Chain-of-thought prompting implementeren

StellarStack staat op het punt zijn vlaggenschip-cloudplatform te lanceren. Marketing wil een e-mail die uitlegt waarom developers moeten migreren. Om Claude’s redenering te sturen, voeg je een checklist met “denk-stappen” toe aan de prompt.

De json-bibliotheek is geïmporteerd.

Deze oefening maakt deel uit van de cursus

Introductie tot Amazon Bedrock

Cursus bekijken

Oefeninstructies

Schets drie stappen die het model moet nemen bij het voorbereiden van het antwoord, met een genummerde lijst (1., 2., 3.).
Gebruik de steps-lijst voor de chain of thought en vul het "text"-veld van text_data door ze toe te voegen achter "\nSteps: ".

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

def generate_message_content(text_data):
    steps = [
        # Outline the steps
        "____",
        "____",
        "____"
    ]
    # Append the steps to the prompt text
    text_data["text"] += "\nSteps: " + ____
    return text_data

text_data = {"text": "Create a promotional email for the StellarStack cloud platform."}
  
print(generate_message_content(text_data))

Code bewerken en uitvoeren

Deze oefening maakt deel uit van de cursus

Introductie tot Amazon Bedrock

SkillTag.level.intermediateSkillTag.label

4.8+

Begin de cursus gratis

Learn the fundamentals of Amazon Bedrock, AWS's fully managed foundation model service. Start with basic setup and authentication, explore available foundation models like Claude and Titan, and master essential API interactions. This chapter covers setting up a development environment, choosing appropriate models for a use case, and implementing proper API request handling, including streaming responses and error management.

Exercise 1: Accessing Amazon Bedrock Exercise 2: Initializing the Bedrock client Exercise 3: Retrieving Bedrock model details Exercise 4: Selecting a foundation model Exercise 5: Using the correct model in Bedrock Exercise 6: Generating text in Bedrock

In this chapter, you'll progress through three key lessons. Starting with the basics of text generation and response handling, you'll then advance to sophisticated prompt engineering techniques like few-shot learning and structured outputs. The final lesson covers model parameter optimization, teaching learners to fine-tune settings like temperature and token limits for optimal results. Each lesson builds upon the previous, moving from fundamental API interactions to advanced parameter control for specific use cases.

Exercise 1: Generating text with language models Exercise 2: Enhancing prompts with role assignment Exercise 3: Processing model responses and error handling Exercise 4: Foundation model report categorization Exercise 5: Applying prompt engineering techniques Exercise 6: Implement few-shot learning for geometry Exercise 7: Creating structured output prompts Exercise 8: Working with different model parameters Exercise 9: Adjusting creativity with temperature control Exercise 10: Controlling output with top_p and max_tokens Exercise 11: Parameters for product descriptions

In this chapter, you’ll build robust and responsible applications with Amazon Bedrock, moving beyond basic API interactions. Starting with conversational AI, you’ll master state management and context handling to create coherent chat experiences. You’ll then explore advanced prompting techniques, learning how to generate diverse content types while maintaining consistent style and tone. The chapter covers critical production aspects like error handling, rate limiting, and efficient response processing. Finally, you will implement ethical guardrails and safety measures, ensuring your AI applications are both powerful and responsible.

Exercise 1: Ethische overwegingen en waarborgen Exercise 2: Moderatie van productreviews Exercise 3: Leeftijdsgebonden contentfilters maken Exercise 4: AI-gesprekken beheren Exercise 5: Een klassenmanager voor gesprekken bouwen Exercise 6: Een gesprek geheugenbewust maken Exercise 7: Geavanceerde prompting-strategieën Exercise 8: Chain-of-thought prompting implementeren

Huidige oefening

Exercise 9: Gestileerde contentgeneratie met Bedrock Exercise 10: API-responses en fouten afhandelen Exercise 11: Omgaan met snelheidslimieten (rate limiting)Exercise 12: Fallbacks voor modellen implementeren Exercise 13: Gefeliciteerd!