A Python GUI tool for systematically improving prompts using LLM feedback. PromptBuster implements a 7-step workflow that leverages the LLM's own weights to generate better prompts than you could write manually.

• 7-Step Workflow: Guided process for prompt improvement
• Multiple LLM Providers: Support for OpenAI, Anthropic, and local models
• Modern GUI: Clean Tkinter interface with CustomTkinter
• Configuration Management: Save/load settings and sessions
• Async Processing: Non-blocking LLM requests

1. Clone the repository:

git clone <repository-url>
cd PromptBuster

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables:

cp .env.example .env
# Edit .env with your API keys

1. Run the application:

python main.py

2. Configure your LLM provider in the left panel
3. Follow the 7-step workflow:
   • Step 1: Define your target role and generate initial prompt guide
   • Step 2: Input 5 examples of desired input/output behavior
   • Step 3: Generate a prompt from your examples
   • Step 4: Generate an evaluation guide for your role
   • Step 5: Evaluate the generated prompt
   • Step 6: Generate 3 improved alternatives
   • Step 7: Select and edit your final prompt

1. Initial Prompt Guide: Generate a detailed prompt engineering guide for your target audience
2. Examples Input: Provide 5 examples of how you want your prompt to work
3. Prompt Generation: Generate a prompt that could produce your examples' outputs
4. Evaluation Guide: Generate a detailed prompt evaluation guide
5. Prompt Evaluation: Evaluate the generated prompt using the evaluation guide
6. Improved Alternatives: Generate 3 improved alternative prompts
7. Final Selection: Select and edit your final prompt

• OpenAI: GPT-4, GPT-3.5-turbo, etc.
• Anthropic: Claude-3 models
• Local: Any OpenAI-compatible API endpoint

Set your API keys in environment variables:

• OPENAI_API_KEY: For OpenAI models
• ANTHROPIC_API_KEY: For Anthropic models

Or configure them directly in the GUI.

• Better than manual: LLM's own weights influence prompt generation
• Systematic: Follows a proven 7-step methodology
• Family-optimized: Best results when using the same model family for generation and final use
• Iterative: Each step builds on the previous to improve quality

• Python 3.8+
• See requirements.txt for full dependency list

See LICENSE file for details.