First steps
A guided walkthrough for someone new to Soliplex: install the skills, generate a
stack, bring it up, and exercise the demo tool in the custom room.
Prerequisites
- A coding agent that supports filesystem Agent Skills (e.g. Claude Code).
- On the machine the agent drives: Docker (with the Compose plugin),
uv, and an Ollama server reachable from the containers that serves the models the stack will reference.
1. Install the skills
Install two skills into your agent. Both are listed at
https://agentskills.io/; each is also published as a rolling latest release
you can install from directly:
soliplex-template— generates and configures a stack (this template).soliplex-docs— Soliplex's documentation, so the agent can answer setup and usage questions as you go.
2. Generate a stack
Ask the agent to generate one — e.g. "generate a new Soliplex stack". The
soliplex-template skill prompts for what it needs (at minimum your Ollama URL),
scaffolds a fresh project directory, and generates its secrets.
(To drive the generator by hand instead, see Generate a project.)
3. Bring it up
From the generated project directory, prepare the databases, then start the stack:
The first run builds the images and initializes Postgres, so it takes a few minutes. Once it settles, the web frontend is at http://localhost:9000. The stack starts with no OIDC provider, so authentication is off and there is no login step.
4. Visit the custom room
Open http://localhost:9000 and select the Custom Tool Demo room. Every
generated stack ships it to demonstrate a tool provided by the project's own
Python package (src/<package>/src/<package>/tools.py).
5. Exercise the greeting tool
Ask the room to greet someone:
Please greet Ada.
The room's agent calls the package's greeting tool and replies with something
like:
Hello, Ada! This greeting came from your own package's tool.
That round-trip — your prompt, the agent calling a tool defined in your package, and the result coming back — shows the stack already running your own code, not just stock Soliplex.
Where next
The next tutorial, Populate RAG, feeds a document set into the stack and asks questions grounded in it.