Prompt Engineering Best Practices

Instruction: [Describe the task clearly]
Context: [Any background or relevant details]
Constraints: [e.g., format, tone, length]
Examples: [Optional: provide 1-2 examples]

Input:
"""
[Text to process]
"""

Output:

# Sentiment Analysis with Few-Shot Examples

Example 1:
Text: "The product arrived damaged and customer service was unhelpful."
Sentiment: Negative
Confidence: 0.95

Example 2:
Text: "Absolutely love this! Exceeded all my expectations."
Sentiment: Positive
Confidence: 0.98

Example 3:
Text: "It's okay, nothing special but does the job."
Sentiment: Neutral
Confidence: 0.85

Now analyze:
Text: "The shipping was fast but the quality is disappointing."
Sentiment:

Question: A store sells apples for $2 each. On Monday, they sold 15 apples. 
On Tuesday, they sold 20% more than Monday. On Wednesday, they sold half 
of what they sold on Tuesday. What was their total revenue for the three days?

Let's solve this step by step:

Step 1: Calculate Monday's sales
- Apples sold: 15
- Revenue: 15 × $2 = $30

Step 2: Calculate Tuesday's sales
- Tuesday sold 20% more than Monday
- 20% of 15 = 0.20 × 15 = 3
- Tuesday apples: 15 + 3 = 18
- Revenue: 18 × $2 = $36

Step 3: Calculate Wednesday's sales
- Wednesday sold half of Tuesday
- Wednesday apples: 18 ÷ 2 = 9
- Revenue: 9 × $2 = $18

Step 4: Calculate total revenue
- Total: $30 + $36 + $18 = $84

Therefore, the total revenue for three days was $84.

# prompt_testing_framework.py

import json
from typing import List, Dict, Any
from dataclasses import dataclass
import openai

@dataclass
class PromptTest:
    name: str
    prompt_template: str
    test_cases: List[Dict[str, Any]]
    expected_qualities: List[str]
    
class PromptTester:
    def __init__(self, api_key: str):
        self.client = openai.Client(api_key=api_key)
        self.results = []
        
    def test_prompt(self, test: PromptTest, model="gpt-4"):
        """Run tests on a prompt template"""
        results = {
            "prompt_name": test.name,
            "model": model,
            "test_results": []
        }
        
        for case in test.test_cases:
            # Format prompt with test case data
            prompt = test.prompt_template.format(**case['inputs'])
            
            # Get response
            response = self.client.chat.completions.create(
                model=model,
                messages=[{"role": "user", "content": prompt}],
                temperature=case.get('temperature', 0.7)
            )
            
            # Evaluate response
            evaluation = self.evaluate_response(
                response.choices[0].message.content,
                case.get('expected_output'),
                test.expected_qualities
            )
            
            results["test_results"].append({
                "input": case['inputs'],
                "output": response.choices[0].message.content,
                "evaluation": evaluation,
                "tokens_used": response.usage.total_tokens
            })
            
        self.results.append(results)
        return results