🔧 PromptForge

AI Prompt Engineering Toolkit

English | 中文

---

English

Inspired by GitHub trending AI projects: prompts.chat (151k⭐), caveman (83.8k⭐), system-prompts-and-models-of-ai-tools (130k⭐)

Introduction

PromptForge is an all-in-one AI prompt engineering toolkit that helps you optimize, test, and manage prompts.

Key differentiators from existing projects:

• prompts.chat only provides static template library → PromptForge adds AI-driven token optimization engine
• caveman only does token compression → PromptForge offers three optimization strategies (minimal/balanced/aggressive)
• system-prompts only collects and displays → PromptForge supports multi-model A/B testing and effectiveness scoring

Features

1. Prompt Management

• Create, edit, delete, search prompts
• Version control, track every modification
• Category and tag system
• Automatic token counting

2. Token Optimization Engine

Inspired by caveman (83.8k⭐) and ponytail (73.8k⭐).

Three optimization strategies:

Strategy	Effect	Description
minimal	~15% compression	Remove redundant phrases and filler words
balanced	~30% compression	Compress instruction format + normalize whitespace
aggressive	~45% compression	Merge short sentences + remove articles

3. Multi-Model Test Bench

• Support OpenAI (GPT-4o, GPT-4o-mini, etc.)
• Support Anthropic Claude (Claude Sonnet, Claude Haiku)
• Mock mode (no API key required)
• Latency, token usage statistics

4. Effectiveness Scoring

Four-dimensional scoring (0-10):

• Clarity - Sentence length and readability
• Specificity - Constraints and format specifications
• Structure - Organization and formatting
• Token Efficiency - Token usage efficiency

5. Curated Template Library

10+ curated prompt templates covering:

• Coding (code review, API design, debugging)
• Writing (blog, story)
• Productivity (meeting notes, project planning)
• AI/ML (prompt engineering, data analysis)
• System prompts (general assistant, coding assistant)

Quick Start

Docker (Recommended)

git clone https://github.com/Eileenes/promptforge.git
cd promptforge
cp .env.example .env  # Optional: configure API keys
docker-compose up -d
# Visit http://localhost:8777

Local

pip install -r requirements.txt
python main.py
# Visit http://localhost:8777

API Documentation

Visit http://localhost:8777/docs for Swagger UI after starting.

Method	Path	Description
GET	/api/prompts	List all prompts
POST	/api/prompts	Create a prompt
PUT	/api/prompts/{id}	Update a prompt
DELETE	/api/prompts/{id}	Delete a prompt
POST	/api/optimize	Optimize a prompt
POST	/api/optimize/save	Optimize and save as new version
GET	/api/test/providers	List LLM providers
POST	/api/test/run	Run a test
POST	/api/test/benchmark	Effectiveness scoring
GET	/api/library	Get template library
POST	/api/library/import	Import a template

Tech Stack

• Backend: Python 3.13 + FastAPI + SQLite
• Frontend: Native HTML/CSS/JS (no build step)
• Deployment: Docker + docker-compose

License

MIT

---

中文

灵感来源于 GitHub 热门 AI 项目：prompts.chat (151k⭐)、caveman (83.8k⭐)、system-prompts-and-models-of-ai-tools (130k⭐)

简介

PromptForge 是一个一站式的 AI 提示词工程工具，帮助你优化、测试、管理提示词。

与现有项目的差异化：

• prompts.chat 只提供静态模板库 → PromptForge 还提供 AI 驱动的 token 优化引擎
• caveman 只做 token 压缩 → PromptForge 提供三级优化策略 (minimal/balanced/aggressive)
• system-prompts 只是收集展示 → PromptForge 支持多模型 A/B 测试和效果评分

功能

1. 提示词管理

• 创建、编辑、删除、搜索提示词
• 版本管理，追踪每次修改
• 分类和标签系统
• 自动 token 计数

2. Token 优化引擎

灵感来自 caveman (83.8k⭐) 和 ponytail (73.8k⭐) 项目。

三级优化策略：

策略	效果	说明
minimal	~15% 压缩	移除冗余短语和填充词
balanced	~30% 压缩	压缩指令格式 + 规范空白
aggressive	~45% 压缩	合并短句 + 去除冠词

3. 多模型测试台

• 支持 OpenAI (GPT-4o, GPT-4o-mini 等)
• 支持 Anthropic Claude (Claude Sonnet, Claude Haiku)
• Mock 模式（无需 API key 即可体验）
• 延迟、token 用量统计

4. 效果评分系统

四个维度评分 (0-10)：

• 清晰度 - 句子长度和可读性
• 具体性 - 约束条件和格式规范
• 结构性 - 组织层次和格式化
• Token 效率 - token 使用效率

5. 精选模板库

内置 10+ 精选提示词模板，覆盖：

• 编程开发 (代码审查、API设计、调试)
• 写作创作 (博客、故事)
• 工作效率 (会议纪要、项目规划)
• AI/ML (提示词工程、数据分析)
• 系统提示词 (通用助手、编程助手)

快速开始

Docker 部署 (推荐)

git clone https://github.com/Eileenes/promptforge.git
cd promptforge
cp .env.example .env  # 可选：配置 API Key
docker-compose up -d
# 访问 http://localhost:8777

本地运行

pip install -r requirements.txt
python main.py
# 访问 http://localhost:8777

API 文档

启动后访问 http://localhost:8777/docs 查看 Swagger UI。

方法	路径	说明
GET	/api/prompts	列出所有提示词
POST	/api/prompts	创建提示词
PUT	/api/prompts/{id}	更新提示词
DELETE	/api/prompts/{id}	删除提示词
POST	/api/optimize	优化提示词
POST	/api/optimize/save	优化并保存为新版本
GET	/api/test/providers	列出 LLM 提供商
POST	/api/test/run	运行测试
POST	/api/test/benchmark	效果评分
GET	/api/library	获取模板库
POST	/api/library/import	导入模板

技术栈

• 后端: Python 3.13 + FastAPI + SQLite
• 前端: 原生 HTML/CSS/JS (无构建步骤)
• 部署: Docker + docker-compose

License

MIT