@neriros/ralphy 2.3.0 → 2.4.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +25 -51
- package/dist/cli/index.js +3406 -6323
- package/dist/mcp/index.js +10880 -9936
- package/package.json +4 -8
- package/dist/checklists/deploy.md +0 -5
- package/dist/checklists/static.md +0 -6
- package/dist/checklists/tests.md +0 -5
- package/dist/phases/done.md +0 -10
- package/dist/phases/exec.md +0 -140
- package/dist/phases/plan.md +0 -119
- package/dist/phases/research.md +0 -99
- package/dist/phases/review.md +0 -115
- package/dist/scaffolds/MANUAL_TESTING.md +0 -9
- package/dist/scaffolds/PLAN.md +0 -20
- package/dist/scaffolds/PROGRESS.md +0 -13
- package/dist/scaffolds/STEERING.md +0 -9
package/README.md
CHANGED
|
@@ -4,15 +4,15 @@ An iterative AI task execution framework. Ralphy orchestrates multi-phase autono
|
|
|
4
4
|
|
|
5
5
|
## How It Works
|
|
6
6
|
|
|
7
|
-
Ralphy
|
|
7
|
+
Ralphy runs a single continuous loop against an OpenSpec change — no phases, no phase transitions.
|
|
8
8
|
|
|
9
9
|
```mermaid
|
|
10
10
|
graph LR
|
|
11
|
-
R[
|
|
12
|
-
|
|
11
|
+
S[Start iteration] --> R[Read Steering] --> T[Find first unchecked task] --> W[Do the work] --> V[Validate] --> C[Check off task] --> S
|
|
12
|
+
T -->|all tasks checked| D[Archive change]
|
|
13
13
|
```
|
|
14
14
|
|
|
15
|
-
Each
|
|
15
|
+
Each iteration reads the `## Steering` section of `proposal.md`, picks the first unchecked item from `tasks.md`, does the work, validates, and checks the item off. When all items are checked the loop archives the change automatically.
|
|
16
16
|
|
|
17
17
|
## Installation
|
|
18
18
|
|
|
@@ -53,15 +53,7 @@ ralph task --name fix-auth --prompt "Fix the JWT validation bug" --claude opus -
|
|
|
53
53
|
|
|
54
54
|
The engine defaults to Claude Opus.
|
|
55
55
|
|
|
56
|
-
###
|
|
57
|
-
|
|
58
|
-
```bash
|
|
59
|
-
ralph task --name fix-auth --prompt "Fix the JWT validation bug" --interactive
|
|
60
|
-
```
|
|
61
|
-
|
|
62
|
-
Runs the research and plan phases interactively (with direct terminal I/O), then switches to automated execution for the remaining phases. Useful when you want to guide early discovery and let the agent execute autonomously.
|
|
63
|
-
|
|
64
|
-
### Resume a Task
|
|
56
|
+
### Resume a Change
|
|
65
57
|
|
|
66
58
|
```bash
|
|
67
59
|
ralph task --name fix-auth
|
|
@@ -76,13 +68,6 @@ ralph list # Table of all tasks
|
|
|
76
68
|
ralph status --name fix-auth # Detailed view of one task
|
|
77
69
|
```
|
|
78
70
|
|
|
79
|
-
### Manual Phase Control
|
|
80
|
-
|
|
81
|
-
```bash
|
|
82
|
-
ralph advance --name fix-auth # Advance to next phase
|
|
83
|
-
ralph set-phase --name fix-auth --phase exec # Jump to a specific phase
|
|
84
|
-
```
|
|
85
|
-
|
|
86
71
|
## CLI Options
|
|
87
72
|
|
|
88
73
|
| Option | Description |
|
|
@@ -93,8 +78,6 @@ ralph set-phase --name fix-auth --phase exec # Jump to a specific phase
|
|
|
93
78
|
| `--claude [model]` | Use Claude engine (haiku/sonnet/opus) |
|
|
94
79
|
| `--codex` | Use Codex engine |
|
|
95
80
|
| `--model <model>` | Set model (haiku/sonnet/opus) |
|
|
96
|
-
| `--no-execute` | Stop after research + plan phases |
|
|
97
|
-
| `--interactive` | Run research + plan interactively, then automate |
|
|
98
81
|
| `--max-iterations <N>` | Stop after N iterations (0 = unlimited) |
|
|
99
82
|
| `--max-cost <N>` | Stop when cost exceeds $N |
|
|
100
83
|
| `--max-runtime <N>` | Stop after N minutes |
|
|
@@ -104,42 +87,32 @@ ralph set-phase --name fix-auth --phase exec # Jump to a specific phase
|
|
|
104
87
|
| `--log` | Log raw JSON stream output |
|
|
105
88
|
| `--verbose` | Verbose output |
|
|
106
89
|
|
|
107
|
-
##
|
|
90
|
+
## OpenSpec Flow
|
|
108
91
|
|
|
109
|
-
|
|
110
|
-
| ------------ | -------------------------------------------- | ------------------------ |
|
|
111
|
-
| **Research** | Study the codebase and gather context | `RESEARCH.md` |
|
|
112
|
-
| **Plan** | Design the implementation approach | `PLAN.md`, `PROGRESS.md` |
|
|
113
|
-
| **Exec** | Implement items from the progress checklist | Code changes |
|
|
114
|
-
| **Review** | Verify the work, loop back to exec if needed | Updated `PROGRESS.md` |
|
|
115
|
-
| **Done** | Terminal phase — task is complete | — |
|
|
92
|
+
There are no phases. One loop, one prompt, one `tasks.md` checklist.
|
|
116
93
|
|
|
117
|
-
|
|
94
|
+
Each change lives in `.ralph/tasks/<name>/`:
|
|
118
95
|
|
|
119
|
-
|
|
96
|
+
| File / Directory | Purpose |
|
|
97
|
+
| ------------------- | --------------------------------------------------------- |
|
|
98
|
+
| `proposal.md` | Description, goals, and the `## Steering` section |
|
|
99
|
+
| `design.md` | Technical design and architecture decisions |
|
|
100
|
+
| `tasks.md` | Checklist driving iteration — one unchecked item per loop |
|
|
101
|
+
| `specs/` | Detailed specifications for individual tasks |
|
|
102
|
+
| `.ralph-state.json` | Loop state (iteration count, status, cost, history) |
|
|
103
|
+
| `STOP` | Create this file to signal the loop to stop |
|
|
120
104
|
|
|
121
|
-
|
|
122
|
-
| ---------------- | ------------------------------------------------ |
|
|
123
|
-
| `state.json` | Task state, usage stats, and history |
|
|
124
|
-
| `STEERING.md` | Your guidance to the agent (editable anytime) |
|
|
125
|
-
| `RESEARCH.md` | Agent's research findings |
|
|
126
|
-
| `PLAN.md` | Agent's implementation plan |
|
|
127
|
-
| `PROGRESS.md` | Checklist tracking execution progress |
|
|
128
|
-
| `INTERACTIVE.md` | Context saved from interactive session (if used) |
|
|
129
|
-
| `STOP` | Create this file to signal the loop to stop |
|
|
105
|
+
Steering is delivered by editing the `## Steering` section of `proposal.md`. The agent reads it at the start of every iteration.
|
|
130
106
|
|
|
131
107
|
## MCP Server
|
|
132
108
|
|
|
133
109
|
Ralphy includes an MCP server that exposes task management tools to Claude agents. It's automatically configured during installation. Available tools:
|
|
134
110
|
|
|
135
|
-
- `
|
|
136
|
-
- `
|
|
137
|
-
- `
|
|
138
|
-
- `
|
|
139
|
-
- `
|
|
140
|
-
- `ralph_update_steering` — Update STEERING.md
|
|
141
|
-
- `ralph_finish_interactive` — Complete interactive session and hand off to automated phases
|
|
142
|
-
- `ralph_list_checklists` / `ralph_apply_checklist` — Manage verification checklists
|
|
111
|
+
- `ralph_list_changes` — List changes with status
|
|
112
|
+
- `ralph_get_change` — Get change details
|
|
113
|
+
- `ralph_create_change` — Create and optionally start a change
|
|
114
|
+
- `ralph_append_steering` — Append a steering message to `proposal.md`
|
|
115
|
+
- `ralph_stop` — Stop a running change
|
|
143
116
|
|
|
144
117
|
## Project Structure
|
|
145
118
|
|
|
@@ -149,11 +122,12 @@ ralphy/
|
|
|
149
122
|
│ ├── cli/ # CLI application
|
|
150
123
|
│ └── mcp/ # MCP server
|
|
151
124
|
├── packages/
|
|
152
|
-
│ ├── core/ # State management
|
|
125
|
+
│ ├── core/ # State management and loop
|
|
153
126
|
│ ├── context/ # Storage abstraction
|
|
127
|
+
│ ├── content/ # Base prompt and task templates
|
|
154
128
|
│ ├── engine/ # Claude/Codex engine spawning
|
|
129
|
+
│ ├── openspec/ # ChangeStore interface and OpenSpec adapter
|
|
155
130
|
│ ├── output/ # Terminal formatting
|
|
156
|
-
│ ├── phases/ # Phase definitions and checklists
|
|
157
131
|
│ └── types/ # Zod schemas and types
|
|
158
132
|
└── Makefile
|
|
159
133
|
```
|