{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"A comprehensive technical guide covering AWS Bedrock, Retrieval-Augmented Generation (RAG), voice AI agents, vector databases, and production-grade CI/CD deployment pipelines."}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"Table of Contents"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#1-cloud-providers--aws-overview","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Cloud Providers & AWS Overview"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#2-aws-bedrock-deep-dive","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"AWS Bedrock Deep Dive"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#3-retrieval-augmented-generation-rag","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Retrieval-Augmented Generation (RAG)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#4-vector-databases--embeddings","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Vector Databases & Embeddings"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#5-ai-voice-agent-architecture-with-pipecat","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"AI Voice Agent Architecture with Pipecat"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#6-backend-python-fastapi--aws-lambda","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Backend: Python, FastAPI & AWS Lambda"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#7-frontend-react-chat-interface","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Frontend: React Chat Interface"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#8-cicd-pipeline--dockerkubernetes-deployment","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"CI/CD Pipeline & Docker/Kubernetes Deployment"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#9-responsible-ai--guardrails","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Responsible AI & Guardrails"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#10-end-to-end-architecture-flow","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"End-to-End Architecture Flow"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/f47b220c-bde6-408a-b7c7-cc2c9888f896#11-troubleshooting-common-issues","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Troubleshooting Common Issues"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"1. Cloud Providers & AWS Overview"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"The Major Cloud Players"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Before diving into AWS Bedrock, understanding the cloud ecosystem helps in choosing the right provider for AI workloads."}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Provider"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Typical Adoption"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Azure"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Enterprise companies; often preferred due to pricing models"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Startups and scale-ups; widest ecosystem of services"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"GCP"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Mid-to-large companies; strong in data and ML workloads"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"DigitalOcean / Heroku"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Smaller teams and hobbyist projects"}]}]}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Definition:"},{"type":"text","text":" A cloud provider offers infrastructure (IaaS), platform (PaaS), and software services (SaaS) that enable organizations to build, deploy, and scale applications without managing physical hardware."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Why AWS for AI Projects?"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS is the preferred choice for AI projects because:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Largest catalog of managed AI/ML services"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bedrock's native integration with Lambda, S3, and API Gateway"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Mature IAM for fine-grained permission control"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Extensive free-tier and credit offerings for experimentation"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"2. AWS Bedrock Deep Dive"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/2bzb4vwunyazbhjklin1xzqr_-angptzj7yoxzesnic0kxxufp_i4xytko2t6z6sagsjjvi_vmwe7ktlwjoqzikhbur3altuk8hw","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"What is AWS Bedrock?"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Definition:"},{"type":"text","text":" Amazon Bedrock is a fully managed service that provides both "},{"type":"text","marks":[{"type":"bold"}],"text":"hosting"},{"type":"text","text":" and "},{"type":"text","marks":[{"type":"bold"}],"text":"provider"},{"type":"text","text":" capabilities for foundation models (LLMs), enabling developers to access, customize, and deploy models via console, SDK, or APIs — without managing any underlying ML infrastructure."}]}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bedrock acts in two roles:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Host"},{"type":"text","text":" – Provides the compute and execution environment to run models"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Provider"},{"type":"text","text":" – Offers a catalog of pre-trained models from multiple vendors"}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Supported Model Providers"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Provider"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Example Model"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Modality"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Amazon"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Titan"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Anthropic"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Claude"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Meta"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Llama 3 / 8B"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cohere"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Command R"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Mistral"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Mistral-7B"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Stability AI"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Stable Diffusion"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Image"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"NVIDIA"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"NeMo"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Audio"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Navigating the Bedrock Console"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Once your AWS account is set up, access Bedrock by:"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Searching \"Amazon Bedrock\" in the AWS console search bar"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The left-hand menu provides:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Overview"},{"type":"text","text":" – Dashboard and getting started guides"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Model Catalog"},{"type":"text","text":" – Browse and enable foundation models"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Guardrails"},{"type":"text","text":" – Define content filters and safety policies"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Agent Builder"},{"type":"text","text":" – Create multi-step AI agents"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Playground"},{"type":"text","text":" – Interactive testing environment"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"API Keys / Credentials"},{"type":"text","text":" – Generate long-term tokens"}]}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Using the Playground"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The Playground is your sandbox for testing models before writing any code:"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Go to "},{"type":"text","marks":[{"type":"bold"}],"text":"Model Catalog → Select Model"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Choose a free-tier model (e.g., "},{"type":"text","marks":[{"type":"code"}],"text":"Nova2Lite"},{"type":"text","text":", "},{"type":"text","marks":[{"type":"code"}],"text":"meta.llama3-8b-instruct"},{"type":"text","text":", or "},{"type":"text","marks":[{"type":"code"}],"text":"amazon.titan-text-express-v1"},{"type":"text","text":")"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Click "},{"type":"text","marks":[{"type":"bold"}],"text":"Apply"},{"type":"text","text":" to enable the model"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Navigate to "},{"type":"text","marks":[{"type":"bold"}],"text":"Playground → Chat or Text tab"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Enter a "},{"type":"text","marks":[{"type":"bold"}],"text":"system prompt"},{"type":"text","text":" to define model behavior:"}]}]}]},{"type":"codeBlock","attrs":{"language":"text"},"content":[{"type":"text","text":"System Prompt Example:\n\"You are a senior marketing manager. Your task is to provide a detailed\nproduct launch plan whenever asked. Always structure your response with\nsections: Objective, Target Audience, Channels, Timeline, and KPIs.\"\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Prompt Engineering Basics"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"System Prompt"},{"type":"text","text":" = Persona / instruction set that defines how the model behaves throughout the session"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Chain-of-thought"},{"type":"text","text":" = Embed reasoning instructions like "},{"type":"text","marks":[{"type":"italic"}],"text":"\"Think step by step before answering\""}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Temperature"},{"type":"text","text":" = Controls creativity (0 = deterministic, 1 = creative)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Top-P"},{"type":"text","text":" = Controls diversity of token selection"}]}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Example: Calling Bedrock via Boto3\nimport boto3\nimport json\n\nbedrock = boto3.client(service_name='bedrock-runtime', region_name='us-east-1')\n\nbody = json.dumps({\n \"inputText\": \"Explain RAG in simple terms.\",\n \"textGenerationConfig\": {\n \"maxTokenCount\": 512,\n \"temperature\": 0.7,\n \"topP\": 0.9\n }\n})\n\nresponse = bedrock.invoke_model(\n body=body,\n modelId='amazon.titan-text-express-v1',\n accept='application/json',\n contentType='application/json'\n)\n\nresponse_body = json.loads(response.get('body').read())\nprint(response_body['results'][0]['outputText'])\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"API Access & Authentication"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bedrock provides two types of API credentials:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Short-term keys"},{"type":"text","text":" – Tied to the console session; expire automatically"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Long-term API keys"},{"type":"text","text":" – Generated from the Bedrock console for programmatic/production use"}]}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Best practice:"},{"type":"text","text":" Always use long-term API keys stored in "},{"type":"text","marks":[{"type":"bold"}],"text":"AWS Secrets Manager"},{"type":"text","text":" for production workloads, never hardcode credentials in source code."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Pricing Model"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Free tier:"},{"type":"text","text":" $100 AWS credit valid for 6 months"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Token-based billing:"},{"type":"text","text":" Charged per 1,000 tokens (input + output)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Short-term vs long-term:"},{"type":"text","text":" Committed usage (monthly/yearly) can unlock 50–80% discounts"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Monitor costs:"},{"type":"text","text":" Use "},{"type":"text","marks":[{"type":"bold"}],"text":"Cost & Usage → Billing Dashboard"},{"type":"text","text":" to avoid surprise bills"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"3. Retrieval-Augmented Generation (RAG)"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/cskq-upwj2s1ymtjkaziotdzsdarmuvjx-f2z5e3vfw_qr3ivhdygxmti5bxdstvit0uknef_da7hajrysavlybeeygd9-thwi9n","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"What is RAG?"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Definition:"},{"type":"text","text":" RAG (Retrieval-Augmented Generation) is a technique that combines a "},{"type":"text","marks":[{"type":"bold"}],"text":"retrieval system"},{"type":"text","text":" (vector search over a knowledge base) with a "},{"type":"text","marks":[{"type":"bold"}],"text":"generative model"},{"type":"text","text":" (LLM), allowing the model to ground its answers in external, up-to-date data rather than relying solely on parametric memory baked into model weights during training."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Why RAG?"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Without RAG, an LLM:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cannot access data beyond its training cutoff"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"May hallucinate facts not in its training corpus"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cannot reference your private/internal documents"}]}]}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"With RAG, the LLM:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Retrieves relevant context at query time"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Produces answers grounded in your actual data"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Can be updated without retraining the model"}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"The RAG Pipeline: Step by Step"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"[Raw Documents]\n │\n ▼\n[Chunking] ──── Split into manageable pieces (e.g., 512 tokens each)\n │\n ▼\n[Embedding] ─── Convert each chunk to a high-dimensional vector\n │\n ▼\n[Vector Store] ─ Persist vectors in MongoDB Atlas / Pinecone / FAISS\n │\n ▼\n[Query Time]\n │\n [User Query]\n │\n ▼\n[Embed Query] ── Convert user question to vector\n │\n ▼\n[Similarity Search] ── Find top-K most relevant chunks\n │\n ▼\n[Augment Prompt] ─── Inject retrieved chunks into LLM context window\n │\n ▼\n[LLM Generation] ─── Generate final grounded answer\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"RAG Implementation with Boto3 and MongoDB Atlas"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from pymongo import MongoClient\nimport boto3\nimport json\nimport numpy as np\n\n# Step 1: Embed a query using Amazon Titan Embeddings\ndef embed_text(text: str) -> list:\n bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')\n body = json.dumps({\"inputText\": text})\n response = bedrock.invoke_model(\n body=body,\n modelId='amazon.titan-embed-text-v1',\n accept='application/json',\n contentType='application/json'\n )\n return json.loads(response['body'].read())['embedding']\n\n# Step 2: Retrieve relevant chunks from MongoDB Atlas Vector Search\ndef retrieve_chunks(query: str, top_k: int = 3) -> list:\n client = MongoClient(\"your_mongodb_atlas_uri\")\n collection = client[\"rag_db\"][\"documents\"]\n \n query_vector = embed_text(query)\n \n pipeline = [\n {\n \"$vectorSearch\": {\n \"index\": \"vector_index\",\n \"path\": \"embedding\",\n \"queryVector\": query_vector,\n \"numCandidates\": 100,\n \"limit\": top_k\n }\n },\n {\"$project\": {\"text\": 1, \"score\": {\"$meta\": \"vectorSearchScore\"}}}\n ]\n \n return list(collection.aggregate(pipeline))\n\n# Step 3: Augment and generate\ndef rag_answer(user_query: str) -> str:\n chunks = retrieve_chunks(user_query)\n context = \"\\n\\n\".join([chunk['text'] for chunk in chunks])\n \n augmented_prompt = f\"\"\"You are a helpful assistant. Use only the following context to answer the question.\n \nContext:\n{context}\n\nQuestion: {user_query}\n\nAnswer:\"\"\"\n \n bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')\n body = json.dumps({\n \"inputText\": augmented_prompt,\n \"textGenerationConfig\": {\"maxTokenCount\": 512, \"temperature\": 0.3}\n })\n \n response = bedrock.invoke_model(\n body=body,\n modelId='amazon.titan-text-express-v1',\n accept='application/json',\n contentType='application/json'\n )\n \n return json.loads(response['body'].read())['results'][0]['outputText']\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"4. Vector Databases & Embeddings"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/ajm5axpx38asyuzakkqupex8if0fxjykkh5x029tcjphuatsm6i-92xg-jg-xxwxazfdj_fxg4cd7wvvntjj8vh_sviuzejjqe35","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"What is a Vector Database?"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Definition:"},{"type":"text","text":" A vector database is a specialized data store that indexes high-dimensional numerical vectors (embeddings) and enables "},{"type":"text","marks":[{"type":"bold"}],"text":"fast similarity search"},{"type":"text","text":" — finding vectors closest to a query vector using distance metrics like cosine similarity or Euclidean distance."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"How Embeddings Work"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Embeddings convert text (or audio/images) into fixed-length numerical vectors:"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"\"AWS Bedrock is a managed LLM service\"\n │\n [Embedding Model]\n │\n ▼\n[0.234, -0.812, 0.091, 0.543, ..., -0.321] ← 1536-dimensional vector\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Semantically similar text produces vectors that are geometrically "},{"type":"text","marks":[{"type":"bold"}],"text":"close"},{"type":"text","text":" in vector space:"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Chunking a document before embedding\ndef chunk_document(text: str, chunk_size: int = 512, overlap: int = 50) -> list:\n \"\"\"Split a document into overlapping chunks for better retrieval.\"\"\"\n words = text.split()\n chunks = []\n for i in range(0, len(words), chunk_size - overlap):\n chunk = \" \".join(words[i:i + chunk_size])\n chunks.append(chunk)\n return chunks\n\n# Ingesting documents into MongoDB Atlas\ndef ingest_document(file_path: str):\n with open(file_path, 'r') as f:\n text = f.read()\n \n chunks = chunk_document(text)\n client = MongoClient(\"your_mongodb_atlas_uri\")\n collection = client[\"rag_db\"][\"documents\"]\n \n documents = []\n for chunk in chunks:\n embedding = embed_text(chunk)\n documents.append({\n \"text\": chunk,\n \"embedding\": embedding,\n \"source\": file_path\n })\n \n collection.insert_many(documents)\n print(f\"Ingested {len(documents)} chunks from {file_path}\")\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"MongoDB Atlas Vector Search Setup"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Create a cluster on "},{"type":"text","marks":[{"type":"link","attrs":{"href":"https://cloud.mongodb.com/","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"MongoDB Atlas"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Enable "},{"type":"text","marks":[{"type":"bold"}],"text":"Vector Search"},{"type":"text","text":" on your collection"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Create a search index:"}]}]}]},{"type":"codeBlock","attrs":{"language":"json"},"content":[{"type":"text","text":"{\n \"mappings\": {\n \"dynamic\": true,\n \"fields\": {\n \"embedding\": {\n \"dimensions\": 1536,\n \"similarity\": \"cosine\",\n \"type\": \"knnVector\"\n }\n }\n }\n}\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"5. AI Voice Agent Architecture with Pipecat"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"What is Pipecat?"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Definition:"},{"type":"text","text":" Pipecat is an open-source Python framework for building "},{"type":"text","marks":[{"type":"bold"}],"text":"real-time voice and multimodal conversation agents"},{"type":"text","text":", handling audio/video pipelines, transport, and orchestration between Speech-to-Text, LLM, and Text-to-Speech components."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Voice Agent Component Stack"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Component"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Service"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Role"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Speech-to-Text (STT)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Deepgram"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Converts microphone audio → text"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"LLM"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Grok / Llama 3 via Bedrock"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Text understanding & generation"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Text-to-Speech (TTS)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"ElevenLabs (11Labs)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Converts generated text → spoken audio"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Orchestrator"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Pipecat"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Coordinates all components in real-time"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Architecture Flow"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"[User Microphone]\n │ (audio stream)\n ▼\n[WebSocket / HTTP Transport]\n │\n ▼\n[Deepgram STT] ──── transcribed text ────▶ [RAG Pipeline]\n │\n retrieved context\n │\n ▼\n [LLM - Llama3/Grok]\n │\n generated text\n │\n ▼\n [ElevenLabs TTS]\n │\n audio response\n │\n ▼\n [Frontend Playback]\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Pipecat Implementation Skeleton"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Install: pip install pipecat-ai deepgram-sdk elevenlabs\nimport asyncio\nfrom pipecat.pipeline.pipeline import Pipeline\nfrom pipecat.pipeline.task import PipelineTask\nfrom pipecat.services.deepgram import DeepgramSTTService\nfrom pipecat.services.elevenlabs import ElevenLabsTTSService\nfrom pipecat.services.openai import OpenAILLMService # or custom Bedrock service\n\nasync def run_voice_agent():\n # Initialize services\n stt = DeepgramSTTService(api_key=\"YOUR_DEEPGRAM_KEY\")\n \n llm = OpenAILLMService(\n api_key=\"YOUR_LLM_KEY\",\n model=\"grok-llama3\"\n )\n \n tts = ElevenLabsTTSService(\n api_key=\"YOUR_ELEVENLABS_KEY\",\n voice_id=\"YOUR_VOICE_ID\"\n )\n \n # Build the pipeline: STT → LLM → TTS\n pipeline = Pipeline([stt, llm, tts])\n task = PipelineTask(pipeline)\n \n await task.run()\n\nif __name__ == \"__main__\":\n asyncio.run(run_voice_agent())\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"6. Backend: Python, FastAPI & AWS Lambda"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Technology Stack"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Technology"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Role"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Python"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Core backend language"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Boto3"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS SDK for Python — calls Bedrock, S3, Lambda APIs"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"FastAPI"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"High-performance async API framework"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS Lambda"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Serverless compute for backend logic"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Amazon S3"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Storage for documents, model artifacts, logs"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"API Gateway"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Expose Lambda/FastAPI endpoints to the frontend"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"FastAPI + Bedrock Integration"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# main.py\nfrom fastapi import FastAPI, HTTPException\nfrom pydantic import BaseModel\nimport boto3\nimport json\n\napp = FastAPI(title=\"RAG AI Agent API\", version=\"1.0.0\")\n\nclass QueryRequest(BaseModel):\n question: str\n use_voice: bool = False\n\nclass QueryResponse(BaseModel):\n answer: str\n sources: list[str]\n\n@app.post(\"/query\", response_model=QueryResponse)\nasync def query_agent(request: QueryRequest):\n \"\"\"Main endpoint: receives user question, runs RAG pipeline, returns answer.\"\"\"\n try:\n # Step 1: Retrieve relevant context\n chunks = retrieve_chunks(request.question)\n sources = [chunk.get('source', 'unknown') for chunk in chunks]\n context = \"\\n\\n\".join([chunk['text'] for chunk in chunks])\n \n # Step 2: Generate answer via Bedrock\n bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')\n prompt = f\"Context:\\n{context}\\n\\nQuestion: {request.question}\\nAnswer:\"\n \n body = json.dumps({\n \"inputText\": prompt,\n \"textGenerationConfig\": {\n \"maxTokenCount\": 512,\n \"temperature\": 0.3\n }\n })\n \n response = bedrock.invoke_model(\n body=body,\n modelId='amazon.titan-text-express-v1',\n accept='application/json',\n contentType='application/json'\n )\n \n answer = json.loads(response['body'].read())['results'][0]['outputText']\n return QueryResponse(answer=answer, sources=sources)\n \n except Exception as e:\n raise HTTPException(status_code=500, detail=str(e))\n\n@app.get(\"/health\")\nasync def health_check():\n return {\"status\": \"ok\"}\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"AWS Lambda Handler Wrapper"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# lambda_handler.py — wraps FastAPI for AWS Lambda deployment\nfrom mangum import Mangum\nfrom main import app\n\n# Mangum adapter converts Lambda events to ASGI requests\nhandler = Mangum(app)\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"IAM Permissions Required"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Your Lambda execution role needs these policies:"}]},{"type":"codeBlock","attrs":{"language":"json"},"content":[{"type":"text","text":"{\n \"Version\": \"2012-10-17\",\n \"Statement\": [\n {\n \"Effect\": \"Allow\",\n \"Action\": [\n \"bedrock:InvokeModel\",\n \"bedrock:ListFoundationModels\"\n ],\n \"Resource\": \"*\"\n },\n {\n \"Effect\": \"Allow\",\n \"Action\": [\n \"s3:GetObject\",\n \"s3:PutObject\"\n ],\n \"Resource\": \"arn:aws:s3:::your-bucket-name/*\"\n }\n ]\n}\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"7. Frontend: React Chat Interface"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Architecture"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The React frontend:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Captures user text input (and microphone audio for voice mode)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Sends requests to the FastAPI backend via API Gateway"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Displays chat transcript with streamed responses"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Plays back synthesized speech from ElevenLabs"}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Core Chat Component"}]},{"type":"codeBlock","attrs":{"language":"jsx"},"content":[{"type":"text","text":"// ChatInterface.jsx\nimport { useState, useRef, useEffect } from \"react\";\n\nconst API_BASE = process.env.REACT_APP_API_URL;\n\nexport default function ChatInterface() {\n const [messages, setMessages] = useState([]);\n const [input, setInput] = useState(\"\");\n const [loading, setLoading] = useState(false);\n const chatEndRef = useRef(null);\n\n // Auto-scroll to latest message\n useEffect(() => {\n chatEndRef.current?.scrollIntoView({ behavior: \"smooth\" });\n }, [messages]);\n\n const sendMessage = async () => {\n if (!input.trim()) return;\n \n const userMessage = { role: \"user\", content: input };\n setMessages(prev => [...prev, userMessage]);\n setInput(\"\");\n setLoading(true);\n\n try {\n const res = await fetch(`${API_BASE}/query`, {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ question: input })\n });\n \n const data = await res.json();\n const aiMessage = { \n role: \"assistant\", \n content: data.answer,\n sources: data.sources \n };\n setMessages(prev => [...prev, aiMessage]);\n } catch (err) {\n console.error(\"API error:\", err);\n } finally {\n setLoading(false);\n }\n };\n\n return (\n

\n {messages.map((msg, i) => (\n

{msg.content}

\n {msg.sources && (\n Sources: {msg.sources.join(\", \")}\n )}\n

\n ))}\n {loading &&

Thinking...

}\n

\n \n

\n setInput(e.target.value)}\n onKeyPress={e => e.key === \"Enter\" && sendMessage()}\n placeholder=\"Ask anything...\"\n />\n \n

\n );\n}\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"8. CI/CD Pipeline & Docker/Kubernetes Deployment"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/etmpthab1rbrpj4u6jqzhsu-qt-kfuxz19rcfiqxco1uffi5hkascdbumd_ji3p078-mpgw1yosvjzvt7sg1xkzq6iyztszxrwug","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Pipeline Overview"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Stage"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Tool / Service"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Purpose"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Source Control"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"GitHub"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Version control, PR reviews"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Build"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"GitHub Actions"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Docker image build & test"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Image Registry"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS ECR"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Store and version Docker images"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Compute"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS ECS / EKS with Fargate"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Run containerized services"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Orchestration"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Kubernetes"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Auto-scaling, rolling updates"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Expose"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS API Gateway"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Route external traffic to containers"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Definitions"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS ECR (Elastic Container Registry):"},{"type":"text","text":" A fully managed Docker container registry for storing, versioning, and deploying container images."}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS ECS (Elastic Container Service):"},{"type":"text","text":" A container orchestration service that runs Docker containers on managed cluster instances."}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS Fargate:"},{"type":"text","text":" A serverless compute engine for containers that abstracts away server provisioning — you only define CPU/memory requirements."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Dockerfile"}]},{"type":"codeBlock","attrs":{"language":"dockerfile"},"content":[{"type":"text","text":"# Dockerfile\nFROM python:3.11-slim\n\nWORKDIR /app\n\n# Install dependencies\nCOPY requirements.txt .\nRUN pip install --no-cache-dir -r requirements.txt\n\n# Copy application code\nCOPY . .\n\n# Expose FastAPI port\nEXPOSE 8000\n\n# Run with uvicorn\nCMD [\"uvicorn\", \"main:app\", \"--host\", \"0.0.0.0\", \"--port\", \"8000\"]\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"GitHub Actions CI/CD Workflow"}]},{"type":"codeBlock","attrs":{"language":"yaml"},"content":[{"type":"text","text":"# .github/workflows/deploy.yml\nname: Build and Deploy to AWS ECS\n\non:\n push:\n branches: [main]\n\nenv:\n AWS_REGION: us-east-1\n ECR_REPOSITORY: rag-ai-agent\n ECS_SERVICE: rag-agent-service\n ECS_CLUSTER: rag-agent-cluster\n CONTAINER_NAME: rag-agent\n\njobs:\n deploy:\n runs-on: ubuntu-latest\n \n steps:\n - name: Checkout code\n uses: actions/checkout@v3\n\n - name: Configure AWS credentials\n uses: aws-actions/configure-aws-credentials@v2\n with:\n aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}\n aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}\n aws-region: ${{ env.AWS_REGION }}\n\n - name: Login to Amazon ECR\n id: login-ecr\n uses: aws-actions/amazon-ecr-login@v1\n\n - name: Build, tag, and push image to ECR\n id: build-image\n env:\n ECR_REGISTRY: ${{ steps.login-ecr.outputs.registry }}\n IMAGE_TAG: ${{ github.sha }}\n run: |\n docker build -t $ECR_REGISTRY/$ECR_REPOSITORY:$IMAGE_TAG .\n docker push $ECR_REGISTRY/$ECR_REPOSITORY:$IMAGE_TAG\n echo \"image=$ECR_REGISTRY/$ECR_REPOSITORY:$IMAGE_TAG\" >> $GITHUB_OUTPUT\n\n - name: Deploy to ECS\n uses: aws-actions/amazon-ecs-deploy-task-definition@v1\n with:\n task-definition: task-definition.json\n service: ${{ env.ECS_SERVICE }}\n cluster: ${{ env.ECS_CLUSTER }}\n wait-for-service-stability: true\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Kubernetes Deployment (Optional Scaling)"}]},{"type":"codeBlock","attrs":{"language":"yaml"},"content":[{"type":"text","text":"# k8s/deployment.yaml\napiVersion: apps/v1\nkind: Deployment\nmetadata:\n name: rag-agent\nspec:\n replicas: 3\n selector:\n matchLabels:\n app: rag-agent\n template:\n metadata:\n labels:\n app: rag-agent\n spec:\n containers:\n - name: rag-agent\n image: /rag-ai-agent:latest\n ports:\n - containerPort: 8000\n env:\n - name: AWS_REGION\n value: us-east-1\n resources:\n requests:\n memory: \"512Mi\"\n cpu: \"250m\"\n limits:\n memory: \"1Gi\"\n cpu: \"500m\"\n---\napiVersion: v1\nkind: Service\nmetadata:\n name: rag-agent-svc\nspec:\n selector:\n app: rag-agent\n ports:\n - port: 80\n targetPort: 8000\n type: LoadBalancer\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Load Balancing"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"AWS Elastic Load Balancer (ELB) distributes incoming traffic across all running container instances, ensuring:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"No single point of failure"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Horizontal scaling to handle thousands of concurrent users"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Health checks to route traffic only to healthy containers"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"9. Responsible AI & Guardrails"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"AWS Bedrock Guardrails"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Guardrails in Bedrock allow you to define policies that are applied at the model level to prevent unwanted or harmful outputs."}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"You can configure:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Content filters"},{"type":"text","text":" – Block toxic, violent, or explicit content"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Topic denial"},{"type":"text","text":" – Prevent the model from discussing certain topics"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Sensitive information redaction"},{"type":"text","text":" – Automatically mask PII (names, emails, phone numbers)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Grounding checks"},{"type":"text","text":" – Verify responses are consistent with provided context"}]}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Applying guardrails when invoking a model\nimport boto3\n\nbedrock_runtime = boto3.client('bedrock-runtime', region_name='us-east-1')\n\nresponse = bedrock_runtime.invoke_model(\n modelId='amazon.titan-text-express-v1',\n body=json.dumps({\"inputText\": user_query, ...}),\n guardrailIdentifier='your-guardrail-id',\n guardrailVersion='DRAFT',\n trace='ENABLED'\n)\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Secrets Management"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Never hardcode API keys or credentials. Use "},{"type":"text","marks":[{"type":"bold"}],"text":"AWS Secrets Manager"},{"type":"text","text":":"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"import boto3\nfrom botocore.exceptions import ClientError\n\ndef get_secret(secret_name: str) -> dict:\n \"\"\"Retrieve a secret from AWS Secrets Manager.\"\"\"\n client = boto3.client('secretsmanager', region_name='us-east-1')\n \n try:\n response = client.get_secret_value(SecretId=secret_name)\n return json.loads(response['SecretString'])\n except ClientError as e:\n raise Exception(f\"Failed to retrieve secret: {e}\")\n\n# Usage\nsecrets = get_secret(\"rag-agent/production\")\nmongo_uri = secrets['MONGODB_URI']\ndeepgram_key = secrets['DEEPGRAM_API_KEY']\nelevenlabs_key = secrets['ELEVENLABS_API_KEY']\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"10. End-to-End Architecture Flow"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":" ┌─────────────────────────────────┐\n │ USER INTERFACE │\n │ React Frontend (Browser) │\n │ Text input + Voice microphone │\n └──────────────┬──────────────────┘\n │ HTTPS / WebSocket\n ▼\n ┌─────────────────────────────────┐\n │ AWS API Gateway │\n │ Route & authenticate requests │\n └──────────────┬──────────────────┘\n │\n ┌────────────────┴────────────────┐\n │ │\n ▼ ▼\n ┌─────────────────┐ ┌──────────────────┐\n │ AWS Lambda │ │ ECS / Fargate │\n │ (Serverless) │ │ (Containerized) │\n │ FastAPI App │ │ Pipecat Voice │\n └────────┬────────┘ └────────┬─────────┘\n │ │\n ▼ ▼\n ┌─────────────────────┐ ┌────────────────────────┐\n │ AWS Bedrock │ │ Deepgram STT │\n │ (LLM: Titan / │◄────────►│ ElevenLabs TTS │\n │ Llama / Claude) │ │ (Voice Pipeline) │\n └──────────┬──────────┘ └────────────────────────┘\n │\n ▼\n ┌─────────────────────┐\n │ MongoDB Atlas │\n │ Vector Database │\n │ (RAG Retrieval) │\n └─────────────────────┘\n │\n ▼\n ┌─────────────────────┐\n │ Amazon S3 │\n │ (Documents, Logs, │\n │ Model Artifacts) │\n └─────────────────────┘\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Complete Request Lifecycle"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"User sends a query"},{"type":"text","text":" via the React UI (text or voice)"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"API Gateway"},{"type":"text","text":" authenticates the request and routes it to Lambda/ECS"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"FastAPI"},{"type":"text","text":" receives the request, invokes the RAG pipeline"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Query is embedded"},{"type":"text","text":" using Amazon Titan Embeddings"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"MongoDB Atlas"},{"type":"text","text":" performs similarity search, returns top-K relevant chunks"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Augmented prompt"},{"type":"text","text":" (context + question) is sent to Bedrock LLM"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"LLM generates"},{"type":"text","text":" a grounded answer"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"For "},{"type":"text","marks":[{"type":"bold"}],"text":"voice responses"},{"type":"text","text":": answer is sent to ElevenLabs TTS → audio returned to frontend"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"CI/CD"},{"type":"text","text":" ensures every code push automatically builds, tests, and deploys via GitHub Actions → ECR → ECS/Fargate"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Guardrails"},{"type":"text","text":" are applied at the Bedrock level before any response is returned to the user"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"11. Troubleshooting Common Issues"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Symptom"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Likely Cause"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Remedy"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"\"many token\""},{"type":"text","text":" error"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Exceeded model token limit or account plan restriction"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Reduce prompt length; verify free-tier credit is active"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Model not visible in catalog"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Region mismatch or credit still pending"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Switch to a region where provider is enabled (e.g., "},{"type":"text","marks":[{"type":"code"}],"text":"ap-south-1"},{"type":"text","text":" Mumbai)"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bedrock access denied (IAM)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Logged in as IAM user lacking permissions"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Re-login with Root user or attach required Bedrock policies to IAM role"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"API calls failing"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Missing or expired API token"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Regenerate long-term token from Bedrock console"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Payment declined on sign-up"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Card doesn't support small authorizations"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Use a different credit/debit card with sufficient balance"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"No free-tier models appear"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Account not fully provisioned"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Wait up to 12 hours after sign-up, then refresh the Model Catalog"}]}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"Key Takeaways"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"AWS Bedrock"},{"type":"text","text":" provides a unified platform for text, image, and audio foundation models without managing any ML infrastructure."}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"RAG"},{"type":"text","text":" is the essential pattern for grounding LLM responses in your own data — combining a vector store (MongoDB Atlas) with a generative model (Bedrock)."}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Pipecat"},{"type":"text","text":" is the central orchestrator for voice AI agents; it wires together Deepgram (STT), any LLM, and ElevenLabs (TTS) into a real-time pipeline."}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"FastAPI + AWS Lambda"},{"type":"text","text":" provides a clean serverless backend layer that's easy to test locally and scales effortlessly in production."}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"CI/CD with GitHub Actions → Docker → ECR → ECS/Fargate"},{"type":"text","text":" ensures reproducible, automated deployments to production."}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Guardrails and Secrets Manager"},{"type":"text","text":" are non-negotiable for responsible AI deployment — never skip them in production."}]}]}]},{"type":"horizontalRule"},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"italic"}],"text":"This blog covers the complete technical stack for building a production-grade RAG AI agent on AWS. Each section builds on the previous, taking you from zero to a fully deployed, voice-capable, retrieval-augmented AI system."}]}]}