Prompt Optimizer

Optimize your LLM prompts to reduce token consumption and save costs.

This VS Code extension helps you write better, more efficient prompts for Claude, ChatGPT, Gemini, and other LLMs by reducing token count while preserving meaning and intent.

Features

Dual Optimization Modes

Local Mode (Offline, Free, Fast)

6-Stage Optimization Pipeline: Sophisticated algorithms that work offline
Text Normalization: Cleans up whitespace and formatting
Smart Deduplication: Removes duplicate and similar sentences using fuzzy matching
Fluff Removal: Strips filler words and unnecessary phrases
Sentence Optimization: Simplifies structure and removes redundancy
Structural Reorganization: Groups and prioritizes content
Token-Specific Optimization: Replaces verbose phrases with concise alternatives

AI Mode (Gemini Flash API)

AI-Powered Optimization: Uses Google's Gemini Flash for intelligent optimization
Free Tier Available: Generous limits (15 RPM, 1M TPM, 1500 RPD)
Automatic Fallback: Falls back to local mode if API fails
Smart Retry Logic: Exponential backoff for reliability

Key Benefits

Average token reduction without losing meaning
Save money on LLM API costs
Fast local mode (sub-3 second optimization)
Secure API key storage in VS Code secrets
Context menu integration for quick access
Before/after token counts to track savings
Diff view to compare original vs optimized

Installation

From Marketplace

Open VS Code
Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X)
Search for "Prompt Optimizer"
Click Install

From .vsix File

Download the latest .vsix file from releases
Open VS Code
Go to Extensions
Click ... menu → "Install from VSIX..."
Select the downloaded file

Quick Start

Using Local Mode (No Setup Required)

Select any text in your editor
Right-click → "Optimize Prompt (Local Mode)"
View the results and choose to replace, show diff, or copy

Keyboard Shortcut: Ctrl+Alt+L (Windows/Linux) or Cmd+Alt+L (Mac)

Using AI Mode (Requires Gemini API Key)

Step 1: Get Your Free Gemini API Key

Visit Google AI Studio
Sign in with your Google account
Click "Create API Key"
Copy your API key (starts with AIza...)

Step 2: Configure the Extension

Open Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
Type "Prompt Optimizer: Setup Gemini API Key"
Paste your API key
Done!

Step 3: Use AI Mode

Select any text in your editor
Right-click → "Optimize Prompt (AI Mode - Gemini)"
Wait for AI optimization
View results and choose action

Keyboard Shortcut: Ctrl+Alt+A (Windows/Linux) or Cmd+Alt+A (Mac)

Auto Mode (Smart Selection)

Select text
Right-click → "Optimize Prompt (Auto-Select Mode)"
Choose Local or AI mode when prompted

Keyboard Shortcut: Ctrl+Alt+O (Windows/Linux) or Cmd+Alt+O (Mac)

Configuration

Open Settings (Ctrl+, / Cmd+,) and search for "Prompt Optimizer"

Available Settings

Setting	Description	Default
Default Mode	Choose `local`, `ai`, or `ask`	`ask`
Show Token Count	Display before/after token counts	`true`
Auto Replace	Automatically replace text without confirmation	`false`
Local Optimization Level	`conservative`, `moderate`, or `aggressive`	`moderate`
Gemini Model	Gemini model to use	`gemini-2.0-flash-exp`
API Timeout	Request timeout in seconds	`30`

Usage Examples

Example 1: Verbose Prompt

Original (87 tokens):

I think that you should maybe consider the fact that it might be a good idea to 
implement a feature that would allow users to be able to search through their 
documents in order to find specific information. This would be very helpful and 
useful for many people.

Optimized (21 tokens, 76% reduction):

Implement document search feature to help users find specific information.

Example 2: Technical Prompt

Original (154 tokens):

I'm working on a project and I need help with implementing a REST API endpoint. 
The endpoint should handle POST requests and it should validate the input data. 
After validation, it should save the data to the database. If there are any errors 
during the process, it should return appropriate error messages. Can you help me 
with this?

Optimized (43 tokens, 72% reduction):

Create REST API POST endpoint that:
- Validates input data
- Saves to database
- Returns appropriate errors

Need implementation help.

Example 3: Code Review Request

Original (92 tokens):

Hi there! I was wondering if you could possibly take a look at my code and maybe 
provide some feedback? I'm not sure if I'm doing things the right way and I think 
it might be helpful to get another perspective on this. Thanks in advance!

Optimized (12 tokens, 87% reduction):

Review my code and provide feedback on implementation approach.

Optimization Levels Explained

Conservative

Removes only obvious filler ("I think that", "obviously")
Minimal changes to sentence structure
Safest option, smallest reduction (~10-15%)

Moderate (Recommended)

Removes common filler and hedge words
Simplifies sentence structure
Replaces verbose phrases
Good balance (~20-30% reduction)

Aggressive

Removes all filler, including politeness
May remove articles (a, an, the)
Maximum compression
Review results carefully (~30-40% reduction)

Commands

Access via Command Palette (Ctrl+Shift+P / Cmd+Shift+P):

Prompt Optimizer: Optimize Prompt (Local Mode) - Use local algorithms
Prompt Optimizer: Optimize Prompt (AI Mode - Gemini) - Use Gemini AI
Prompt Optimizer: Optimize Prompt (Auto-Select Mode) - Choose mode
Prompt Optimizer: Setup Gemini API Key - Configure API key

Privacy & Security

API Key Storage: Stored securely using VS Code's secret storage
No Data Collection: Your prompts are never stored or logged
Local Mode: Works completely offline, no external requests
AI Mode: Only sends selected text to Gemini API when you explicitly choose AI mode

Tips for Best Results

Select Complete Thoughts: Select full sentences or paragraphs for better optimization
Review Results: Always review optimized text before use, especially in aggressive mode
Use Local Mode First: Try local mode first to see if it meets your needs
Combine Modes: Use local mode for quick edits, AI mode for complex prompts
Token Savings Add Up: Even 20% reduction saves significant costs over time

Troubleshooting

"Gemini API key is required"

Run "Setup Gemini API Key" command
Check that your API key starts with "AIza"
Verify you copied the complete key

"API request failed"

Check your internet connection
Verify API key is still valid at Google AI Studio
Check rate limits (free tier: 15 requests/minute)
Extension will automatically fall back to local mode

"No text selected"

Select some text before running optimization
Ensure you have an active editor window

Local mode results unexpected

Try different optimization levels
Report issues on GitHub

Token Counting

Token counts are estimates based on:

~4 characters per token (GPT-style tokenization)
Word-based estimation for validation
Averaged for better accuracy

Note: Exact token counts may vary by LLM model, but estimates are reliable for comparison.

Contributing

Contributions are welcome! See DEVELOPMENT.md for setup and development guidelines.

License

MIT License - See LICENSE for details

Acknowledgments

Built with VS Code Extension API
Powered by Google Gemini for AI mode
Inspired by the need to optimize LLM costs across the developer community

Changelog

Version 1.0.0

Initial release
Local optimization with 6-stage pipeline
AI optimization using Gemini Flash
Context menu integration
Token counting and statistics
Diff view support
Secure API key storage

Made with ❤️ by Sushant Bansode

Save tokens. Save money. Write better prompts.

Prompt Token Saver

Sushant Bansode