Bedrock

🧠 What is Amazon Bedrock?

Amazon Bedrock is a serverless platform that lets you build and scale generative AI applications using pre-trained Foundation Models (FMs) from leading AI companies, via a unified API.
You don’t need to manage GPUs, train models, or fine-tune from scratch.

✅ It supports text generation, summarization, chat, image generation, embeddings, and more.

---

🔗 Supported Foundation Model Providers (as of 2024)

Provider	Models Offered
Anthropic	Claude 1, 2, 3 (for reasoning, chat, docs)
Meta	Llama 2 & 3
Cohere	Command R+, Embed models
AI21 Labs	Jurassic-2 (for text generation)
Mistral AI	Mistral 7B, Mixtral (Mixture-of-Experts)
Stability AI	Stable Diffusion (image generation)
Amazon Titan	Titan Text, Titan Embeddings, Titan Image

---

🔧 Core Capabilities

Feature	Description
Text generation	Chatbots, summarization, content creation
Embeddings	Semantic search, vector databases
Image generation	AI image creation with Stability AI
Agents	Automatically execute tasks with grounding and orchestration
Knowledge bases	Retrieval-Augmented Generation (RAG) over your enterprise data
Model evaluation	Compare models side by side in Studio
Fine-tuning	Train custom versions of Titan models
Secure API access	IAM-based auth, no model weights exposed

---

✨ Example Use Cases

Use Case	Example
Chatbot	Claude or Llama-based assistant
Document Q&A	Titan + RAG on S3 + OpenSearch
Image generator	Stability AI → Generate art from prompts
Embedding + search	Titan Embeddings + vector DB (e.g., Pinecone, FAISS)
Summarization	Upload PDFs → Claude → Text summary
Code generation	LLMs for code, CLI, function writing

---

🧪 Sample Bedrock Python SDK (Boto3)

Invoke Claude 3

import boto3

bedrock = boto3.client('bedrock-runtime')

response = bedrock.invoke_model(
    modelId="anthropic.claude-3-sonnet-20240229-v1:0",
    contentType="application/json",
    accept="application/json",
    body=json.dumps({
        "messages": [
            {"role": "user", "content": "Explain quantum computing simply."}
        ],
        "max_tokens": 512
    })
)

print(json.loads(response['body'].read()))

---

Embed text using Titan Embeddings

response = bedrock.invoke_model(
    modelId="amazon.titan-embed-text-v1",
    contentType="application/json",
    accept="application/json",
    body=json.dumps({"inputText": "The capital of France is Paris."})
)

print(json.loads(response['body'].read())["embedding"])

---

📚 Knowledge Bases (RAG) Support

You can link Bedrock with:

• Amazon OpenSearch

• Pinecone

• Redis Enterprise Cloud

To build Retrieval-Augmented Generation (RAG) pipelines with vector search over:

• S3 PDFs, TXT

• Internal documents

• FAQs, Wikis

✅ Bedrock natively supports RAG workflows via “Knowledge Bases”.

---

⚙️ Bedrock Agents

Bedrock Agents allow models to:

• Use tools (APIs, functions) to complete tasks (e.g., book a ticket, fetch weather)

• Perform reasoning over multi-step prompts

• Combine RAG + tools + models with memory

---

🧾 Pricing Overview

Feature	Price (2024 estimate)
Text generation (Claude 3)	$0.003 - $0.015 / 1K tokens (input/output)
Embeddings (Titan)	$0.0001 / 1K tokens
Image generation (Stable)	~$0.02 per image
Agents & KB	Usage-based + underlying model cost

✅ You pay only for what you use (tokens in/out).

---

🔐 Security & Compliance

Feature	Supported
IAM + API Gateway	✅ Yes
VPC support	✅ Yes via PrivateLink
Encryption (KMS)	✅ Yes
Logging (CloudTrail)	✅ Yes
HIPAA eligible	✅ Yes (Titan only)
Data stays in AWS	✅ Models don’t retain inputs

---

🧱 Terraform Support

Bedrock resources are not natively available in Terraform (as of now), but you can:

• Use Lambda + Bedrock Runtime API

• Create knowledge base infrastructure using:

  • aws_opensearch_domain

  • aws_s3_bucket

  • aws_lambda_function

Terraform Sample (S3 for RAG):

resource "aws_s3_bucket" "bedrock_knowledge_data" {
  bucket = "bedrock-rag-files"
  force_destroy = true
}

---

🧠 Bedrock vs Alternatives

Service	Bedrock	OpenAI (API)	Vertex AI
Managed by AWS	✅ Yes	❌ No	✅ Yes
Multi-model hub	✅ Yes	❌ (OpenAI only)	✅ (limited)
Fine-tuning	✅ Titan only	✅ GPT-3.5/4 Turbo	✅
RAG Support	✅ Native KB + Agents	❌ (DIY)	✅
VPC Support	✅ Yes (PrivateLink)	❌ No	✅ Yes

---

✅ TL;DR Summary

Feature	Amazon Bedrock
Foundation Models	Claude, Llama, Titan, Cohere, AI21
Image generation	✅ Stability AI (SDXL)
Embeddings	✅ Titan, Cohere
Tools/Agents	✅ Orchestration of LLMs
RAG pipeline	✅ Built-in with Knowledge Base
Real-time inference	✅ Yes (no infra to manage)
Fine-tuning	✅ Titan Text/Image only
Terraform support	❌ Not yet full
Free tier	❌ No free tier yet

---