Tutorial: your first session with mini-me

By the end of this tutorial, you will have built the mm binary, started it against Docker Model Runner, asked it one question that makes it run a shell command, and left cleanly. No prior knowledge of the project is required.

Prerequisites

Go 1.25 or later on your PATH (go version answers).
Docker Desktop with Docker Model Runner enabled and reachable on http://localhost:12434.
The model named in agent.yaml pulled locally. The file shipped with the project names huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M; pull it with docker model pull <that name>.
A terminal open at the root of the repository (the directory that holds main.go and agent.yaml).

Step 1 — Build the binary

Type:

go build -o mm .

You should see no output at all. A file named mm now sits in the current directory.

We have just compiled the whole agent into a single executable.

Step 2 — Start the agent

Type:

./mm

You should see a banner like this one (the model name is the one from agent.yaml):

config: agent.yaml | provider: dmr | model: huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M | ctx: unknown | tools: bash, read_file, write_file, edit_file | skills: 0
Agent ready. Commands: "/quit" to exit, "/abort" to stop current generation, "/compact" to compress the history, "/new" to start a new session.
Shortcut: Ctrl+C to abort generation, or Ctrl+C at prompt to quit.

>

If a dimmed [warning: dmr: nothing answers at …] line appears above the banner, Docker Model Runner is not running: start it, then run ./mm again. The agent starts anyway, but the next step would fail.

We have just loaded the configuration, connected to the model server, and declared the tools the model may call.

Step 3 — Ask a question that needs a command

At the > prompt, type:

list the files in this directory

You should see, in order: a spinner while the model thinks, a line starting with 🛠️ bash: showing the command the model chose (typically ls or ls -la), the first lines of that command's output indented and dimmed, then the model's answer in plain text. The turn ends with a dimmed recap such as:

⚙ 1 command(s)

followed by the numbered list of the commands that ran, in green.

We have just watched the agent loop once: the model asked for a tool, the tool ran, its output went back to the model, and the model answered.

Step 4 — Leave

At the > prompt, type:

/quit

The program exits. Pressing Ctrl+C at the prompt does the same and prints Goodbye. first.

What now?

You have built mm, run one tool-calling turn, and exited. To go further:

To run it on llama.cpp, host it in an editor, or add skills → see the how-to guides
To look up a flag, a YAML key or a slash command → see the reference
To understand how the loop and the two front ends fit together → see the explanation

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80


# Tutorial: your first session with mini-me

By the end of this tutorial, you will have built the `mm` binary, started it against Docker Model Runner, asked it one question that makes it run a shell command, and left cleanly. No prior knowledge of the project is required.

## Prerequisites

- Go 1.25 or later on your `PATH` (`go version` answers).
- Docker Desktop with Docker Model Runner enabled and reachable on `http://localhost:12434`.
- The model named in `agent.yaml` pulled locally. The file shipped with the project names `huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M`; pull it with `docker model pull <that name>`.
- A terminal open at the root of the repository (the directory that holds `main.go` and `agent.yaml`).

## Step 1 — Build the binary

Type:

```bash
go build -o mm .
```

You should see no output at all. A file named `mm` now sits in the current directory.

We have just compiled the whole agent into a single executable.

## Step 2 — Start the agent

Type:

```bash
./mm
```

You should see a banner like this one (the model name is the one from `agent.yaml`):

```
config: agent.yaml | provider: dmr | model: huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M | ctx: unknown | tools: bash, read_file, write_file, edit_file | skills: 0
Agent ready. Commands: "/quit" to exit, "/abort" to stop current generation, "/compact" to compress the history, "/new" to start a new session.
Shortcut: Ctrl+C to abort generation, or Ctrl+C at prompt to quit.

>
```

If a dimmed `[warning: dmr: nothing answers at …]` line appears above the banner, Docker Model Runner is not running: start it, then run `./mm` again. The agent starts anyway, but the next step would fail.

We have just loaded the configuration, connected to the model server, and declared the tools the model may call.

## Step 3 — Ask a question that needs a command

At the `>` prompt, type:

```
list the files in this directory
```

You should see, in order: a spinner while the model thinks, a line starting with `🛠️  bash:` showing the command the model chose (typically `ls` or `ls -la`), the first lines of that command's output indented and dimmed, then the model's answer in plain text. The turn ends with a dimmed recap such as:

```
⚙ 1 command(s)
```

followed by the numbered list of the commands that ran, in green.

We have just watched the agent loop once: the model asked for a tool, the tool ran, its output went back to the model, and the model answered.

## Step 4 — Leave

At the `>` prompt, type:

```
/quit
```

The program exits. Pressing `Ctrl+C` at the prompt does the same and prints `Goodbye.` first.

## What now?

You have built `mm`, run one tool-calling turn, and exited. To go further:

- To run it on llama.cpp, host it in an editor, or add skills → see the [how-to guides](../how-to/)
- To look up a flag, a YAML key or a slash command → see the [reference](../reference/)
- To understand how the loop and the two front ends fit together → see the [explanation](../explanation/)