@graypark/ralph-codex 0.8.0 → 0.8.1
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 +21 -0
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -8,14 +8,35 @@
|
|
|
8
8
|
<a href="https://github.com/vcz-Gray/ralph-codex/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg?style=flat-square" alt="license" /></a>
|
|
9
9
|
<img src="https://img.shields.io/badge/node-%3E%3D18-brightgreen.svg?style=flat-square" alt="node version" />
|
|
10
10
|
<img src="https://img.shields.io/badge/platform-Codex%20CLI%20%7C%20Claude%20Code-purple.svg?style=flat-square" alt="platform" />
|
|
11
|
+
<img src="https://img.shields.io/badge/tests-32%20passing-brightgreen.svg?style=flat-square" alt="tests" />
|
|
11
12
|
</p>
|
|
12
13
|
|
|
13
14
|
<p align="center">
|
|
14
15
|
<b>Iterative AI development loops with PRD-driven task tracking, multi-agent orchestration, and interactive command generation.</b>
|
|
16
|
+
<br/>
|
|
17
|
+
<sub>Based on <a href="https://ghuntley.com/ralph/">Geoffrey Huntley's Ralph Wiggum technique</a></sub>
|
|
15
18
|
</p>
|
|
16
19
|
|
|
17
20
|
---
|
|
18
21
|
|
|
22
|
+
## Why ralph-codex?
|
|
23
|
+
|
|
24
|
+
AI coding agents struggle with:
|
|
25
|
+
|
|
26
|
+
| Problem | What happens |
|
|
27
|
+
| ------------------------ | ---------------------------------------------------------- |
|
|
28
|
+
| **Context rot** | Long conversations accumulate noise, agent gets confused |
|
|
29
|
+
| **No checkpoints** | All-or-nothing execution — can't resume after interruption |
|
|
30
|
+
| **Lost learnings** | Previous iterations' insights overwritten by new context |
|
|
31
|
+
| **Completion ambiguity** | Agent says "done" but tests still fail |
|
|
32
|
+
|
|
33
|
+
ralph-codex solves this:
|
|
34
|
+
|
|
35
|
+
- **Fresh context per iteration** — Each cycle reads PRD + progress from disk, no degradation
|
|
36
|
+
- **Git-enforced safety** — Atomic commits per story, rollback at any point
|
|
37
|
+
- **Append-only learnings** — `progress.txt` accumulates knowledge across iterations
|
|
38
|
+
- **Test-verified completion** — Agent can only exit when `<promise>COMPLETE</promise>` is genuinely true
|
|
39
|
+
|
|
19
40
|
## What is Ralph Loop?
|
|
20
41
|
|
|
21
42
|
An AI agent works on a task in a continuous loop. Each iteration starts with **fresh context** — reading the PRD and progress files to decide what to do next. The agent implements one story, commits, updates progress, and exits. The stop hook intercepts the exit and re-injects the prompt. Repeat until all stories pass.
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@graypark/ralph-codex",
|
|
3
|
-
"version": "0.8.
|
|
3
|
+
"version": "0.8.1",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "Ralph Loop for Codex CLI & Claude Code — iterative dev loops with multi-agent orchestration, interactive interview, and stop hooks",
|
|
6
6
|
"license": "MIT",
|