vibe-action 0.2.0

Command router — execute shell commands and LLM prompts via simple YAML actions.
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
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150

# Configuration

Vibe Action uses a single YAML config file at `~/.vibe-action/config.yaml`. It is created automatically on first run.

## Action

Runtime settings for all flows.

```yaml
action:
  system: |
    You are Vibe Action — a CLI tool, not a chatbot.
    Work fast. Don't think too much. Just do the task.
    Output ONLY the requested result.
  retries: 2
```

| Field     | Type    | Description                                      |
| --------- | ------- | ------------------------------------------------ |
| `system`  | string  | Global system prompt for all LLM requests        |
| `retries` | integer | Number of retries for failed LLM steps (0 = off) |

## Cluster

Define one or more LLM providers. The engine sends prompts to all nodes in parallel.
Use `role` to assign models to specific complexity levels.

```yaml
cluster:
  - provider: ollama
    host: http://localhost:11434
    model: qwen2.5-coder:14b-instruct
    role: medium
    timeout_secs: 60
    temperature: 0.1
    seed: 42
    num_ctx: 4096
    num_predict: 2048
    parallel: 1
```

### Cluster Node Fields

| Field          | Type    | Description                                   |
| -------------- | ------- | --------------------------------------------- |
| `provider`     | string  | `ollama`, `deepseek`, `qwen`, `kimi`, `zhipu` |
| `host`         | string  | API endpoint URL                              |
| `model`        | string  | Model name                                    |
| `role`         | string  | `tiny`, `small`, `medium`, `large`, `vision`  |
| `timeout_secs` | integer | Request timeout in seconds                    |
| `temperature`  | float   | 0.0-2.0, lower = more deterministic           |
| `seed`         | integer | Random seed for reproducibility               |
| `num_ctx`      | integer | Context window size in tokens                 |
| `num_predict`  | integer | Max tokens to generate                        |
| `api_key`      | string  | API key for cloud providers (optional)        |
| `parallel`     | integer | Concurrent connections (default: 1)           |

### Multi-Node Cluster with Roles

```yaml
cluster:
  - provider: ollama
    host: http://localhost:11434
    model: qwen2.5-coder:3b-instruct
    role: tiny
    timeout_secs: 30
    temperature: 0.0
    seed: 42
    num_ctx: 4096
    num_predict: 512
    parallel: 2

  - provider: zhipu
    host: https://open.bigmodel.cn/api/paas/v4
    model: glm-4-flash
    role: small
    timeout_secs: 60
    temperature: 0.1
    seed: 42
    num_ctx: 8192
    num_predict: 2048
    api_key: sk-...
    parallel: 2

  - provider: ollama
    host: http://192.168.1.10:11434
    model: qwen2.5-coder:14b-instruct
    role: medium
    timeout_secs: 60
    temperature: 0.1
    seed: 42
    num_ctx: 8192
    num_predict: 2048
    parallel: 1

  - provider: deepseek
    host: https://api.deepseek.com/v1
    model: deepseek-v4-flash
    role: large
    timeout_secs: 120
    temperature: 0.1
    seed: 42
    num_ctx: 16384
    num_predict: 8192
    api_key: sk-...
    parallel: 2

  - provider: kimi
    host: https://api.moonshot.cn/v1
    model: moonshot-v1-8k
    role: large
    timeout_secs: 120
    temperature: 0.1
    seed: 42
    num_ctx: 8192
    num_predict: 2048
    api_key: sk-...
    parallel: 1

  - provider: ollama
    host: http://localhost:11434
    model: qwen2.5vl:7b
    role: vision
    timeout_secs: 120
    temperature: 0.1
    seed: 42
    num_ctx: 4096
    num_predict: 2048
    parallel: 1
```

Nodes with `role: tiny` are used for `run: tiny`, `role: small` for `run: small`, `role: medium` for `run: medium`, `role: large` for `run: large`, `role: vision` for `run: vision`.
Nodes without a `role` respond to all requests.

## Actions Directory

Actions are stored in `~/.vibe-action/actions/`. The directory is created on first run with default actions. Override with `VIBE_ACTION_PATH`.

```
~/.vibe-action/
├── config.yaml
└── actions/
    ├── comment.yaml
    ├── commit.yaml
    ├── describe.yaml
    ├── explain.yaml
    ├── ...
```

Add your own `.yaml` files here — they will be loaded automatically.