DSPy: Problems & Solutions - Testable Examples

This repository contains practical examples demonstrating the problems DSPy solves and how it solves them. Perfect for learning DSPy through hands-on examples.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

📁 Structure

dspy_problems_solutions/
├── README.md                          # This file
├── requirements.txt                   # Dependencies
├── problem_01_brittle_prompts/        # Problem 1: Manual prompt engineering
├── problem_02_few_shot_examples/      # Problem 2: Finding good examples
├── problem_03_prompt_optimization/    # Problem 3: Manual optimization
├── problem_04_model_portability/      # Problem 4: Model-specific code
├── problem_05_complex_pipelines/       # Problem 5: Building pipelines
├── problem_06_systematic_improvement/  # Problem 6: Performance improvement
└── problem_07_reproducibility/        # Problem 7: Version control

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Set Up API Keys (Optional - for real LLM testing)

export OPENAI_API_KEY="your-key-here"
# or
export ANTHROPIC_API_KEY="your-key-here"

3. Run Examples

Each problem folder contains:

• traditional_approach.py - Shows the problem
• dspy_solution.py - Shows the solution
• README.md - Explanation

Run any example:

cd problem_01_brittle_prompts
python traditional_approach.py
python dspy_solution.py

📊 Problem Overview

Problem	Folder	Key Issue	DSPy Solution
1. Brittle Prompts	problem_01_brittle_prompts/	Manual prompt crafting	Declarative signatures
2. Few-Shot Examples	problem_02_few_shot_examples/	Manual example selection	Automatic optimization
3. Prompt Optimization	problem_03_prompt_optimization/	Manual iteration	Systematic search
4. Model Portability	problem_04_model_portability/	Model-specific code	Model-agnostic framework
5. Complex Pipelines	problem_05_complex_pipelines/	Manual composition	Modular building blocks
6. Systematic Improvement	problem_06_systematic_improvement/	Manual A/B testing	Automatic optimization
7. Reproducibility	problem_07_reproducibility/	Scattered prompts	Version-controlled modules

🎯 Learning Path

1. Start with Problem 1 - Understand the basic issue with manual prompts
2. Progress sequentially - Each problem builds on previous concepts
3. Compare approaches - Run both traditional and DSPy versions
4. Experiment - Modify examples to see how DSPy adapts

💡 Key Concepts

• Signatures: Declarative way to define input/output
• Modules: Composable building blocks
• Optimizers: Automatic prompt/example optimization
• Metrics: Define what "good" means for your task

🔧 Requirements

• Python 3.8+
• DSPy library
• (Optional) LLM API keys for real testing

📝 Notes

• Examples use mock responses when API keys aren't available
• All examples are runnable and testable
• Each example includes clear comments explaining the problem

🤝 Contributing

Contributions are welcome! Feel free to:

• Add more problem examples
• Improve existing examples
• Fix bugs or typos
• Enhance documentation

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Happy Learning! 🚀