🚀 prompt-autopilot

Turn AI tool output from "templates" to "professional tools"

All steps are LLM-powered · No API Key required · Chinese & English

中文说明 · English · Integration Guide · License

---

⚡ See It In Action

$ pma "帮我写排序算法"

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✨ 优化后的编程指令
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

## 🎯 任务
用 Python 实现快速排序算法

## 📥 输入
类型：整数数组，范围：1-100000，示例：[3, 6, 8, 10, 1, 2, 1]

## 📤 输出
类型：整数数组（升序），示例：[1, 1, 2, 3, 6, 8, 10]

## ⚡ 性能
时间：O(n log n) | 空间：O(log n)

## 🛡️ 边界
空数组 → [] | 单元素 → [x] | 重复 → 保持相对顺序

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Score: 9.0/10

---

📦 Quick Start

pip install prompt-autopilot

# Core commands
pma think "做个登录功能"    # Dual-perspective analysis
pma optimize "写排序算法"   # Smart optimization

---

🧠 LLM Tier System

All steps are LLM-powered. The difference is only in model selection:

Tier	Model	Response Time	Use Case
⚡ fast	gpt-3.5-turbo	1-2s	Simple instructions (< 10 words)
⚡ medium	gpt-3.5-turbo + detailed prompt	2-3s	Medium instructions
🧠 deep	gpt-4	5-10s	Complex instructions (>= 30 words)

CLI Usage

# Auto-select tier based on instruction complexity (default)
pma optimize "做个登录"

# Force fast LLM (gpt-3.5-turbo, ~1-2s)
pma optimize "做个登录" --fast

# Force deep LLM (gpt-4, ~5-10s)
pma optimize "做个登录" --deep

# Explicit tier selection
pma optimize "做个登录" --tier auto|fast|medium|deep

---

🔐 API Key Configuration (Optional)

For LLM-powered optimization, configure your API key securely:

# Method 1: Environment variables (recommended)
export PROMPT_AUTOPILOT_API_KEY="sk-..."
export PROMPT_AUTOPILOT_FAST_MODEL="gpt-3.5-turbo"
export PROMPT_AUTOPILOT_DEEP_MODEL="gpt-4"
export PROMPT_AUTOPILOT_ENDPOINT="https://api.openai.com/v1/chat/completions"

# Method 2: Local config file (not committed to git)
# File: ~/.prompt-autopilot/config.json
{
  "llm_api_key": "sk-...",
  "fast_model": "gpt-3.5-turbo",
  "deep_model": "gpt-4",
  "llm_endpoint": "https://api.openai.com/v1/chat/completions"
}

Without API key: Falls back to built-in smart templates. With API key: All steps use LLM with auto tier selection.

⚠️ Security: API keys are never committed to git. The config.json file and .env are in .gitignore.

---

⭐ Features

• 🧠 think — Dual perspective analysis (engineer + product)
• ⚡ optimize — LLM-powered with tier selection (fast/medium/deep)
• 📊 analyze — Quick gap detection
• 🎯 No API Key needed — Built-in smart inference engine
• 🔧 Category-aware — Code, writing, explanation, Q&A
• 🔐 Secure — API key via env vars or local config, never in git

---

🔗 Links

• 📖 Chinese README
• 📖 Integration Guide
• 📖 License
• 🐛 Issues

---

⭐ Star if useful!