AWS Certified Cloud Practitioner

Explanation

Correct Answer: D - Upload PDF documents to an Amazon Bedrock knowledge base. Use the knowledge base to provide context when users submit prompts to Amazon Bedrock.

Why Option D is Correct:

1. Cost-Effectiveness: Amazon Bedrock knowledge bases are designed specifically for Retrieval-Augmented Generation (RAG) applications. They provide an efficient way to retrieve relevant information from documents without needing to include entire documents in every prompt.

2. Efficient Context Management: Knowledge bases automatically:

   • Convert PDF files into embeddings
   • Store them in a vector database
   • Retrieve only the most relevant sections when processing user queries
   • This avoids the high token costs of including entire PDFs in prompts

3. Scalability: As the number of PDF files grows, knowledge bases handle the scaling efficiently without increasing costs linearly.

Why Other Options Are Not Most Cost-Effective:

Option A & B (Prompt Engineering with PDF context):

• Token Cost: Including PDF content directly in prompts consumes a large number of tokens
• Context Window Limits: LLMs have token limits (typically 4K-128K tokens), and entire PDFs may exceed these limits
• Inefficient: Every query would need to include the PDF content, leading to redundant processing

Option C (Fine-tuning a model):

• High Initial Cost: Fine-tuning requires significant computational resources and is expensive
• Maintenance Cost: Each time manuals are updated, the model would need to be re-fine-tuned
• Overkill for This Use Case: Fine-tuning is better for changing model behavior or style, not for adding factual knowledge

Key Benefits of Amazon Bedrock Knowledge Base:

• Automatic Document Processing: Handles PDF parsing, chunking, and embedding generation
• Smart Retrieval: Uses semantic search to find the most relevant information
• Cost Optimization: Only retrieves and processes relevant document sections
• Real-time Updates: Easy to add or update documents without retraining models

This approach provides the best balance of accuracy, performance, and cost for a chat interface with product manuals.