Prompt Engineering

Crafting effective prompts is crucial for getting the best results from language models.

Recommendations:

• Be specific and clear - Provide detailed instructions and specify the desired format.
• Use system messages effectively - Define the assistant’s role and constraints in the system message.
• Include examples - For complex tasks, include examples of desired outputs.
• Structure matters - Organize information logically, with the most important details first.

Example: Well-Structured Prompt

response = openai.ChatCompletion.create(
  model="deepseek-ai/DeepSeek-R1",
  messages=[
    {"role": "system", "content": "You are a data analysis assistant that provides insights in a structured format."},
    {"role": "user", "content": """
      Analyze the following quarterly sales data and provide:
      1. Key trends
      2. Comparison to previous quarter
      3. Recommendations for next quarter

      DATA:
      Q3 2023: $1.2M (North: $400K, South: $350K, East: $250K, West: $200K)
      Q2 2023: $1.0M (North: $300K, South: $300K, East: $200K, West: $200K)
    """}
  ],
  temperature=0.2
)