📄 工具详情 ⚙️ 安装教程 📚 使用教程

能力标签

🛠

AI工具

PromptOS

基于 TypeScript · 开源免费，本地部署，数据完全自主可控

⭐ 16 Stars 💻 TypeScript 📄 未公布协议 🏷 AI 8.0分

8.0AI 综合评分

aigeneratellmprompttypescript

🌐 访问官网 🔍 查看原项目

✦ AI Skill Hub 推荐

经 AI Skill Hub 精选评估，PromptOS 获评「强烈推荐」。这款AI工具在功能完整性、社区活跃度和易用性方面表现出色，AI 评分 8.0 分，适合有一定技术背景的用户使用。

📚 深度解析

PromptOS 是一款基于 TypeScript 的开源工具，在 GitHub 上收获 0k+ Star，是ai、generate、llm、prompt领域中的优质开源项目。开源工具的最大优势在于代码完全透明，你可以审计每一行代码的安全性，也可以根据自身需求进行二次开发和定制。

**为什么要使用开源工具而非商业 SaaS？**
对于个人开发者和有隐私需求的用户，本地部署的开源工具意味着数据不离本机，不受第三方服务商的数据政策约束。同时，开源工具通常没有使用次数限制和月度费用，一次安装即可长期使用，对于高频使用场景的总拥有成本（TCO）远低于订阅制商业工具。

**安装与环境准备**
PromptOS 依赖 TypeScript 运行环境。建议通过 pyenv（Python）或 nvm（Node.js）管理 TypeScript 版本，避免全局环境污染。对于新手用户，推荐先创建虚拟环境（python -m venv venv && source venv/bin/activate），再安装依赖，这样即使出现问题也可以随时删除虚拟环境重新开始，不影响系统稳定性。

**社区与维护**
GitHub Issue 和 Discussion 是获取帮助的最快渠道。在提问前建议先检查 Closed Issues（已关闭的问题），大多数常见问题都已有解答。遇到 Bug 时，提供 pip list 的输出、完整错误堆栈和最小可复现示例，能显著提高开发者响应速度。AI Skill Hub 将持续追踪 PromptOS 的版本更新，及时通知重要功能变化。

📋 工具概览

PromptOS 是一款基于 TypeScript 开发的开源工具，专注于 ai、generate、llm 等核心功能。作为 GitHub 开源项目，它拥有活跃的社区支持和持续的版本迭代，代码完全透明可审计，支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流，都能提供稳定可靠的解决方案。

GitHub Stars

⭐ 16

开发语言

TypeScript

支持平台

Windows / macOS / Linux

维护状态

轻量级项目，按需更新

开源协议

未公布

AI 综合评分

8.0 分

工具类型

AI工具

Forks

—

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

📌 核心特色

开源免费，支持本地部署，数据完全自主可控
活跃的 GitHub 开源社区，持续迭代更新
提供详细文档和使用示例，新手友好
支持自定义配置，灵活适配不同使用环境
可作为基础组件集成进现有技术栈或进行二次开发

🎯 主要使用场景

本地部署运行，保护数据隐私，满足合规要求
自定义集成到现有系统，扩展技术栈能力
作为开源基础组件进行商业化二次开发

以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。

安装命令

# 方式一：npm 全局安装
npm install -g promptos

# 方式二：npx 直接运行（无需安装）
npx promptos --help

# 方式三：项目依赖安装
npm install promptos

# 方式四：从源码运行
git clone https://github.com/staypossesed/PromptOS
cd PromptOS
npm install
npm start

📋 安装步骤说明

访问 GitHub 仓库页面
按照 README 文档完成依赖安装
根据系统环境完成初始化配置
参考官方示例或文档开始使用
遇到问题可在 GitHub Issues 中查找解答

以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。

常用命令 / 代码示例

# 命令行使用
promptos --help

# 基本用法
promptos [options] <input>

# Node.js 代码中使用
const promptos = require('promptos');

const result = await promptos.run(options);
console.log(result);

以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。

配置示例

# promptos 配置说明
# 查看配置选项
promptos --config-example > config.yml

# 常见配置项
# output_dir: ./output
# log_level: info
# workers: 4

# 环境变量（覆盖配置文件）
export PROMPTOS_CONFIG="/path/to/config.yml"

📑 README 深度解析真实文档完整度 80/100 查看 GitHub 原文 →

以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

Umprompt

Turn rough ideas into execution-ready AI prompts — scored, optimized, and saved.

Umprompt is a prompt engineering workspace for developers, automation builders, and AI power users. Describe what you want in plain language, pick your target AI tool, and get a structured prompt that actually performs — scored across six quality dimensions, one-click optimized, and saved to your account.

---

Features

Feature	What it does
Generate	Describe your goal in plain English. Umprompt applies tool-specific profiles (Cursor, Claude, ChatGPT) to produce a structured, execution-ready prompt.
Score	Every prompt is scored 0–100 across Clarity, Context, Constraints, Examples, Output Format, and Tool Fit. Each dimension includes one actionable improvement tip.
Optimize	Click "Optimize weak dimensions" to rewrite the prompt targeting every low-scoring dimension automatically. Confirms improvement with a before → after score.
Save & Reopen	Prompts are saved to your account with full score data. Reopen any prompt from History to refine and update it.
Context Panel	Optionally supply project type, audience, constraints, output format, and examples — fed directly into generation and restored when you reopen a saved prompt.

---

Prerequisites

Node.js 20+
A Supabase project (free tier is fine)
An Anthropic API key

Local Setup

1. Clone and install

git clone <your-repo-url>
cd promptos
npm install

Setup

Get an OpenRouter API key.
Add to .env.local (and Vercel env vars):

OPENROUTER_API_KEY=sk-or-v1-...
KIMI_MODEL=moonshotai/kimi-k2.6

Navigate to /model-lab while signed in.

Deployment (Vercel)

Supabase URL Configuration (after first deploy)

Go to Supabase → Authentication → URL Configuration and set:

Field	Value
Site URL	`https://<your-vercel-domain>`
Redirect URLs	`https://<your-vercel-domain>/auth/callback`

Keep http://localhost:3000/auth/callback in Redirect URLs for local development.

---

Pre-launch demo checklist

[ ] Visit /demo — all 5 steps render correctly, before/after scores show
[ ] Landing page loads at root — hero CTA reads "Try Umprompt free"
[ ] /builder opens without auth redirect
[ ] Generate → score → optimize → save flow completes end-to-end
[ ] History shows saved prompt; clicking reopens in builder
[ ] Feedback button appears in sidebar; modal opens and saves to Supabase
[ ] PostHog Live Events shows events firing during the above flow
[ ] /privacy, /terms, /help, /demo all load without auth
[ ] Mobile layout works on 390px viewport
[ ] No console.error during any of the above

2. Configure environment

cp .env.example .env.local

Open .env.local and fill in all values (see table below). The file is gitignored — never commit it.

Environment Variables

Variable	Required	Description
`NEXT_PUBLIC_SUPABASE_URL`	Yes	Supabase project URL (safe for browser)
`NEXT_PUBLIC_SUPABASE_ANON_KEY`	Yes	Supabase anon/public key (safe for browser)
`NEXT_PUBLIC_SITE_URL`	Yes	App base URL — used for magic link redirects
`ANTHROPIC_API_KEY`	Yes	Anthropic API key — server-only, never expose to browser
`DEFAULT_AI_PROVIDER`	No	`anthropic` or `openrouter` (default: `anthropic`)
`DEFAULT_AI_MODEL`	No	Registered model ID (default: `claude-sonnet-4-6`)
`SCORE_AI_MODEL`	No	Model used for scoring + optimization (default: `claude-sonnet-4-6`)
`OPENROUTER_API_KEY`	No	Required only when `DEFAULT_AI_PROVIDER=openrouter`

Copy .env.example for the full template with comments.

---

Vercel environment variables

Paste these into Project Settings → Environment Variables:

Variable	Value
`NEXT_PUBLIC_SUPABASE_URL`	Your Supabase project URL
`NEXT_PUBLIC_SUPABASE_ANON_KEY`	Your Supabase anon/public key
`NEXT_PUBLIC_SITE_URL`	Your Vercel domain, e.g. `https://umprompt.vercel.app`
`ANTHROPIC_API_KEY`	Your Anthropic API key (server-only — never use `NEXT_PUBLIC_`)
`DEFAULT_AI_PROVIDER`	`anthropic`
`DEFAULT_AI_MODEL`	`claude-sonnet-4-6`
`SCORE_AI_MODEL`	`claude-sonnet-4-6`
`OPENROUTER_API_KEY`	OpenRouter key for Kimi (optional — only needed for Model Lab)
`KIMI_MODEL`	`moonshotai/kimi-k2.6` (informational — actual ID is hardcoded in registry)

Important: ANTHROPIC_API_KEY must NOT have the NEXT_PUBLIC_ prefix — it is server-only and must never be exposed to the browser.

New env vars for production hardening

Add these to Vercel Project Settings → Environment Variables:

Variable	Required	Description
`NEXT_PUBLIC_POSTHOG_KEY`	Optional	PostHog project API key — enables analytics
`NEXT_PUBLIC_POSTHOG_HOST`	Optional	PostHog ingest host (default: `https://us.i.posthog.com`)
`UPSTASH_REDIS_REST_URL`	Optional	Upstash Redis URL — enables persistent rate limiting
`UPSTASH_REDIS_REST_TOKEN`	Optional	Upstash Redis token

---

API Routes

Endpoint	Method	Auth	Description
`/api/prompts`	GET	Required	List the signed-in user's prompts
`/api/prompts`	POST	Required	Save a new prompt
`/api/prompts/[id]`	GET	Required	Fetch a single prompt
`/api/prompts/[id]`	PATCH	Required	Update title, prompt, score, or context
`/api/prompts/[id]`	DELETE	Required	Delete a prompt
`/api/prompts/generate`	POST	Required	Stream an AI-generated prompt (text/plain)
`/api/prompts/score`	POST	Required	Score a prompt across 6 dimensions
`/api/prompts/optimize`	POST	Required	Rewrite prompt targeting weak dimensions
`/api/model-lab/compare`	POST	Required	Compare N models on the same idea (10/day limit)
`/api/model-lab/comparisons`	GET	Required	List last 10 saved comparisons (metadata only)
`/api/model-lab/comparisons`	POST	Required	Save a comparison result to the dataset
`/api/model-lab/comparisons/[id]`	PATCH	Required	Set winner and reason on a saved comparison

---

Supported Models

Registered in lib/ai/providers.ts. Override via env vars:

Model ID	Provider	Notes
`claude-sonnet-4-6`	Anthropic	Default — best quality/cost balance
`claude-opus-4-7`	Anthropic	Highest quality, higher cost
`claude-haiku-4-5`	Anthropic	Fastest, lowest cost
`moonshotai/kimi-k2.6`	OpenRouter	Requires `OPENROUTER_API_KEY`

To add a new model: add an entry to MODEL_REGISTRY in lib/ai/providers.ts.

---

Running comparisons and building the evaluation dataset

Navigate to /model-lab.
Enter an idea and target tool.
Select Claude Sonnet and Kimi K2.6 (or any combination of 1–4 models).
Click Generate comparison — all models run in parallel. Each comparison is automatically saved to your account.
Review the outputs side by side: quality score (0–100), latency, and estimated cost are shown per result.
Click Mark as winner on the result card you would actually ship, then pick the reason.