VibeBase

A Prompt Management & Model Selection Platform for LLM Engineering

VibeBase is a local-first desktop application that helps you manage prompts, compare multiple LLM models side-by-side, and debug your prompt engineering workflow with precision.

🎯 What is VibeBase?

A desktop tool for prompt engineers and LLM developers to:

• 📁 Manage and organize prompts in Markdown format
• 🏟️ Compare multiple models side-by-side (Arena Mode)
• 🔍 Select the best model for your specific use case
• 🐛 Debug prompts with detailed metrics and execution history

Features

Core Capabilities

• 🏟️ Arena Mode: Compare 2-10 LLM models simultaneously to find the best one for your prompt
• 📝 Prompt Management: Organize, version, and manage prompts in .vibe.md Markdown format
• 🔍 Model Selection: Evaluate different models side-by-side with voting and statistics
• 🐛 Engineering Debug: Track execution history, metadata, and performance metrics
• ⚡ Multi-Provider Support: OpenAI, Anthropic, DeepSeek, OpenRouter, Ollama, and more
• 🎨 Monaco Editor: Professional editing experience with syntax highlighting
• 🔐 Secure API Keys: Store credentials safely in system Keychain/Credential Manager
• 🗂️ Workspace Management: Organize prompts with folder structure and file tree navigation

Engineering & Debug Tools

• 📊 Performance Metrics: Track latency, token usage, and cost for every execution
• 📈 Arena Statistics: Analyze model performance with votes, win rates, and comparisons
• 🔄 File History: Version control for prompt iterations with rollback capability
• 🧪 Global Variables: Define reusable variables across all prompts for testing
• ⚙️ Metadata Management: Configure provider, model, temperature, and parameters per prompt
• 💾 Execution History: Complete audit trail of all prompt executions
• 🌙 Dark Mode: Adaptive theme (Light/Dark/System) for comfortable debugging
• 🗄️ Workspace Database: SQLite-based local storage for all your engineering data

Why VibeBase?

Problem: When developing LLM applications, you need to:

• Test the same prompt across multiple models
• Find the best model for your specific use case
• Track performance and cost across different providers
• Version and organize your prompts effectively

Solution: VibeBase provides a unified workspace where you can:

1. Manage all your prompts in one place
2. Compare models side-by-side with Arena mode
3. Debug with detailed metrics and execution history
4. Optimize by analyzing performance and cost data

Perfect for prompt engineers, LLM developers, and AI product teams who need to make data-driven decisions about model selection and prompt optimization.

Quick Start

Prerequisites

• Node.js 18+
• Rust 1.70+
• macOS, Windows, or Linux

Installation

Download Pre-built Binaries

Download the latest release from GitHub Releases.

macOS

1. Download the DMG file for your architecture:
   • Apple Silicon (M1/M2/M3): VibeBase_aarch64-apple-darwin.dmg
   • Intel: VibeBase_x86_64-apple-darwin.dmg
2. Open the DMG and drag VibeBase to Applications
3. Remove quarantine attribute (required for unsigned apps):

   xattr -c /Applications/VibeBase.app

4. Launch VibeBase from Applications

Alternative method: Right-click on VibeBase.app → Select "Open" → Click "Open" in the security dialog

Windows

Download and run the installer:

• MSI: VibeBase_x86_64-pc-windows-msvc.msi - Traditional Windows Installer
• NSIS: VibeBase_x86_64-pc-windows-msvc.exe - Modern installation wizard

Linux

• Debian/Ubuntu: Download and install the .deb package

  sudo dpkg -i VibeBase_x86_64-unknown-linux-gnu.deb

• Other distributions: Use the AppImage (universal format)

  chmod +x VibeBase_x86_64-unknown-linux-gnu.AppImage
  ./VibeBase_x86_64-unknown-linux-gnu.AppImage

Build from Source

# Clone the repository
git clone https://github.com/Geoion/VibeBase.git
cd VibeBase

# Install dependencies
npm install

# Run in development mode
npm run tauri dev

Build for Production

# Build for your platform
npm run tauri build

# Output will be in src-tauri/target/release/bundle/

Usage

1. Open a Workspace

Click "Open Workspace" and select a folder containing your prompts.

2. Create a Prompt File

Create a new file with .vibe.md extension:

# Customer Greeting

Generate personalized greeting messages for customers.

## System Message

You are a friendly customer service representative.
Always be warm and professional.

## User Message

Customer: {{customer_name}}
Account Type: {{account_type}}

Generate a greeting message.

3. Configure LLM Provider

1. Open Settings (⚙️ icon)
2. Go to Providers tab
3. Add a new provider:
   • Provider: OpenAI
   • API Key: Your OpenAI API key
   • Models: Select models to enable (e.g., gpt-4o, gpt-4o-mini)
4. Save configuration

4. Execute a Prompt

1. Click on your prompt file in the Navigator
2. Fill in variable values in the Execution panel
3. Select a model from the dropdown
4. Click Run button
5. View the response with metadata (latency, tokens, cost)

5. Use Arena Mode

1. Open Settings → Arena
2. Enable concurrent execution and configure settings
3. Select multiple models in the Execution panel
4. Click Run Arena
5. Compare responses side-by-side
6. Vote for the best response

Architecture

Tech Stack

Frontend

• React 18 + TypeScript
• Vite (Build tool)
• Tailwind CSS (Styling)
• Monaco Editor (Code editor)
• Zustand (State management)
• react-i18next (Internationalization)

Backend

• Tauri 1.5 (Rust)
• SQLite (Database)
• keyring (System Keychain)
• reqwest (HTTP client)
• serde (Serialization)

Project Structure

VibeBase/
├── src/                    # Frontend (React + TypeScript)
│   ├── components/         # React components
│   │   ├── arena/          # Arena mode components
│   │   ├── editor/         # Monaco editor
│   │   ├── execution/      # Execution panel
│   │   ├── history/        # File history
│   │   ├── metadata/       # Metadata management
│   │   ├── settings/       # Settings panels
│   │   └── variables/      # Global variables
│   ├── stores/             # Zustand stores
│   ├── i18n/               # Translations
│   └── App.tsx             # Main app component
├── src-tauri/              # Backend (Rust)
│   ├── src/
│   │   ├── commands/       # Tauri commands
│   │   ├── db/             # Database operations
│   │   ├── llm/            # LLM integrations
│   │   └── main.rs         # Entry point
│   └── Cargo.toml
└── examples/               # Example workspaces

Data Storage

~/.vibebase/                # Global application data
├── app.db                  # Global database
│   ├── llm_providers       # Provider configurations
│   ├── global_variables    # Global variables
│   └── workspace_history   # Recent workspaces
└── logs/                   # Application logs

{workspace}/.vibebase/      # Workspace-specific data
└── workspace.db            # Workspace database
    ├── prompt_metadata     # Prompt configurations
    ├── execution_history   # Execution records
    ├── file_history        # Version history
    └── arena_battles       # Arena results

Supported LLM Providers

Provider	Status	Base URL	Notes
OpenAI	✅	api.openai.com/v1	Full support
Anthropic	✅	api.anthropic.com	Claude models
DeepSeek	✅	api.deepseek.com	Chinese LLM
OpenRouter	✅	openrouter.ai/api/v1	Multi-provider gateway
Ollama	✅	localhost:11434/v1	Local models
AiHubMix	✅	aihubmix.com/v1	OpenAI-compatible
Custom	✅	User-defined	OpenAI-compatible APIs

Custom Provider Support

VibeBase supports any OpenAI-compatible API endpoint. You can add custom providers by:

1. Settings → Providers → Add Custom Provider
2. Enter provider ID, display name, and base URL
3. Configure API key and models

Key Features

Arena Mode - Find Your Best Model

The killer feature for model selection and prompt engineering.

Arena Mode allows you to compare multiple LLM models simultaneously with the same prompt, helping you make data-driven decisions about which model performs best for your specific use case.

How it works:

1. Select 2-10 models from different providers
2. Run the same prompt across all models concurrently
3. View responses side-by-side in a clean comparison view
4. Vote for the best response based on your criteria
5. Analyze statistics: win rates, average latency, cost per model
6. Make informed decisions about model selection

Arena Settings:

• Concurrent Execution: Run models in parallel for faster results
• Max Concurrent Models: Choose 1-10 models (default: 3)
• Cost Warning: Get alerts when execution cost exceeds threshold
• Auto-Save Results: Automatically save all battle results to database
• Remember Selection: Restore previously selected models
• Card Density: Customize information density (Compact/Normal/Detailed)

Use Cases:

• Compare GPT-4 vs Claude vs DeepSeek for your specific task
• Find the most cost-effective model that meets quality standards
• A/B test different providers for production deployment
• Evaluate new models as they are released

Global Variables

Define variables once, use everywhere:

# In any prompt file
Customer: {{company_name}}
API Endpoint: {{api_base_url}}

Variable Management:

• Create global variables in Settings
• Use {{variable_name}} syntax in prompts
• Override with custom values during execution
• Stored in ~/.vibebase/app.db

File History

Every save creates a version snapshot:

• View all historical versions
• Preview differences
• Rollback to any previous version
• Automatic timestamping

Metadata Management

Configure per-prompt settings:

• Tags: Organize and filter prompts
• Provider Reference: Link to global LLM configuration
• Model Override: Override default model
• Parameters: Temperature, max tokens
• Test Data: Path to test data file
• Variables: Detected variables

Troubleshooting

Prompt files not showing up

Solution:

1. Ensure files have .vibe.md extension
2. Click Refresh button in Navigator
3. Check file is not in hidden directory

Database errors

Solution:

1. Close the application
2. Delete .vibebase/ directory
3. Restart application (database will be recreated)

API key issues

Solution:

1. Go to Settings → Providers
2. Re-enter API key
3. Test connection with "Test Connection" button

Arena mode not working

Solution:

1. Ensure multiple models are selected
2. Check all providers have valid API keys
3. Verify network connection
4. Check Settings → Arena for configuration

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

How to Contribute

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

MIT License - see LICENSE file for details

---

⭐ Star us on GitHub if you find VibeBase useful!