Install and create your first project

Install the Robo Architect Lite desktop app and launch it for the first time, check that it brings up the graph DB and analysis services in Docker, then create your first project, “hanbit_delivery_en”, on Home, choose its code folder and open it.

What gets installed#

Lite runs as one desktop app and seven Docker containers. When you start the app, it brings up the containers itself; when you quit, it leaves them running (so the next start is faster).

Neo4j is not used. Earlier versions of Robo Architect kept the design in Neo4j, but Lite uses Ontological DB, where the graph DB itself also connects to your data sources.

Prerequisites#

Hands-on 1 — Install#

The Lite installer comes as three files (ask uEngine for them). These steps are for Windows.

1. Open PowerShell in the download folder and check that the hashes of both files match SHA256SUMS.

   Get-Content .\SHA256SUMS
   Get-ChildItem .\Robo-Architect-Setup-*.exe, .\robo-images.tar |
     ForEach-Object { "{0}  {1}" -f (Get-FileHash -Algorithm SHA256 $_).Hash.ToLower(), $_.Name }
   

2. Run Robo-Architect-Setup-<version>.exe. You don't need administrator rights. Uncheck the “Run” option on the last screen — an app launched from the installer inherits the environment the installer started with.
3. Move the image bundle into the app's data folder. If you skip this step, the app stops at start-up — Home shows “Ontological stopped · Could not start — see the logs folder”, and the log in the logs folder says docker.image_archive_missing.

   $dst = "$env:APPDATA\robo-architect-desktop\runtime"
   New-Item -ItemType Directory -Force -Path $dst | Out-Null
   Copy-Item .\robo-images.tar -Destination $dst -Force
   

4. (Only if you will use document upload in the Requirements tab) Set your OpenAI key as a user environment variable, then sign out and sign back in. Each service reads a different name, so put the same value under five names.

   foreach ($n in 'OPENAI_API_KEY','LLM_API_KEY','ROBO_LLM_API_KEY','ROBO_EMBED_API_KEY','ROBO_SEARCH_LLM_API_KEY') {
     [Environment]::SetEnvironmentVariable($n, '<your key>', 'User')
   }
   

Hands-on 2 — First launch#

Start the app from the Robo-Architect icon on your desktop. The top of the window is the tab strip, and the first tab that opens is Home. While “Starting the graph DB and services…” is shown in the middle of Home, the app is bringing up the seven Docker containers. On the computer used for this tutorial, the first start took about 50 seconds (with the installer, the first start takes 4–5 minutes because the image bundle is unpacked). Until then the feature cards are dimmed and locked.

When everything is up, the status at bottom left changes to Ontological connected, and the project picker opens in the middle. There is no sign-in — Lite is a single-user app with no accounts, and your name and email are read from Git on your computer.

Check in a terminal that all containers are up. You should get seven lines, all healthy.

docker ps --filter "label=com.docker.compose.project=robo-architect-desktop" --format "{{.Names}}  {{.Status}}"

robo-architect-desktop-gateway-1     Up 17 seconds (healthy)
robo-architect-desktop-analyzer-1    Up 28 seconds (healthy)
robo-architect-desktop-catalog-1     Up 34 seconds (healthy)
robo-architect-desktop-fabric-1      Up 39 seconds (healthy)
robo-architect-desktop-graph-bolt-1  Up 45 seconds (healthy)
robo-architect-desktop-graph-db-1    Up 50 seconds (healthy)
robo-architect-desktop-parser-1      Up 50 seconds (healthy)

Hands-on 3 — Create your first project#

A Robo Architect project is one graph inside the graph DB. There is one graph DB, but it can hold many graphs, so each new project gets its own graph and they never mix. At first there is one project, the default robo, created at install time.

1. In the New project field on Home, type hanbit_delivery_en and click ＋ Create. Names must start with a lowercase letter and use only lowercase letters, digits and underscores. “Created project hanbit_delivery_en.” appears and that project is selected.
2. Under Code folder, click Choose folder… and pick the folder where the implementation code will go. This tutorial uses a Git repository folder named hanbit-delivery-en. In Chapter 6 Claude Code writes code into this folder, and each proposal uses a Git worktree of this folder as its sandbox.
3. Click Open project. The app points the backend at this project's graph (a few seconds) and shows the open project at the top of Home. The feature cards are enabled.

The next time you start the app, it reopens the project you opened last. To move to another project, click Switch project on Home.

What to check#

• All seven containers are healthy.
• The hanbit_delivery_en project and its code folder are shown at the top of Home.
• The status at bottom left reads “Ontological connected — graph hanbit_delivery_en · 7 Docker containers · 127.0.0.1” and “Agent Claude Code”.

Quiz#

1. In Lite, what is a “project” stored as?
2. Do proposal analysis, planning and implementation need an OpenAI key?