Amazon Bedrock¶

Overview¶

Fully managed service for accessing foundation models via API.

Available Foundation Models¶

Provider	Models	Strengths
Amazon	Titan Text, Titan Embeddings	Cost-effective, AWS-native
Anthropic	Claude 3 (Opus, Sonnet, Haiku)	Complex reasoning, long context
Meta	Llama 3	Open weights, fine-tunable
Mistral	Mistral, Mixtral	Efficient, multilingual
Cohere	Command, Embed	RAG, enterprise
Stability AI	Stable Diffusion	Image generation

Key Features¶

Model Access¶

import boto3
import json

bedrock_runtime = boto3.client("bedrock-runtime")

response = bedrock_runtime.invoke_model(
    modelId="anthropic.claude-3-sonnet-20240229-v1:0",
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 1000,
        "messages": [{"role": "user", "content": "Hello!"}]
    })
)

Fine-tuning¶

Customize models with your data.

Model	Fine-tuning Support
Titan Text	Yes
Llama 3	Yes
Cohere	Yes
Claude	No (as of 2024)

Guardrails¶

Content moderation and safety.

Topic blocking
Content filters (hate, violence, etc.)
PII detection and masking
Word filters

Knowledge Bases¶

RAG (Retrieval Augmented Generation).

graph LR
    A[Documents] --> B[Chunking]
    B --> C[Embeddings]
    C --> D[Vector Store]
    E[Query] --> F[Search]
    D --> F
    F --> G[Context + Query]
    G --> H[LLM]
    H --> I[Response]

Agents¶

Autonomous task completion.

Action groups (Lambda functions)
Knowledge base integration
Multi-step reasoning

Pricing Model¶

Pricing Type	Description
On-demand	Pay per token
Provisioned	Reserved capacity
Batch	Async processing at discount

Use Cases¶

Use Case	Recommended Approach
Text generation	Claude, Titan Text
Embeddings	Titan Embeddings, Cohere Embed
RAG	Knowledge Bases
Automation	Agents
Image generation	Stable Diffusion, Titan Image

Exam Focus Areas¶

!!! warning "Key Topics" - When to use Bedrock vs SageMaker - Guardrails for content safety - Knowledge Bases for RAG patterns - Fine-tuning capabilities and limitations