋 Task management

# List tasks
/task list

# Create a task
/task create "Finish documentation" "Write user guide"

# Update task status
/task update <task_id> completed

⚡ Running commands

# Run a shell command
/bash ls -la

# Specify timeout and working directory
/bash --timeout 10000 --cwd /project git status

🤖 Subagents (background tasks)

# Run an analysis task in the background (does not block the current session)
/subagent-run run explore "Analyze project structure" --background

# View background tasks
/subagent-run list

# Stop a task
/subagent-run stop <agent_id>

---

Run modes

Liri supports several ways to start:

Command	Mode	Use case
bun run dev	Development mode (hot reload)	Development and debugging
bun run start	Production run	Day-to-day use
bun run memory	Memory management CLI	Managing skills and tools
bun run src/index.ts	Direct run	Quick startup

---

Best practices

• Use /task to manage long-running tasks
• Use /config to adjust application settings
• Check the logs regularly to understand system health
• Use the skill system to boost your productivity