# ReadmeAI — Generate README from a Repo URL

> ReadmeAI generates a README from a local repo path or GitHub URL, with offline mode plus OpenAI/Anthropic/Gemini/Ollama backends for LLM polish.

## Install

Save the content below to `.claude/skills/` or append to your `CLAUDE.md`:

## Quick Use

1. Install (Python):
   ```bash
   pip install -U readmeai
   ```
2. Generate a README with no API keys (offline mode):
   ```bash
   readmeai --api offline -o README.generated.md -r https://github.com/eli64s/readme-ai
   ```
3. (Optional) Use OpenAI:
   ```bash
   export OPENAI_API_KEY=<your_api_key>
   readmeai --api openai -o README.openai.md -r https://github.com/eli64s/readme-ai
   ```

## Intro

ReadmeAI generates a README from a local repo path or GitHub URL, with offline mode plus OpenAI/Anthropic/Gemini/Ollama backends for LLM polish.

- **Best for:** maintainers who need consistent READMEs across many repos, and teams onboarding new contributors
- **Works with:** Python 3.9+, local repos or GitHub URLs, optional LLM API keys (OpenAI/Anthropic/Gemini) or Ollama
- **Setup time:** 5–10 minutes

## Practical Notes

- README states Python 3.9+ requirement and supports pip/pipx/uv tool installs.
- Usage examples include offline mode and OpenAI/Anthropic/Gemini/Ollama backends.

## How to Use It Without “Doc Drift”

If your README is generated once and never revisited, it quickly diverges from reality. A practical loop:

1. Keep a `.readmeaiignore` (ReadmeAI supports ignore patterns) so vendor folders don’t pollute your docs.
2. Re-run on release candidates (or weekly) and commit diffs like any other docs change.
3. Start with offline mode to get structure, then switch to an LLM backend only when you want tone/wording improvements.

## What to Check in the Output

Before you ship the generated README:

- Does it include the *minimum runnable command* for your project?
- Are environment variables and ports correct?
- Are “Installation” steps actually reproducible on a clean machine?


### FAQ

**Q: Do I need an API key?**
A: No. The README documents an offline mode; add API keys only for LLM-backed generation.

**Q: What inputs are supported?**
A: The CLI accepts a local repository path or a GitHub URL as the repository source.

**Q: How do I isolate installs?**
A: Use pipx or uv tool install so readmeai runs in an isolated environment.

## Source & Thanks

> Source: https://github.com/eli64s/readme-ai
> License: MIT
> GitHub stars: 2,898 · forks: 283

---

<!-- ZH -->

## 快速使用

1. 安装（Python）：
   ```bash
   pip install -U readmeai
   ```
2. 不用 API key 也能生成（离线模式）：
   ```bash
   readmeai --api offline -o README.generated.md -r https://github.com/eli64s/readme-ai
   ```
3.（可选）切到 OpenAI：
   ```bash
   export OPENAI_API_KEY=<your_api_key>
   readmeai --api openai -o README.openai.md -r https://github.com/eli64s/readme-ai
   ```

## 简介

ReadmeAI 可以从本地目录或 GitHub URL 自动生成结构化 README：支持离线模式，也可切换 OpenAI/Anthropic/Gemini/Ollama 等后端做更“像人写的”文档。

- **适合谁：** 需要批量统一 README 风格的维护者，以及想降低新人上手成本的团队
- **可搭配：** Python 3.9+、本地仓库或 GitHub URL、可选的 LLM API key（OpenAI/Anthropic/Gemini）或 Ollama
- **准备时间：** 5–10 分钟

## 实战建议

- README 标注要求 Python 3.9+，并提供 pip / pipx / uv tool 等安装方式。
- 示例覆盖离线模式，以及 OpenAI/Anthropic/Gemini/Ollama 多后端切换。

## 如何避免“文档漂移”

README 只生成一次、不再更新，很快就会与现实不符。更实用的做法：

1. 保持一份 `.readmeaiignore`（ReadmeAI 支持忽略规则），避免把 vendor 目录写进文档。
2. 在发布前（或每周）重跑一次，把 README diff 当作常规文档更新来审。
3. 先用离线模式拿到结构，再切换到 LLM 后端做措辞与表达优化。

## 输出结果的检查清单

发布前至少确认：

- 是否包含项目的“最小可运行命令”？
- 环境变量与端口是否准确？
- 安装步骤是否能在干净环境复现？


### FAQ

**一定要 API key 吗？**
答：不需要。README 提到离线模式；只有想用 LLM 后端润色时才需要 key。

**支持哪些输入？**
答：CLI 支持本地仓库路径与 GitHub URL 作为输入源。

**怎么做到隔离安装？**
答：可以用 pipx 或 uv tool install，把 readmeai 安装到隔离环境中。

## 来源与感谢

> Source: https://github.com/eli64s/readme-ai
> License: MIT
> GitHub stars: 2,898 · forks: 283


---
Source: https://tokrepo.com/en/workflows/readmeai-generate-readme-from-a-repo-url
Author: Script Depot