Quickstart¶

This path creates a working AgentMuru application with deterministic output. It does not use an API key or make a model request over the network.

Check the installation¶

muru doctor

Both Python and Workspace assets should report ready.

Create the starter¶

muru init my-agent --name "My first AgentMuru app"
cd my-agent

The directory contains:

app.py, which exports application.
requirements.txt, constrained to AgentMuru 0.3.
README.md, with the exact run command.

The default model is FakeModel. Open app.py and change its fixed response if you want a different local result.

Start the runtime¶

python -m pip install -r requirements.txt
muru run app:application

Open http://127.0.0.1:8000 in a browser.

Send the first message¶

Create a session in the left rail, enter a message, and submit it. The fixed assistant reply appears while the Timeline records the run, model request, streamed content, and completion.

What you should see¶

A session remains available in the session list.
The conversation contains your message and the deterministic assistant reply.
Timeline entries have a stable sequence and event type.
The run ends in completed.

Refresh the page. The in-memory starter remains useful for one process lifetime. Add SQLite persistence when sessions must survive a restart.

Stop the server¶

Return to the terminal and press Ctrl+C.