AI Prompt Refiner
Transform your messy prompts into clear, effective AI instructions
✨ What It Does
AI Prompt Refiner takes your rough, unclear prompts and transforms them into well-structured, effective instructions that get better results from AI assistants like ChatGPT, Claude, or GitHub Copilot.
🚀 Installation
VS Code / Cursor / Trae / Antigravity
- Download the
.vsix file from Releases
- Press
Ctrl+Shift+P (or Cmd+Shift+P on Mac)
- Type "Install from VSIX" and select the file
- Done! ✨
Or install from marketplace: (coming soon)
🎯 How to Use
Method 1: Quick Refine (Fastest)
- Select any text in your editor
- Right-click → "Refine Selection"
- Get a perfectly structured prompt instantly
Method 2: Chat Interface with Multi-Session Support
- Open the "Chat" panel from the sidebar
- Create multiple sessions for different tasks
- Switch between sessions without losing context
- Type your rough prompt and get an optimized version
- All your conversation history is preserved
Method 3: Templates
Choose from smart templates based on your task:
- 💻 Code - Programming and development
- 📝 Writing - Documentation and content
- 📊 Analysis - Data and research tasks
- ⚡ Strict - No conversational filler
- 🎨 Custom - Create your own
💡 Why Use It?
Save Time
- ⏱️ 10 seconds to refine vs 10 minutes rewriting manually
- 🎯 Get better results on the first try
- 🧠 Stop guessing what the AI needs
Better Results
- 📋 Structured format that AI understands
- 🎨 Consistent quality across all prompts
- 🚀 More productive conversations with AI
Multi-Session Chat
- 💬 Create separate sessions for different tasks
- 🔄 Switch between conversations seamlessly
- 💾 Never lose your chat history
- 📤 Export individual sessions
🌍 Works With
| Editor |
Compatible |
| VS Code |
✅ Yes |
| Cursor |
✅ Yes |
| Trae |
✅ Yes |
| Antigravity |
✅ Yes |
| Windsurf |
✅ Yes |
🎨 Key Features
💬 AI-Powered Refinement
- Works with 6 AI providers (GitHub, OpenAI, Gemini, Groq, HuggingFace, Ollama)
- Local AI support for privacy with Ollama
- Premium providers for best results
💾 Smart Multi-Session History
- Create and manage multiple chat sessions
- Never lose your best prompts
- Search through previous refinements
- Export your favorites
- Archive old sessions
🎭 Templates
Choose from smart templates based on your task:
- 💻 Code - Programming and development
- 📝 Writing - Documentation and content
- 📊 Analysis - Data and research tasks
- ⚡ Strict - No conversational filler
- 🎨 Custom - Create your own
🎨 Roles
Each session can have a role that defines how the AI assistant approaches your prompts. When you create a new session, you'll choose a role that best fits your needs:
| Role |
Best For |
Approach |
| 💻 Programmer |
Coding tasks |
Technical precision, best practices, edge cases |
| ✍️ Writer |
Content creation |
Clarity, tone, structure, accessibility |
| 🔬 Researcher |
Analysis & investigation |
Accuracy, evidence, multiple perspectives |
| 📊 Analyst |
Problem solving |
Pattern recognition, data-driven insights |
How it works: The role adds a "personality" to your session. The same prompt will be refined differently depending on the role:
- A Programmer will add code examples and technical considerations
- A Writer will focus on clarity and audience
- A Researcher will emphasize sources and accuracy
- An Analyst will break down problems into components
Tip: Create different sessions with different roles for different types of work!
📖 Example Use Cases
For Developers 💻
Messy prompt:
fix this bug where the button doesn't work
Refined:
[Context]
React login form, submit button
[Objective]
Debug and fix non-functional button
[Constraints]
- Maintain existing styling
- Preserve accessibility
- No breaking changes
[Expected Output]
Working button with proper event handlers
For Writers 📝
Messy prompt:
write about AI
Refined:
[Context]
Technical blog, intermediate audience
[Objective]
Explain AI prompt engineering best practices
[Scope]
800-1000 words
[Constraints]
- Practical examples
- Common mistakes to avoid
- Actionable takeaways
[Expected Output]
Complete blog post with code examples
For Researchers 🔬
Messy prompt:
analyze this data
Refined:
[Context]
Q1-Q4 2024 sales data, CSV format
[Objective]
Identify trends and anomalies
[Scope]
Time series analysis with forecasting
[Constraints]
- Statistical significance testing
- Executive summary format
[Expected Output]
Analysis report with charts and recommendations
🔧 Configuration
Click the ⚡ icon in your status bar to:
- Choose AI provider
- Select different models
- Add API keys (required for all providers except Ollama)
Available Providers
- GitHub Marketplace - Free tier available with GitHub token
- OpenAI - GPT-4o, GPT-4o Mini (requires API key)
- Google Gemini - Gemini 1.5 Flash/Pro (requires API key)
- Groq - Ultra-fast inference (requires API key)
- HuggingFace - Open source models (requires API key)
- Ollama - Local AI, no API key needed
Note: Free providers (DuckDuckGo, HuggingFace public inference) are temporarily disabled due to service unavailability.
🤔 FAQ
Is it free?
Yes and no. GitHub Marketplace has a free tier. Other providers require API keys. Ollama is completely free and private.
Does it work offline?
Yes! Use Ollama with local AI models.
Is my data private?
Absolutely. We don't store or see your data. Your prompts go directly to your chosen AI provider.
Which languages are supported?
English and Spanish, with more coming soon!
How do I get help?
GitHub Issues for bugs, Discussions for questions.
⭐ Pro Tips
- Multi-Session Workflow - Create separate sessions for different projects or topics
- Choose the Right Role - Select a role that matches your task (Programmer for code, Writer for content, etc.)
- Save Sessions - Export important sessions as JSON files
- Use Templates - Match the template to your task type for best results
- History - All your prompts are saved automatically - search and reuse!
- Clear Sessions - Archive or delete old sessions to keep your workspace organized
🛠️ Troubleshooting
"Extension not showing"
→ Reload VS Code: Ctrl+Shift+P → "Reload Window"
"No provider configured"
→ Click the ⚡ icon in status bar and select a provider, then add your API key
"API Key not working"
→ Make sure you've entered the correct API key format for your provider
"Slow responses"
→ Switch to a different provider or use local Ollama
📝 Changelog
See CHANGELOG.md for detailed version history.
🤝 Contributing
Found a bug? Have an idea? We welcome contributions!
📜 License
MIT © Gperez88
Made with ❤️ for better AI conversations
⭐ Star us on GitHub
