<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Dineshrajdhanapathy on Medium]]></title>
        <description><![CDATA[Stories by Dineshrajdhanapathy on Medium]]></description>
        <link>https://medium.com/@dineshrajdhanapathy?source=rss-2d8438cdd725------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*i70k4Q06XTfrH3CZLrHuXg.jpeg</url>
            <title>Stories by Dineshrajdhanapathy on Medium</title>
            <link>https://medium.com/@dineshrajdhanapathy?source=rss-2d8438cdd725------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 06:59:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@dineshrajdhanapathy/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Generative AI Fundamentals: From FMs to AWS Applications]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/generative-ai-fundamentals-from-fms-to-aws-applications-6bfb781a36a5?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/6bfb781a36a5</guid>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[amazon-web-services]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sun, 11 Oct 2026 12:47:07 GMT</pubDate>
            <atom:updated>2026-10-11T12:47:07.039Z</atom:updated>
            <content:encoded><![CDATA[<p>Explore foundation models, transformers, RAG, prompt engineering, responsible AI, and how AWS services like Amazon Bedrock enable generative AI applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q3QtXFgxdk4wFyfcxCBiaQ.png" /></figure><h3>Introduction</h3><p>Generative AI has evolved from being a research topic into a core technology powering chatbots, coding assistants, document analysis, image generation, enterprise search, and autonomous AI agents.</p><p>Services like ChatGPT, Amazon Q Developer, Amazon Kiro, Claude, and Amazon Nova are all built upon <strong>Foundation Models (FMs)</strong> large AI models trained on enormous datasets capable of performing many different tasks without being designed specifically for each one.</p><p>In this article, you’ll learn:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/824/0*wsmkeobKrQ0eEMey" /></figure><h3>Why Traditional Machine Learning Isn’t Enough</h3><p>Traditional ML follows a straightforward pipeline:</p><pre>Collect Data<br>      │<br>      ▼<br>Label Data<br>      │<br>      ▼<br>Train Model<br>      │<br>      ▼<br>   Deploy</pre><p>Every new task requires:</p><ul><li>New labeled datasets</li><li>New model training</li><li>New deployment</li></ul><p>For example:</p><ul><li>Spam detection</li><li>Image classification</li><li>Fraud detection</li></ul><p>Each needs its own model. Generative AI changes this entirely.</p><h3>Foundation Models</h3><p>A Foundation Model is a massive neural network trained once on trillions of tokens.</p><p>Instead of solving one task, it learns:</p><ul><li>Language</li><li>Programming</li><li>Mathematics</li><li>Reasoning</li><li>Images</li><li>Audio</li><li>General world knowledge</li></ul><p>Later, the same model can perform:</p><ul><li>Question answering</li><li>Summarization</li><li>Translation</li><li>Coding</li><li>Document analysis</li><li>Chatbots</li><li>Content creation</li></ul><p>Examples available through <strong>Amazon Bedrock</strong> include:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/775/1*qD6F6Ge5UQSvUnGQ_d2mYg.png" /></figure><h3>How Foundation Models Are Trained</h3><p>Pretraining uses enormous amounts of <strong>unlabeled data</strong>.</p><pre>Books<br>Wikipedia<br>Research Papers<br>GitHub<br>Web Pages<br>Images<br>Videos<br>Audio</pre><p>↓</p><p>Transformer Training</p><p>↓</p><p>Foundation Model</p><p>No one labels every sentence manually. Instead, the model learns by predicting missing words over billions of examples.</p><h3>Understanding Tokens</h3><p>Large Language Models don’t read text.</p><p>They read <strong>tokens</strong>.</p><p>Example:</p><pre>Amazon Bedrock is awesome.</pre><p>might become</p><pre>Amazon<br>Bed<br>rock<br>is<br>awesome<br>.</pre><p>Each token becomes an integer.</p><pre>1023<br>194<br>556<br>92<br>812<br>9</pre><p>Pricing in most LLM APIs is also token-based.</p><p>That means:</p><ul><li>Longer prompts cost more</li><li>Longer outputs cost more</li></ul><p>Prompt optimization reduces latency and cost.</p><h3>Word Embeddings</h3><p>After tokenization, tokens become vectors.</p><p>Instead of storing words directly:</p><pre>Dog</pre><p>becomes</p><pre>[0.34,<br>-0.81,<br>0.91,<br>...<br>]</pre><p>Words with similar meanings occupy nearby positions in vector space.</p><pre>Dog -------- Puppy<br><br><br>Cat -------- Kitten<br><br><br>King ------ Queen</pre><p>This semantic representation enables reasoning beyond simple keyword matching.</p><h3>Transformers: The Real Breakthrough</h3><p>Before 2017, AI relied heavily on recurrent neural networks (RNNs), which processed words one by one.</p><pre>Word 1<br>↓<br><br><br>Word 2<br>↓<br><br><br>Word 3</pre><p>Transformers process every word simultaneously.</p><pre>Entire Sentence<br>        │<br>        ▼<br>Self Attention</pre><p>Advantages include:</p><ul><li>Parallel GPU execution</li><li>Faster training</li><li>Better context retention</li><li>Scales to trillions of parameters</li></ul><p>This architecture powers:</p><ul><li>GPT</li><li>Claude</li><li>Nova</li><li>Gemini</li><li>Llama</li><li>DeepSeek</li></ul><h3>Self-Attention</h3><p>Consider:</p><p><em>The programmer fixed the bug because </em><strong><em>it</em></strong><em> caused the crash.</em></p><p>What does “it” mean?</p><ul><li>Programmer?</li><li>Bug?</li><li>Crash?</li></ul><p>Self-attention evaluates relationships between every word.</p><p>Plain text</p><p>Code language</p><pre>Programmer ←→ Bug<br>Programmer ←→ Crash<br>Bug ←→ Crash</pre><p>The model determines that “it” refers to “bug.”</p><p>This contextual understanding is a key reason transformers generate coherent responses.</p><h3>From Prompt to Response</h3><p>Suppose the prompt is:</p><p><em>Write a Python Lambda function.</em></p><p>The processing flow is:</p><pre>Prompt<br><br><br>↓<br><br><br>Tokenization<br><br><br>↓<br><br><br>Embedding<br><br><br>↓<br><br><br>Transformer Layers<br><br><br>↓<br><br><br>Probability Distribution<br><br><br>↓<br><br><br>Decoder<br><br><br>↓<br><br><br>Generated Code</pre><p>At each step, the model predicts the most probable next token.</p><h3>Context Windows</h3><p>Large Language Models maintain temporary memory called the <strong>context window</strong>.</p><p>Example:</p><pre>User:<br><br><br>My name is Alice.<br><br><br>Assistant:<br>Hello Alice.<br><br><br>User:<br>What&#39;s my name?</pre><p>The model answers correctly because the earlier message remains within the current context.</p><p>Once the context window fills, older information may be discarded.</p><p>This is why long conversations can lose earlier details.</p><h3>Prompt Engineering</h3><p>Prompt engineering is often the easiest way to improve output quality.</p><p>Instead of:</p><pre>Explain AWS.</pre><p>Use:</p><pre>You are an AWS Solutions Architect.<br><br><br>Explain Amazon VPC to beginners.<br><br><br>Include:<br>- diagrams<br>- real-world examples<br>- interview questions<br>- best practices<br><br><br>Keep it under 700 words.</pre><p>Clear prompts consistently produce better results.</p><h3>Fine-Tuning</h3><p>Sometimes prompts aren’t enough.</p><p>Organizations may need:</p><ul><li>Medical terminology</li><li>Banking policies</li><li>Legal documents</li><li>Internal knowledge</li></ul><p>Fine-tuning updates model parameters using domain-specific labeled datasets.</p><p>Advantages:</p><ul><li>Better accuracy</li><li>Consistent responses</li><li>Specialized expertise</li></ul><p>Trade-offs:</p><ul><li>High cost</li><li>Longer training</li><li>Ongoing maintenance</li></ul><h3>Retrieval-Augmented Generation (RAG)</h3><p>Fine-tuning isn’t always necessary. Instead, many organizations use RAG.</p><pre>User Question<br><br><br>↓<br><br><br>Vector Search<br><br><br>↓<br><br><br>Relevant Documents<br><br><br>↓<br><br><br>Foundation Model<br><br><br>↓<br><br><br>Answer</pre><p>Example:</p><p>Employee asks:</p><p><em>What’s our leave policy?</em></p><p>Instead of relying on pretrained knowledge, the application retrieves the latest HR policy document and sends it to the model as context.</p><p>Benefits:</p><ul><li>Up-to-date answers</li><li>Uses private data</li><li>No retraining</li><li>Lower cost</li></ul><p>AWS services commonly used:</p><ul><li>Amazon Bedrock</li><li>Amazon OpenSearch Serverless</li><li>Amazon S3</li><li>Amazon Knowledge Bases</li><li>AWS Lambda</li></ul><h3>Building with Amazon Bedrock</h3><p>Typical architecture:</p><pre>                  User<br>                    │<br>                    ▼<br>             React / Web App<br>                    │<br>              Amazon API Gateway<br>                    │<br>                AWS Lambda<br>                    │<br>            Amazon Bedrock API<br>                    │<br>          Claude / Nova / Llama<br>                    │<br>                Response</pre><p>If using RAG:</p><pre>User<br><br><br>↓<br><br><br>API Gateway<br><br><br>↓<br><br><br>Lambda<br><br><br>↓<br><br><br>Knowledge Base<br><br><br>↓<br><br><br>Amazon OpenSearch Vector Store<br><br><br>↓<br><br><br>Amazon Bedrock<br><br><br>↓<br><br><br>Answer</pre><h3>Sample Python Code (Amazon Bedrock)</h3><pre>import boto3<br>import json<br><br>client = boto3.client(&quot;bedrock-runtime&quot;)<br><br>response = client.invoke_model(<br>    modelId=&quot;amazon.nova-lite-v1:0&quot;,<br>    body=json.dumps({<br>        &quot;messages&quot;: [<br>            {<br>                &quot;role&quot;: &quot;user&quot;,<br>                &quot;content&quot;: [<br>                    {<br>                        &quot;text&quot;: &quot;Explain Amazon EC2 in simple words.&quot;<br>                    }<br>                ]<br>            }<br>        ]<br>    })<br>)<br><br>print(response[&quot;body&quot;].read().decode())</pre><p>With only a few lines of code, you can invoke a foundation model hosted on Amazon Bedrock without managing infrastructure.</p><h3>Responsible AI</h3><p>Generative AI introduces new challenges.</p><p>Common risks include:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/799/1*XlMe_uzguNevemO9wGNnNg.png" /></figure><p>Mitigation strategies:</p><ul><li>Human review</li><li>Content filtering</li><li>Amazon Bedrock Guardrails</li><li>Prompt validation</li><li>Retrieval-Augmented Generation</li><li>Continuous monitoring</li></ul><p><strong>AWS Services for Generative AI</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/776/1*MeL0ASL-jf92hatTKGufEA.png" /></figure><h3>Best Practices</h3><ul><li>Start with a business problem, not the model.</li><li>Use prompt engineering before considering fine-tuning.</li><li>Prefer RAG for proprietary knowledge instead of retraining.</li><li>Monitor token usage to control costs.</li><li>Build with human oversight for high-impact decisions.</li><li>Apply guardrails, logging, and evaluation throughout the AI lifecycle.</li><li>Regularly review compliance, privacy, and responsible AI practices.</li></ul><h3>Conclusion</h3><p>Generative AI has fundamentally changed how we build intelligent applications. Powered by Foundation Models and transformer architectures, these systems can understand language, generate code, summarize documents, and answer complex questions with remarkable fluency.</p><p>For most organizations, the fastest path to production is not building or training a model from scratch, but combining <strong>Amazon Bedrock</strong>, <strong>Knowledge Bases</strong>, <strong>Guardrails</strong>, and <strong>AWS serverless services</strong> to create secure, scalable, and cost-effective AI solutions. By pairing strong technical design with responsible AI practices, you can deliver applications that are not only innovative but also trustworthy, maintainable, and ready for enterprise use.</p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments; I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud, AI and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️ <a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍 <a href="https://builder.aws.com/community/@dd1"><strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6bfb781a36a5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[TurfArena : The Operating System for Local Sports Communities]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/turfarena-the-operating-system-for-local-sports-communities-fa1aac088479?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/fa1aac088479</guid>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[dynamodb]]></category>
            <category><![CDATA[vercel]]></category>
            <category><![CDATA[sports]]></category>
            <category><![CDATA[v0]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sun, 28 Jun 2026 18:50:25 GMT</pubDate>
            <atom:updated>2026-06-28T18:51:37.264Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>TurfArena</strong> is a modern full-stack sports venue booking platform that connects players, teams, tournament organizers, and venue owners through a fast, scalable, and AI-enhanced experience. It simplifies discovering sports facilities, booking available slots, managing tournaments, and engaging with the local sports community all from a single application.</p><p><strong>Live Demo:</strong> <a href="https://turf-arena-gilt.vercel.app/">turf-arena-gilt.vercel.app</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5Y9aZ0kHyB1BfwArWty06Q.jpeg" /></figure><h3>Why TurfArena?</h3><p>Finding and booking sports venues is often fragmented, requiring multiple phone calls, messaging apps, or manual scheduling. Venue owners also struggle with managing bookings, cancellations, and customer communication.</p><p>TurfArena addresses these challenges with:</p><ul><li>Real-time venue discovery</li><li>Instant online booking</li><li>AI-powered coaching assistant</li><li>Tournament management</li><li>Community engagement</li><li>Venue owner dashboard</li><li>Scalable cloud-native architecture</li></ul><h3>The Problem We’re Solving</h3><p>Every weekend across India, thousands of turf grounds host football, cricket, basketball, and badminton matches. But behind the scenes, it’s chaos:</p><ul><li>Tournaments managed via <strong>WhatsApp groups</strong> and Google Sheets</li><li>No way for players to <strong>track performance</strong> across seasons</li><li>Turf owners handling bookings through <strong>phone calls</strong> and paper registers</li><li>Double-bookings are common two teams show up for the same slot</li><li>No centralized <strong>rankings or statistics</strong> for local sports communities</li></ul><p>There’s no “operating system” for this ecosystem. We built one.</p><h3>What is TurfArena?</h3><p>TurfArena is a production-deployed platform with <strong>22 API endpoints</strong>, <strong>44 pages</strong>, and <strong>9 DynamoDB tables</strong> serving four distinct user roles:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/863/1*udv3VDZ-DLjFJ3B29Uo4Bg.png" /></figure><p>Each role gets a dedicated dashboard with role-specific navigation, data, and actions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A4-MWjZfP97fDpk2Hd0yeg.png" /></figure><h3>Key Features</h3><h3>1. Real-Time Turf Booking with Double-Booking Prevention</h3><p>The core booking challenge: what happens when two people try to book the same slot at the exact same moment?</p><p>We solved this with <strong>Valkey</strong> (Redis-compatible cache) using atomic locking:</p><pre>User clicks &quot;Book 7 PM slot&quot;<br>  → API checks Valkey: EXISTS slot:turf1:2026-06-25:19:00<br>  → If free: SET slot:turf1:2026-06-25:19:00 NX EX 300<br>  → Only ONE user gets the lock (NX = &quot;set if Not eXists&quot;)<br>  → Lock holder&#39;s booking is written to DynamoDB<br>  → Slot marked as booked for 24 hours<br>  → Second user gets &quot;423 Slot Locked&quot; response</pre><p>This prevents race conditions that DynamoDB alone cannot handle at the millisecond level.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8NpYpti2Vn6Qk-G7FPiyrQ.png" /><figcaption>Turf booking flow with calendar date picker, time slot selection, and real-time availability</figcaption></figure><h3>👤 Player Experience</h3><ul><li>Search nearby sports venues</li><li>Browse available slots</li><li>Book courts instantly</li><li>View booking history</li><li>Favorite venues</li><li>GPS navigation</li><li>Live match schedules</li></ul><h3>2. AI Coach (Amazon Bedrock — Nova Micro)</h3><p>Every player gets a personal AI coach that analyzes their <strong>real statistics</strong> from DynamoDB:</p><p><strong>How it works:</strong></p><ol><li>Player requests coaching tips</li><li>API queries TurfArena_PlayerStats table (composite key: playerId + sport)</li><li>Builds a context prompt with actual numbers: wins, losses, goals, MVP awards</li><li>Sends to Amazon Bedrock Nova Micro (us.amazon.nova-micro-v1:0)</li><li>Returns personalized recommendations</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y3eIeGmt2g7npBgrYwICzA.png" /><figcaption>AI-generated performance insights based on real player statistics from DynamoDB</figcaption></figure><p><strong>AI capabilities:</strong></p><ul><li>Performance analysis (strengths, weaknesses, weekly goals)</li><li>Match predictions (win probability from real stats)</li><li>Team builder (optimal formations, chemistry scoring)</li><li>Auto-commentary (narrative from live match events)</li><li>Strategy suggestions (counter-tactics based on opponent data)</li></ul><p>The AI isn’t generic it’s grounded in the player’s actual performance data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*grMUtgnohej55UN6TnNyeg.png" /><figcaption>AI Coach providing personalized training recommendations and match predictions</figcaption></figure><h3>🤖 AI Features</h3><p>Powered by Amazon Bedrock.</p><ul><li>AI Sports Coach</li><li>Training recommendations</li><li>Match insights</li><li>Personalized suggestions</li><li>Natural language assistance</li></ul><h3>3. For Turf Owners</h3><ul><li>Manage turf bookings with real-time slot availability (Valkey-powered)</li><li>Track revenue, occupancy rates, and expenses</li><li>Receive booking confirmations via EventBridge events</li><li>View both live DynamoDB bookings and historical data</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5Qy2zFNgSyYZkBy0DXfXgg.png" /><figcaption>Owner dashboard showing real bookings (green DB badge) alongside historical data</figcaption></figure><h3>🏢 Venue Owner Dashboard</h3><ul><li>Manage venues</li><li>Configure pricing</li><li>Monitor bookings</li><li>View occupancy analytics</li><li>Handle cancellations</li><li>Tournament management</li></ul><h3>4. Tournament Lifecycle Management</h3><p>Full tournament flow from creation to results:</p><ol><li><strong>Create</strong> : Organizer sets format (knockout/league/group), entry fee, prize pool</li><li><strong>Registration</strong> : Teams register via 3-step wizard (team info → roster → payment review)</li><li><strong>Brackets</strong> : Auto-generated based on registered teams</li><li><strong>Live Matches</strong> : Real-time score updates with event timeline</li><li><strong>Results</strong> : Final standings, player stats updated, leaderboards recalculated</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IkVXXaCNR-uz4qfqo5l0Nw.png" /><figcaption>Organizer dashboard with tournament analytics, revenue metrics, and team registrations from DynamoDB</figcaption></figure><p><strong>DynamoDB operations per registration:</strong></p><ul><li>PutItem → Registrations table (team details)</li><li>UpdateItem → Tournaments table (atomic teamsJoined + 1)</li><li>EventBridge → team.registered event (for notifications)</li></ul><p>The atomic counter ensures accurate team counts even under concurrent registrations.</p><h3>🏆 Tournament Management</h3><ul><li>Create tournaments</li><li>Team registration</li><li>Fixture generation</li><li>Match scheduling</li><li>Live scores</li><li>Leaderboards</li></ul><h3>5. GPS-Based Location Features</h3><ul><li><strong>Nearby turfs</strong> : Find turfs within X km using Haversine formula</li><li><strong>200m geofencing</strong> : Players can only check in to a match if GPS confirms they’re within 200 meters of the venue</li><li><strong>Live tracking</strong> : Store player positions during match (foundation for future heatmaps)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/907/1*nFr67JJFJs1VpcD8ynVKQQ.png" /><figcaption>Interactive map showing turf location with distance indicator</figcaption></figure><h3>6. Community &amp; Social</h3><ul><li>Post match results, achievements, and highlights</li><li>Like and comment on posts</li><li>Achievement badges (MVP, Hat-trick, Clean Sheet, 50 Goals)</li><li>Leaderboards with weekly/monthly/all-time rankings</li></ul><h3>Architecture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V7H4pvBWGI3lMSTtaNhmyA.gif" /><figcaption>Full system architecture</figcaption></figure><h3>Architecture connection diagram</h3><ul><li>DynamoDB box shows which API endpoint writes to which table</li><li>Bedrock has a dashed line back to PlayerStats showing it reads stats before generating AI</li><li>Legend box explains the 6-step data flow order</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uTOnO5A69UtwhgB1xIbJkA.gif" /></figure><h3>Why This Architecture?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ijxLMh5qzLFtW2zv3R9k9w.png" /></figure><h3>Why Amazon DynamoDB?</h3><p>Amazon DynamoDB was selected as the primary backend because it provides:</p><ul><li>Single-digit millisecond latency</li><li>Automatic scaling</li><li>High availability</li><li>Flexible schema</li><li>Serverless operation</li><li>Excellent performance for booking workloads</li></ul><h3>Key entities include:</h3><ul><li>Users</li><li>Venues</li><li>Bookings</li><li>Tournaments</li><li>Teams</li><li>Reviews</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xHpxYZgeCfE9KLUBlAOLpA.png" /><figcaption>AWS Console showing all 9 TurfArena tables with active status</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YWKAB5ZjNQsOIVLZOAZQiw.png" /><figcaption>DynamoDB real-data store</figcaption></figure><h3>DynamoDB Design</h3><p>9 Tables with Purpose-Built Access Patterns</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gLdCDg1EBazvQy033eAyWg.png" /></figure><h3>Advanced DynamoDB Features Used</h3><ul><li><strong>Atomic counters</strong> : SET teamsJoined = teamsJoined + :inc (no race conditions)</li><li><strong>List append</strong> : list_append(events, :newEvent) (add match events without overwriting)</li><li><strong>Composite keys</strong> : PlayerStats uses playerId + sport for per-sport queries</li><li><strong>Conditional writes</strong> : Tournament capacity check before registration</li><li><strong>7 GSIs</strong> : Efficient queries without table scans</li></ul><h3>Why Vercel?</h3><p>Vercel enables:</p><ul><li>Global edge deployment</li><li>Fast page loads</li><li>Seamless CI/CD</li><li>Serverless API execution</li><li>Optimized Next.js performance</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M7JdI8I6t7e1TEug-Lej6A.png" /><figcaption>Deploy to Vercel</figcaption></figure><h3>Deployed on Vercel</h3><p>TurfArena is deployed on <a href="https://vercel.com/">Vercel</a> with full GitHub integration every push to main triggers an automatic production build and deployment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yxX3bht0gV8TFC4ssClC1g.png" /></figure><h3>OBSERVABILITY</h3><ul><li>Vercel Analytics, Structured logging, CloudWatch</li><li>logging patterns and HTTP error codes for each</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SQCS3xlKr3uvYIi3SMuTiQ.png" /></figure><h3>Valkey</h3><ul><li>Valkey checks if slot is already booked (instant, &lt;1ms)</li><li>Real-time booking slot locking prevents double-bookings</li><li>Writes to DynamoDB</li><li>Marks slot permanently booked in Valkey cache</li><li>Fires EventBridge notification</li></ul><h3>scalability</h3><p>TurfArena is designed with cloud-native principles.</p><p>The architecture supports:</p><ul><li>Thousands of concurrent users</li><li>Event-driven processing</li><li>AI-assisted workflows</li><li>Horizontal scaling</li><li>Serverless backend services</li></ul><h3>Future enhancements may include:</h3><ul><li>Online payments</li><li>Real-time notifications</li><li>Live venue availability</li><li>Multi-city support</li><li>International expansion</li><li>Mobile applications</li></ul><h3>Business Model</h3><h3>Potential revenue streams include:</h3><ul><li>Booking commissions</li><li>Premium memberships</li><li>Venue subscriptions</li><li>Tournament hosting fees</li><li>Sponsored listings</li><li>Advertising partnerships</li></ul><h3>Event-Driven (Amazon EventBridge)</h3><p>Amazon EventBridge acts as a central event bus. When something important happens in the app (booking confirmed, score updated, team registered), the API route publishes an event to the TurfArena-Events bus.</p><p>Every significant action publishes an event to EventBridge:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/886/1*cwXm81mPspvUbxehxEFhhA.png" /></figure><p>EventBridge is a decoupled event bus. Events are published successfully, but without rules + targets, nothing reacts to them. Adding new consumers (Lambda, SQS, SNS) requires <strong>zero code changes</strong> just add an EventBridge rule.</p><h3>Tech Stack</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ltUeMguQrgapV8an25BnRg.png" /></figure><h3>Challenges &amp; Solutions</h3><h3>Double-Booking Race Condition</h3><p><strong>Problem:</strong> Two users booking same slot within milliseconds both succeed. <strong>Solution:</strong> Valkey SET NX EX 300 atomic lock that only one request can acquire.</p><h3>Turbopack Build Failure</h3><p><strong>Problem:</strong> Space in directory path (D:\github work\) broke Next.js Turbopack. <strong>Solution:</strong> next build --webpack flag bypasses Turbopack entirely.</p><h3>Silent DynamoDB Fallback</h3><p><strong>Problem:</strong> Vercel’s built-in DynamoDB integration overrode env vars with empty strings app silently used mock data. <strong>Solution:</strong> Removed integration, set env vars via CLI. Added AWS_ENABLED feature flag for explicit control.</p><h3>Dark Theme Contrast</h3><p><strong>Problem:</strong> text-gray-600 on bg-white became invisible with dark theme. <strong>Solution:</strong> Audited all 44 pages, replaced with theme-aware CSS variables.</p><h3>Real-Time Without WebSocket</h3><p><strong>Problem:</strong> DynamoDB doesn’t support push. Live scores needed instant updates. <strong>Solution:</strong> EventBridge for event bus + client-side polling (5s). WebSocket planned for Phase 2.</p><h3>Cost</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/703/1*uv5UPAyV8FjfyXiYioDdzw.png" /></figure><p>The entire production platform runs at zero cost for development and demos.</p><h3>What’s Next</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/668/1*9KSnvJBgwMmQK_ZJq1YmKg.png" /></figure><h3>Future Enhancements</h3><p>While TurfArena already provides a complete sports venue booking experience, there are several opportunities to further enhance the platform and support larger communities.</p><p>Future improvements include:</p><ul><li><strong>Secure Online Payments</strong> with multiple payment gateways for seamless booking transactions.</li><li><strong>Real-Time Notifications</strong> using AWS event-driven services for booking confirmations, reminders, and tournament updates.</li><li><strong>Live Venue Availability</strong> with instant slot updates and dynamic scheduling.</li><li><strong>Advanced AI Features</strong> such as personalized training plans, match analysis, smart venue recommendations, and chatbot assistance powered by Amazon Bedrock.</li><li><strong>Analytics Dashboard</strong> for venue owners to monitor bookings, revenue, customer engagement, and business performance.</li><li><strong>Multi-City and Global Expansion</strong> with support for multiple regions, languages, currencies, and time zones.</li><li><strong>Native Mobile Applications</strong> for Android and iOS to provide a seamless mobile experience.</li><li><strong>Enhanced Tournament Ecosystem</strong> including automated fixture generation, live scoring, team rankings, and event management.</li><li><strong>Community Features</strong> such as player profiles, social interactions, team formation, reviews, and achievement badges.</li><li><strong>Enterprise-Scale Architecture</strong> with additional AWS services, caching, monitoring, and global content delivery to support millions of users while maintaining high availability and performance.</li></ul><p>These enhancements will transform TurfArena from a booking platform into a comprehensive digital ecosystem for sports players, venue owners, and tournament organizers.</p><h3>Conclusion</h3><p>TurfArena demonstrates how modern cloud-native technologies can transform the sports venue booking experience into a scalable, intelligent, and user-friendly platform. By combining <strong>Vercel</strong> for high-performance frontend deployment with <strong>Amazon DynamoDB</strong> as the primary backend database, the application delivers fast, reliable, and scalable booking management while supporting real-world business needs.</p><p>Beyond venue reservations, TurfArena introduces AI-powered coaching, tournament management, community engagement, and venue owner tools, making it more than just a booking application. Its serverless architecture, event-driven design, and modular implementation provide a strong foundation for future enhancements such as secure online payments, live notifications, analytics, personalized recommendations, and global expansion.</p><p>Built for the <strong>H0: Hack the Zero Stack Hackathon</strong>, TurfArena showcases thoughtful full-stack engineering, practical AWS database integration, and a production-ready architecture capable of supporting thousands of users. It reflects our vision of making sports more accessible, organized, and connected through innovative cloud technologies while delivering a platform that is ready to grow into a sustainable and impactful business.</p><h3>Try It</h3><pre># Clone and run locally (no AWS needed — uses mock data)<br>git clone https://github.com/dineshrajdhanapathyDD/TurfArena.git<br>cd TurfArena<br>npm install<br>npm run dev<br># Open http://localhost:3000<br><br># Login as different roles:<br># Player:    customer@turf.com / customer123<br># Organizer: organizer@turf.com / organizer123<br># Owner:     owner@turf.com / owner123</pre><pre># Connect to AWS (optional)<br># Set AWS_REGION, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY in .env.local<br>npm run aws:init   # Creates 9 tables + seeds demo data<br>npm run dev        # Now connected to real DynamoDB</pre><p>Built for the <strong>H0: Hack the Zero Stack with Vercel v0 and AWS Databases</strong> hackathon.</p><h3>Judging Criteria Alignment</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/813/1*1VIf6G28PG_Lgar1fj9fPQ.png" /></figure><h3>Technologies &amp; Official Documentation:</h3><h3>Frontend &amp; Backend</h3><ul><li><strong>Next.js</strong> — <a href="https://nextjs.org/docs">https://nextjs.org/docs</a></li><li><strong>TypeScript</strong> — <a href="https://www.typescriptlang.org/docs">https://www.typescriptlang.org/docs</a></li><li><strong>Next.js API Routes</strong> — <a href="https://nextjs.org/docs/app/building-your-application/routing/route-handlers">https://nextjs.org/docs/app/building-your-application/routing/route-handlers</a></li><li><strong>Node.js</strong> — <a href="https://nodejs.org/docs/latest/api/">https://nodejs.org/docs/latest/api/</a></li><li>V0 — <a href="https://v0.app/docs">What is v0? | v0 Docs</a></li></ul><h3>AWS Cloud Services</h3><ul><li><strong>Amazon DynamoDB</strong> — <a href="https://docs.aws.amazon.com/dynamodb/">https://docs.aws.amazon.com/dynamodb/</a></li><li><strong>Amazon Bedrock</strong> — <a href="https://docs.aws.amazon.com/bedrock/">https://docs.aws.amazon.com/bedrock/</a></li><li><strong>Amazon EventBridge</strong> — <a href="https://docs.aws.amazon.com/eventbridge/">https://docs.aws.amazon.com/eventbridge/</a></li><li><strong>AWS SDK for JavaScript v3</strong> — <a href="https://docs.aws.amazon.com/sdk-for-javascript/v3/developer-guide/">https://docs.aws.amazon.com/sdk-for-javascript/v3/developer-guide/</a></li></ul><h3>Deployment</h3><ul><li><strong>Vercel</strong> — <a href="https://vercel.com/docs">https://vercel.com/docs</a></li></ul><h3>Additional technologies</h3><ul><li><strong>Valkey</strong> — <a href="https://valkey.io/">https://valkey.io/</a></li><li><strong>AWS CLI</strong> — <a href="https://docs.aws.amazon.com/cli/">https://docs.aws.amazon.com/cli/</a></li><li><strong>Vercel CLI</strong> — <a href="https://vercel.com/docs/cli">https://vercel.com/docs/cli</a></li></ul><p><strong>GitHub:</strong> <a href="https://github.com/dineshrajdhanapathyDD/TurfArena">github.com/dineshrajdhanapathyDD/TurfArena</a></p><p>Built for the <strong>#H0Hackathon</strong></p><p><strong>H0: Hack the Zero Stack with Vercel v0 and AWS Databases</strong> hackathon.</p><p><strong>Vercel v0</strong> was used to generate the initial UI components describe what you want, v0 generates production-ready React/Next.js code with Tailwind CSS, and deploy to Vercel in one click.</p><p><em>Built with ❤️ TurfArena — where every player has a profile, every match has a story, and every turf has a platform.</em></p><h3>Team</h3><ul><li><strong>Dineshraj Dhanapathy — </strong><a href="https://medium.com/@dineshrajdhanapathy">Dineshrajdhanapathy — Medium</a></li><li><strong>Abinaya SV — </strong><a href="https://medium.com/@abinayasv">Abinaya SV — Medium</a></li><li><strong>Mariyam Usmani</strong></li></ul><p><em>Developed by</em> <a href="https://kiro.dev/">Kiro</a> <em>— AI-powered development environment.</em></p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments; I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud, AI and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️<a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa1aac088479" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an AWS Study Buddy with MCP + Strands Agents SDK]]></title>
            <link>https://aws.plainenglish.io/building-an-aws-study-buddy-with-mcp-strands-agents-sdk-aa81aeefcdce?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa81aeefcdce</guid>
            <category><![CDATA[strands-agents]]></category>
            <category><![CDATA[mcp-server]]></category>
            <category><![CDATA[amazon-bedrock]]></category>
            <category><![CDATA[amazon-web-services]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Fri, 05 Jun 2026 18:01:07 GMT</pubDate>
            <atom:updated>2026-06-06T05:09:11.598Z</atom:updated>
            <content:encoded><![CDATA[<p>An AI-powered learning assistant that can search AWS documentation, explain cloud concepts in simple terms, track learning progress, and generate quizzes for better understanding.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yaZ0abzQpHzdJI6TRgccNg.png" /></figure><h3>📖 Introduction</h3><p>What if you had a personal AWS tutor that could search official documentation, explain services in simple terms, remember what you’ve studied, and quiz you on it later?</p><p>That’s exactly what I built for an <strong>AWS Study Buddy</strong> powered by the Model Context Protocol (MCP) and Amazon Nova Pro.</p><p>In this blog, I’ll walk you through the entire journey from concept to working demo.</p><h3>🎯 The Problem</h3><p>Learning AWS is overwhelming. There are 200+ services, thousands of documentation pages, and it’s hard to track what you’ve already studied. Most chatbots just answer questions — they don’t help you <em>learn</em> systematically.</p><p>I wanted an agent that:</p><ol><li>Searches <strong>real, official</strong> AWS documentation (not hallucinated answers)</li><li><strong>Remembers</strong> what I’ve studied across sessions</li><li>Can <strong>quiz me</strong> to reinforce learning</li><li>Does <strong>cost calculations</strong> for pricing questions</li><li>Streams responses in <strong>real time</strong></li></ol><h3>🧠 The Solution: AWS Study Buddy</h3><p>An interactive AI agent that combines:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/943/1*BXudHLaPUTUALIuhvQbCSg.png" /></figure><h3>🏗️ Architecture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*F0CpV7kEIvE-hwQy" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ul-51LOdpc4L7fncymsDyQ.png" /></figure><h3>🔧 Tech Stack</h3><ul><li><strong>Strands Agents SDK</strong> — The agent framework</li><li><strong>Amazon Bedrock</strong> — Nova Pro v1 as the LLM</li><li><strong>MCP (Model Context Protocol)</strong> — Connects to the AWS Documentation server</li><li><strong>Python</strong> — The glue that holds it all together</li><li><strong>JSON</strong> — Simple persistent memory (no cloud DB needed)</li></ul><h3>📝 Step-by-Step Build Process</h3><h3>Step 1: Setting Up the MCP Connection</h3><p>The Model Context Protocol lets our agent talk to external tools via a standardized interface. The AWS Documentation MCP server provides tools to search and read official AWS docs.</p><pre>from strands.tools.mcp import MCPClient <br>from mcp import StdioServerParameters, stdio_client <br> <br>aws_docs_mcp = MCPClient( <br>    lambda: stdio_client( <br>        StdioServerParameters( <br>            command=&quot;uvx&quot;, <br>            args=[&quot;awslabs.aws-documentation-mcp-server@latest&quot;], <br>            env={**os.environ, &quot;FASTMCP_LOG_LEVEL&quot;: &quot;ERROR&quot;} <br>        ) <br>    ) <br>)</pre><p>This starts the MCP server as a subprocess and communicates via stdio. The uvx command downloads and runs the server without needing a separate install step.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MJtU1FpUfVZHDgy-W5jgDg.png" /></figure><h3>Step 2: Building Custom Tools</h3><p>I created 4 custom tools using the @tool decorator:</p><p><strong>save_note</strong> — Stores a study summary with timestamp:</p><pre>@tool <br>def save_note(topic: str, summary: str) -&gt; str: <br>    &quot;&quot;&quot;Save a study note about an AWS topic.&quot;&quot;&quot; <br>    notes = _load_notes() <br>    notes[&quot;notes&quot;][topic.lower()] = { <br>        &quot;summary&quot;: summary, <br>        &quot;studied_at&quot;: datetime.now().strftime(&quot;%Y-%m-%d %H:%M&quot;) <br>    } <br>    _save_notes(notes) <br>    return f&quot;📝 Note saved for &#39;{topic}&#39;!&quot;</pre><p><strong>get_notes</strong> — Retrieves past study notes:</p><pre>@tool <br>def get_notes(topic: str) -&gt; str: <br>    &quot;&quot;&quot;Retrieve study notes. Use &#39;all&#39; to see everything.&quot;&quot;&quot; <br>    notes = _load_notes() <br>    if topic.lower() == &quot;all&quot;: <br>        return f&quot;📚 You&#39;ve studied {len(notes[&#39;topics&#39;])} topics...&quot;</pre><p><strong>record_quiz_score</strong> — Tracks quiz performance:</p><pre>@tool <br>def record_quiz_score(topic: str, score: int, total: int) -&gt; str: <br>    &quot;&quot;&quot;Record quiz results for progress tracking.&quot;&quot;&quot;</pre><p><strong>calculator</strong> — For cost estimates:</p><pre>@tool <br>def calculator(expression: str) -&gt; str: <br>    &quot;&quot;&quot;Evaluate math expressions for pricing calculations.&quot;&quot;&quot;</pre><h3>Step 3: Streaming Callback</h3><p>The callback handler prints text as it arrives — no waiting for the full response:</p><pre>def streaming_callback(**kwargs): <br>    if &quot;data&quot; in kwargs: <br>        print(kwargs[&quot;data&quot;], end=&quot;&quot;, flush=True) <br>    if &quot;current_tool_use&quot; in kwargs and kwargs[&quot;current_tool_use&quot;].get(&quot;name&quot;): <br>        print(f&quot;\n🔧 Using tool: {kwargs[&#39;current_tool_use&#39;][&#39;name&#39;]}&quot;)</pre><p><strong>Key insight:</strong> The callback uses **kwargs (keyword arguments), not positional args. The SDK passes data, current_tool_use, init_event_loop, etc., as keyword arguments.</p><h3>Step 4: The Agent System Prompt</h3><p>The system prompt is crucial — it tells the agent <em>how</em> to behave as a tutor:</p><pre>system_prompt = &quot;&quot;&quot;You are AWS Study Buddy 📚🤖 <br> <br>Your behavior: <br>1. When asked about AWS, SEARCH the docs first, then explain <br>2. After explaining, ALWAYS save a note using save_note <br>3. When asked to quiz, create questions from saved notes <br>4. Be encouraging, use emojis, make learning fun! <br>&quot;&quot;&quot;</pre><h3>Step 5: Combining MCP + Custom Tools</h3><p>The magic happens when we combine MCP tools with custom tools:</p><pre>with aws_docs_mcp: <br>    mcp_tools = aws_docs_mcp.list_tools_sync() <br>    all_tools = mcp_tools + [save_note, get_notes, record_quiz_score, calculator] <br>     <br>    agent = Agent( <br>        model=model, <br>        tools=all_tools, <br>        callback_handler=streaming_callback <br>    )</pre><p>The agent now has access to both AWS documentation search AND persistent memory — it can search docs, explain, save notes, and quiz, all in one conversation.</p><h3>Step 6: Interactive Chat Loop</h3><pre>while True: <br>    user_input = input(&quot;\n🧑 You: &quot;).strip() <br>    if user_input.lower() in (&quot;quit&quot;, &quot;exit&quot;): <br>        break <br>    print(&quot;\n🤖 Study Buddy: &quot;, end=&quot;&quot;, flush=True) <br>    response = agent(user_input) <br>    print()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nukpK7ws4NsPpPp0NMAqRg.png" /></figure><h3>🎬 Demo: A Real Conversation</h3><pre>🧑 You: Explain S3 in simple terms <br> <br>🤖 Study Buddy: <br>🔧 Using tool: search_documentation <br>🔧 Using tool: read_documentation <br> <br>Amazon S3 (Simple Storage Service) is object storage! 🪣 <br> <br>Key points: <br>• Store unlimited data as &quot;objects&quot; in &quot;buckets&quot; <br>• 99.999999999% (11 nines) durability <br>• Use cases: backups, static websites, data lakes <br>• Storage classes: Standard, IA, Glacier <br> <br>🔧 Using tool: save_note <br>📝 Note saved for &#39;S3&#39;! You&#39;ve now studied 1 topic.</pre><pre>🧑 You: How much for 100GB in S3 Standard? <br> <br>🤖 Study Buddy: <br>🔧 Using tool: calculator <br> <br>S3 Standard: $0.023/GB/month <br>100 × $0.023 = $2.30/month 💰</pre><pre>🧑 You: Quiz me on S3 <br> <br>🤖 Study Buddy: <br>🔧 Using tool: get_notes <br> <br>Q1: How many nines of durability does S3 have? <br>Q2: What are objects stored inside? <br>Q3: Name 2 storage classes.</pre><pre>🧑 You: 11 nines, buckets, Standard and Glacier <br> <br>🤖 Study Buddy: <br>🔧 Using tool: record_quiz_score <br>🌟 Excellent! 3/3 (100%) on S3!</pre><h3>🧩 Challenges I Faced</h3><h3>1. mem0ai Version Incompatibility</h3><p>Initially, I tried using mem0_memory from strands-agents-tools for persistent memory. But mem0ai version 1.x removed FAISS and aws_bedrock support, breaking everything.</p><p><strong>Solution:</strong> Replaced with a simple JSON file-based memory. It’s simpler, always works, and has zero dependencies.</p><h3>2. Streaming Callback Signature</h3><p>The callback handler signature changed across versions. Using def callback(event) or def callback(event, **kwargs) Both failed.</p><p><strong>Solution:</strong> Use def callback(**kwargs) - The SDK passes everything as keyword arguments.</p><h3>3. MCP Server Startup</h3><p>The MCP server needs uvx (from the uv package manager) to run without pre-installation.</p><p><strong>Solution:</strong> Install uv with pip install uv, then the script uses uvx to auto-download and run the MCP server.</p><h3>💡 Key Learnings</h3><ol><li><strong>MCP is powerful</strong> — It gives agents access to real, structured data sources without hallucination</li><li><strong>Simple memory beats complex memory</strong> — A JSON file is more reliable than fighting version conflicts</li><li><strong>System prompts matter</strong> — The agent’s behaviour is 90% determined by how you prompt it</li><li><strong>Streaming makes UX great</strong> — Real-time output feels much more responsive than waiting</li><li><strong>Tool chaining is magic</strong> — The agent naturally chains search → explain → save in one turn</li></ol><h3>📊 Judging Criteria</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/924/1*QSbbEGXsdXS-tWpK4YyLyA.png" /></figure><h3>🚀 Future Improvements</h3><ul><li>Add more MCP servers (CDK, Cost Analysis) for a multi-MCP super-agent</li><li>Spaced repetition algorithm for quiz scheduling</li><li>Export study notes as markdown files</li><li>Add a web UI with Flask/Streamlit</li><li>Connect to Bedrock Knowledge Bases for custom documentation</li></ul><h3>Conclusion</h3><p>Building an AI agent with MCP + Strands SDK is surprisingly straightforward. The hardest part wasn’t the code it was getting the right package versions installed! Once that’s sorted, the SDK handles tool routing, streaming, and conversation management beautifully.</p><p>The Model Context Protocol is a game-changer for AI agents. Instead of hoping the LLM knows the answer, you can connect it to authoritative data sources. For AWS documentation, this means accurate, up-to-date answers every time.</p><p>If you’re learning AWS, try building your own study buddy. The act of building the agent teaches you more about AWS than using it! 🎓</p><p>GitHub link: <a href="https://github.com/dineshrajdhanapathyDD/Building-AI-Agents-from-Zero-to-Hero/tree/main/challenge-5-mcp-agent">AWS Study Buddy with MCP</a></p><h3>📎 Resources</h3><ul><li><a href="https://strandsagents.com/">Strands Agents SDK</a></li><li><a href="https://github.com/awslabs/mcp">AWS Documentation MCP Server</a></li><li><a href="https://modelcontextprotocol.io/">Model Context Protocol</a></li><li><a href="https://docs.aws.amazon.com/bedrock/">Amazon Bedrock</a></li></ul><p><em>Built with ❤️ by DD</em></p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments; I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud, AI and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️<a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa81aeefcdce" width="1" height="1" alt=""><hr><p><a href="https://aws.plainenglish.io/building-an-aws-study-buddy-with-mcp-strands-agents-sdk-aa81aeefcdce">Building an AWS Study Buddy with MCP + Strands Agents SDK</a> was originally published in <a href="https://aws.plainenglish.io/">AWS in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Serenova AI Mental Health Companion]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/serenova-ai-mental-health-companion-81f103429356?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/81f103429356</guid>
            <category><![CDATA[mental-health]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[amazon-nova-sonic]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Thu, 19 Mar 2026 06:52:50 GMT</pubDate>
            <atom:updated>2026-03-19T06:52:50.899Z</atom:updated>
            <content:encoded><![CDATA[<p>A deep dive into building a real-time, speech-to-text AI companion for mental health support using AWS Bedrock and Amazon Nova models.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F-YX0F_3QCDk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D-YX0F_3QCDk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F-YX0F_3QCDk%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/b9fe9c2145e06c312b7eb58f042b76b8/href">https://medium.com/media/b9fe9c2145e06c312b7eb58f042b76b8/href</a></iframe><h3>A Story of Technology, Empathy, and AWS</h3><h3>The Beginning</h3><p>Imagine a moment late at night.</p><p>Someone is feeling overwhelmed, anxious, or alone.<br>They want to talk to someone — but no one is available.</p><p>Therapy appointments take weeks.<br>Friends may be busy.<br>And sometimes, people just need someone to <strong>listen without judgment</strong>.</p><p>This is where <strong>Serenova</strong> was born.</p><p>Serenova is an <strong>AI-powered mental health companion</strong> designed to provide <strong>emotional support, guided conversations, and calming interactions</strong> anytime, anywhere.</p><p>And the entire experience is powered by <strong>AWS and Amazon Nova AI models</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fBlRBmG5VDQz-zhGkEZ1Zg.png" /></figure><h3>The Vision</h3><p>Our vision was simple:</p><blockquote><em>Build an AI companion that feels </em><strong><em>safe, responsive, and human-like</em></strong><em>.</em></blockquote><p>Serenova helps users:</p><ul><li>Share their thoughts</li><li>Reduce stress</li><li>Practice breathing exercises</li><li>Receive supportive AI responses</li><li>Detect emotional distress early</li></ul><p>But to achieve this, we needed <strong>powerful AI and scalable infrastructure</strong>.</p><p>So we built Serenova on <strong>AWS Cloud</strong>.</p><h3>Introduction</h3><p>What if you could talk to an AI that actually listens not just to your words, but to the emotion in your voice?</p><p>That’s the idea behind Serenova. We built a voice-first mental health companion powered by Amazon Nova models on AWS Bedrock. It combines real-time speech-to-speech conversation (Nova Sonic), intelligent text chat (Nova Pro), and a calming UI designed to make users feel safe.</p><h3>The Problem</h3><p>Mental health support has a massive accessibility gap. Therapy is expensive, waitlists are long, and many people don’t feel comfortable reaching out. Text-based chatbots exist, but they miss something fundamental that the human voice carries: emotion that text can’t capture. The trembling, the pauses, the pace of speech all tell a story.</p><p>We wanted to build something that meets people where they are: a companion that’s always available, feels warm and safe, and can understand not just what you say but how you say it.</p><p>This post walks through the technical decisions, architecture, and lessons learned from building and deploying Serenova.</p><h3>How Serenova Works (Powered by AWS)</h3><h3>AI Brain — Amazon Bedrock</h3><p>At the heart of Serenova is <strong>Amazon Bedrock</strong>.</p><p>We use <strong>Amazon Nova models</strong> to power intelligent conversations.</p><h3>Why Amazon Nova?</h3><p>We chose Amazon Nova models for three reasons:</p><h3>Nova Sonic — Speech-to-Speech</h3><p>Traditional voice AI requires three separate services: speech-to-text, language model, text-to-speech. Each adds latency and loses context. Nova Sonic (amazon.nova-sonic-v1:0) handles the entire pipeline in one model - user speaks, AI responds with voice, under 500ms. It also exposes audio features (pitch, volume, pace) that we use for emotion detection.</p><h3>Nova Pro — Text Chat Fallback</h3><p>Not every user has a microphone, and not every environment supports voice. Nova Pro (amazon.nova-pro-v1:0) provides intelligent text chat via the Bedrock Converse API. We use it as the primary model in our production deployment on App Runner.</p><h3>Nova Lite — Fast Responses</h3><p>For quick, lightweight interactions where speed matters more than depth, Nova Lite (amazon.nova-lite-v1:0) handles fast text responses.</p><p>ModelModel IDLatencyUse CaseNova Sonicamazon.nova-sonic-v1:0&lt;500msVoice conversationNova Proamazon.nova-pro-v1:0~300msText chat, reasoningNova Liteamazon.nova-lite-v1:0~150msQuick responses</p><h3>Architecture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/625/0*dSxl0lbQYqJmAb6J" /></figure><h3>Backend Intelligence AWS App Runner</h3><p>The Serenova backend runs on <strong>AWS App Runner</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Y54ipgEKaSorNnX0" /></figure><p>Why App Runner?</p><p>Because it allows us to:</p><ul><li>Deploy <strong>FastAPI services easily</strong></li><li>Automatically scale with traffic</li><li>Handle secure API communication</li></ul><p>The backend connects the frontend to <strong>Amazon Bedrock</strong> to generate AI responses.</p><p>React frontend on AWS Amplify talks to a FastAPI backend on App Runner. The backend uses Nova Pro for text chat. Browser SpeechRecognition handles voice-to-text on the client side.</p><h3>Frontend Experience AWS Amplify</h3><p>The Serenova interface is built with <strong>React</strong> and hosted on <strong>AWS Amplify</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*fptW5uEUXRQkjxpv" /></figure><p>Amplify provides:</p><ul><li>Fast global hosting</li><li>CI/CD deployment</li><li>Secure HTTPS access</li></ul><p>Users can simply open the website and immediately start interacting with Serenova.</p><h3>Container Deployment Amazon ECR + Docker</h3><p>To solve deployment challenges, we packaged the backend using <strong>Docker</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*8P2yDUM4ssed5iKN" /></figure><p>The container image is stored in: <strong>Amazon Elastic Container Registry (ECR)</strong></p><p>App Runner then pulls the image directly from ECR and runs the application reliably.</p><p><strong>Local Development:</strong> Full backend with Nova Sonic voice streaming over WebSocket, emotion detection, crisis detection, and MCP tool integration.</p><h3>Three-Tier Graceful Degradation</h3><p>We designed the system to never break:</p><ol><li><strong>Tier 1: Full Voice</strong> — Nova Sonic speech-to-speech, &lt;500ms latency</li><li><strong>Tier 2: Text Chat</strong> — Nova Pro via Converse API, browser SpeechRecognition</li><li><strong>Tier 3: Demo Mode</strong> — friendly message, crisis resources still available</li></ol><p>If Nova Sonic isn’t available, it falls to text. If the backend is down, it enters demo mode. The user always sees a working app.</p><h3>The Voice Orb</h3><h3>Voice Interaction Web Audio + AI</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*bvdJzCcUzh2tCq7W" /></figure><p>Serenova introduces a <strong>Voice Orb</strong>, a glowing animated circle that reacts to the user’s voice.</p><p>When a user speaks:</p><p>1️⃣ Browser captures voice<br>2️⃣ AI analyzes speech patterns<br>3️⃣ Serenova generates a supportive response</p><p>This creates a <strong>natural, human-like conversation experience</strong>.</p><h3>How It Works</h3><pre>// Web Audio API for mic level monitoring<br>const audioContext = new AudioContext();<br>const analyser = audioContext.createAnalyser();<br>analyser.fftSize = 256;<br><br>const source = audioContext.createMediaStreamSource(stream);<br>source.connect(analyser);<br><br>// Read frequency data every frame<br>const dataArray = new Uint8Array(analyser.frequencyBinCount);<br>analyser.getByteFrequencyData(dataArray);<br>const avg = dataArray.reduce((a, b) =\&gt; a + b, 0) / dataArray.length;<br>const level = Math.min(100, Math.round((avg / 128) * 100));</pre><p>The orb scales based on input level, creating a visual feedback loop that makes speaking feel responsive. It has four states:</p><pre>| State      | Animation                       | Meaning           |<br>| ---------- | ------------------------------- | ----------------- |<br>| Idle       | Soft pulse, mic icon            | Ready to listen   |<br>| Listening  | Animated bars, glow intensifies | Recording voice   |<br>| Thinking   | Bouncing dots                   | AI processing     |<br>| Responding | Wave animation                  | AI response ready |</pre><p>The dark theme with glassmorphism and ambient background orbs creates a calm, safe atmosphere important for a mental health app.</p><h3>Support Modes and Prompt Engineering</h3><p>Each support mode has a carefully crafted system prompt that shapes the AI’s personality:</p><pre>mode_prompts = {<br>    &quot;crisis&quot;: &quot;You are a crisis support counselor. Be calm, direct, and prioritize safety.&quot;,<br>    &quot;cbt&quot;: &quot;You are a CBT therapist. Help identify thought patterns and suggest reframing.&quot;,<br>    &quot;regulation&quot;: &quot;You are an emotion regulation coach. Guide breathing and grounding techniques.&quot;,<br>    &quot;companion&quot;: &quot;You are a warm, empathetic mental health companion. Listen actively, validate feelings. Keep responses short (2-3 sentences).&quot;,<br>}</pre><p>The key insight: mental health prompts need to be warm without being patronizing, supportive without being prescriptive. We iterated on these prompts extensively. Short responses (2–3 sentences) work better than long ones — they feel more like a real conversation.</p><h3>Breathing Exercise</h3><p>The Regulation mode automatically activates a guided breathing exercise:</p><ul><li>Inhale: 4 seconds</li><li>Hold: 4 seconds</li><li>Exhale: 6 seconds</li></ul><p>This 4–4–6 pattern is based on clinical breathing techniques. The animated circle with countdown gives users a visual anchor.</p><h3>Crisis Detection</h3><p>Safety is non-negotiable in a mental health app. Serenova monitors every message for crisis indicators:</p><pre>crisis_keywords = [&quot;hurt&quot;, &quot;harm&quot;, &quot;suicide&quot;, &quot;kill&quot;, &quot;die&quot;, &quot;end it&quot;]<br>detected = any(keyword in text.lower() for keyword in crisis_keywords)</pre><p>When detected:</p><ul><li>A red banner appears at the top of the screen</li><li>A modal opens with crisis resources (988 Lifeline, Crisis Text Line, SAMHSA, Veterans Crisis Line, Trevor Project)</li><li>The 🆘 button is always accessible in the header</li></ul><p>This is a keyword-based first layer. The full local backend adds Bedrock AI analysis and emotion-based detection as additional layers.</p><h3>The Deployment Battle: App Runner and Docker</h3><p>This was our biggest technical challenge. We needed to deploy a Python FastAPI backend to AWS App Runner. Sounds simple. It wasn’t.</p><h3>What Failed: Source-Based Deployment</h3><p>App Runner’s Python runtime builds your code in one environment and runs it in another. Pip-installed binaries like uvicorn aren&#39;t in PATH at runtime.</p><p>We tried everything:</p><ul><li>uvicorn main:app → &quot;executable not found in $PATH&quot;</li><li>python3 -m uvicorn main:app → same error</li><li>/usr/local/bin/uvicorn → still failed</li></ul><h3>What Worked: Docker via ECR</h3><p>The solution was building our own Docker image:</p><pre>FROM python:3.11-slim<br>WORKDIR /app<br>COPY backend/requirements.txt .<br>RUN pip install --no-cache-dir --upgrade pip &amp;&amp; \<br>    pip install --no-cache-dir -r requirements.txt<br>COPY backend/main.py .<br>EXPOSE 8080<br>CMD [&quot;uvicorn&quot;, &quot;main:app&quot;, &quot;--host&quot;, &quot;0.0.0.0&quot;, &quot;--port&quot;, &quot;8080&quot;]</pre><p>Inside the Docker container, uvicorn is properly installed and in PATH. We push the image to Amazon ECR, and App Runner pulls it. No PATH issues, no build/run environment mismatch.</p><h3>The Standalone Backend</h3><p>We also created a separate backend/main.py a minimal FastAPI app with zero complex imports. No Nova Sonic SDK dependency, no relative imports, just boto3 for Bedrock. This made the Docker image small and the deployment reliable.</p><h3>Nova Pro Integration: The Converse API</h3><p>The text chat uses Bedrock’s Converse API, which provides a clean conversation interface:</p><pre>response = bedrock_client.converse(<br>    modelId=&quot;amazon.nova-pro-v1:0&quot;,<br>    messages=conversation_history,  # sliding window of 20 messages<br>    inferenceConfig={<br>        &quot;maxTokens&quot;: 300,<br>        &quot;temperature&quot;: 0.7,<br>        &quot;topP&quot;: 0.9,<br>    },<br>)<br>assistant_text = response[&quot;output&quot;][&quot;message&quot;][&quot;content&quot;][0][&quot;text&quot;]</pre><p>We keep a 20-message sliding window per session. This gives the AI enough context to maintain a coherent conversation without unbounded memory growth.</p><p>Temperature 0.7 gives responses that are warm and varied without being unpredictable — important for mental health where consistency builds trust.</p><h3>Lessons Learned</h3><h3>1. Docker solves deployment headaches</h3><p>If your deployment platform has environment quirks, skip the source-based approach and use Docker. You control the entire runtime.</p><h3>2. Browser APIs are powerful</h3><p>SpeechRecognition and Web Audio API gave us voice features with zero backend dependencies. The browser is an underrated platform.</p><h3>3. Graceful degradation is a feature</h3><p>Users should never see a broken app. Design for failure from the start — voice fails? use text. Backend down? demo mode. Always show something useful.</p><h3>4. Keep production simple</h3><p>The standalone backend/main.py with 5 dependencies was the key to a successful deployment. The full backend with Nova Sonic, MCP tools, and emotion detection is for local dev.</p><h3>5. Mental health AI needs careful tone</h3><p>The difference between “I understand you’re feeling sad” and “That sounds really tough. I’m here.” is huge. Short, warm, human-sounding responses work best.</p><h3>What’s Next</h3><ul><li>Full Nova Sonic voice streaming in production (ECS or App Runner with larger image)</li><li>Persistent conversation history with DynamoDB</li><li>Real-time emotion detection from audio features</li><li>Mobile app (React Native)</li><li>Multi-language support</li><li>AI-guided journaling with mood tracking</li><li>Optional therapist dashboard for professional oversight</li></ul><h3>Tech Stack</h3><pre>| Layer | Technology |<br>|------|---------|<br>| AI | Amazon Nova Sonic, Nova Pro, Nova Lite (Bedrock)|<br>| Backend | Python 3.11, FastAPI, Uvicorn, boto3|<br>| Frontend |React 18, CSS3, Web Audio API, SpeechRecognition API |<br>| Hosting | AWS App Runner (Docker/ECR), AWS Amplify |<br>| Infrastructure | AWS CDK, DynamoDB, S3, EventBridge, SNS, CloudWatch |<br>| Container | Docker, Amazon ECR |</pre><h3>Final Thought</h3><p>We built Serenova because we believe AI can make mental health support more accessible, not as a replacement for human therapists, but as a bridge. A 3 AM companion. A judgment-free space to process your thoughts out loud.</p><p>Serenova proves that <strong>AI can be compassionate</strong>.</p><p>By combining:</p><ul><li><strong>Amazon Nova AI</strong></li><li><strong>AWS Cloud infrastructure</strong></li><li><strong>Human-centered design</strong></li></ul><p>We created a system that helps people feel <strong>heard, supported, and less alone</strong>.</p><p>Serenova isn’t just technology. It’s a <strong>companion for the mind</strong>.</p><p>Resources :</p><p><a href="https://docs.aws.amazon.com/nova/">Amazon Nova Documentation</a></p><p><a href="https://docs.aws.amazon.com/apprunner/">AWS App Runner Documentation</a></p><p><a href="https://docs.amplify.aws/">Amplify Documentation — AWS Amplify Gen 2 Documentation</a></p><p><em>Serenova is an AI companion and is not a replacement for professional mental health care. If you or someone you know is in crisis, call 988 or text HOME to 741741.</em></p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments; I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud, AI and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️<a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81f103429356" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an Agentic AI Weather Assistant Using Amazon Bedrock]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/building-an-agentic-ai-weather-assistant-using-amazon-bedrock-4f496da35dcb?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/4f496da35dcb</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[amazon-bedrock]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[amazon-web-services]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Mon, 02 Mar 2026 06:40:23 GMT</pubDate>
            <atom:updated>2026-03-02T06:48:40.110Z</atom:updated>
            <content:encoded><![CDATA[<p>Agentic AI weather assistant built with Amazon Bedrock that plans API calls, fetches real-time weather data, and delivers human-readable forecasts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y4PPTItTBy4SFS3_rbvZAg.png" /></figure><p>In the evolving landscape of artificial intelligence, the shift from traditional “smart encyclopedias” to <strong>Agentic AI</strong> represents a significant leap in capability. While traditional AI systems provide static responses based on training data, agentic systems are designed to <strong>think, plan, and act</strong> to solve complex problems. This article explores the core building blocks of Agentic AI, using a hands-on weather assistant project built with <strong>Amazon Bedrock</strong> and Python as a primary example.</p><p><strong>Understanding Agentic AI</strong></p><p>Agentic AI moves beyond the “Question → Answer” model toward a <strong>“Problem → Plan → Action → Result”</strong> workflow. According to the sources, true agentic systems are defined by three key characteristics:</p><ul><li><strong>Autonomy:</strong> The ability to make independent decisions, such as choosing appropriate tools or interpreting location descriptions without constant human guidance.</li><li><strong>Reactivity:</strong> The capacity to respond to environmental changes, handle errors, and adjust strategies based on real-time results.</li><li><strong>Proactivity:</strong> Taking the initiative to achieve goals through multi-step planning and anticipating user needs</li></ul><p><strong>The Architecture of an AI Agent</strong></p><p>A functional AI agent requires a “brain” for reasoning and “hands” to interact with the world. The sources outline a four-step process for a weather-focused agent:</p><ol><li><strong>User Input:</strong> The user provides a location in any format (e.g., ZIP code, city name, or coordinates).</li><li><strong>AI Planning:</strong> Using a model like <strong>Claude 4.5 Sonnet</strong>, the agent analyzes the input and determines the necessary coordinates and API endpoints.</li><li><strong>Action (API Calls):</strong> The agent executes the plan. For weather data, this involves a two-step process: first calling the <strong>NWS Points API</strong> to find the correct forecast office, then calling the <strong>Forecast API</strong> for the actual data.</li><li><strong>AI Summary:</strong> The agent processes technical JSON data into a human-friendly, readable summary</li></ol><p><strong>Why Amazon Bedrock?</strong></p><p><strong>Amazon Bedrock</strong> serves as a foundation for these systems by providing access to high-performing models like Claude, which excels at complex reasoning and structured data processing. Bedrock offers <strong>managed infrastructure</strong> and <strong>enterprise-grade security</strong>, allowing developers to build agents that are both scalable and cost-effective.</p><p><strong>Real-World Impact and Future Steps</strong></p><p>The patterns used to build a weather assistant are highly transferable to other industries:</p><ul><li><strong>Travel and Logistics:</strong> Agents can check weather, monitor traffic, and automatically adjust itineraries or routes.</li><li><strong>Agriculture:</strong> Monitoring environmental conditions to recommend specific farming actions.</li><li><strong>Emergency Management:</strong> Tracking severe weather to coordinate rapid responses.</li><li><strong>Business Operations:</strong> Automating market analysis, customer support, and risk assessment.</li></ul><p><strong>Development and Implementation</strong></p><p>Building an agent can be approached through different interfaces depending on the end-user:</p><ul><li><strong>Command-Line Interface (CLI):</strong> Ideal for developers and automation, focusing on direct interaction and raw workflow steps.</li><li><strong>Web Application:</strong> Using tools like Streamlit, developers can create interactive experiences that visualize the agent’s “thinking” phases, such as planning, execution, and analysis.</li></ul><p>The sources emphasize that building from scratch using pure Python and the AWS SDK (boto3) without complex framework abstractions allows developers to truly understand the “aha moment” of how an AI agent functions.</p><h3>Environment Setup</h3><h3>🏠 Your Own AWS Account</h3><p>To set up the development environment for the weather agent in your <strong>own AWS account</strong>, you will need to follow a series of steps to configure your local machine and AWS permissions. This process typically takes about <strong>12 minutes</strong>.</p><p><strong>Prerequisites</strong></p><p>Before starting, ensure you have the following:</p><ul><li><strong>Python 3.7+</strong> installed locally.</li><li>An <strong>AWS account</strong> with appropriate administrative or developer permissions.</li><li>The <strong>AWS CLI</strong> installed on your machine.</li></ul><p><strong>Step 1: Development Environment and Python</strong></p><p>You should use a code editor that supports Python syntax, such as <strong>Visual Studio Code</strong> or <strong>PyCharm</strong>. Once your editor is ready, verify your Python installation in the terminal using python --version. On macOS, you may specifically need to use python3 and pip3.</p><p><strong>Step 2: AWS Credential Configuration</strong></p><p>You must provide your agent with access to AWS services using one of two methods</p><ul><li><strong>AWS CLI (Recommended):</strong> Run aws configure In your terminal, enter your <strong>Access Key ID, Secret Access Key</strong>, and set the default region to us-west-2.</li><li><strong>Environment Variables:</strong> Manually export AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_DEFAULT_REGION in your terminal session.</li></ul><p><strong>Step 3: Enable Claude 4.5 Sonnet in Amazon Bedrock</strong></p><p>This is a <strong>critical step</strong>; without it, your agent will not function.</p><ol><li>Log into the <strong>Amazon Bedrock console</strong>.</li><li>Ensure you are in the <strong>us-west-2 (Oregon)</strong> region.</li><li>Navigate to <strong>Model Access</strong> and request access for <strong>Claude 4.5 Sonnet</strong>.</li><li>Ensure your IAM user has the permissions bedrock:InvokeModel, bedrock:Converse, and bedrock:ListFoundationModels</li></ol><p><strong>Step 4: Project Setup and Dependencies</strong></p><p>Create a dedicated workspace for your agent and set up a virtual environment to manage packages:</p><ol><li><strong>Create a directory:</strong> mkdir agentic-ai-workshop &amp;&amp; cd agentic-ai-workshop.</li><li><strong>Initialize a Virtual Environment:</strong> Run python -m venv .venv and activate it. You should see (.venv) it in your terminal prompt once active.</li><li><strong>Install Packages:</strong> Create a requirements.txt file (or install directly) containing the necessary libraries like <strong>boto3</strong>, <strong>requests</strong>, and <strong>streamlit</strong></li></ol><p><strong>Step 5: Verification</strong></p><p>To confirm everything is configured correctly, run aws sts get-caller-identity it in your terminal. This should return your AWS account details without errors. Once verified, you are ready to begin building the agent logic using <strong>boto3</strong> to interface with the model python -c &quot;import boto3, streamlit, requests; print(&#39;✅ All packages installed&#39;)&quot;</p><h3>Troubleshooting Common Issues</h3><ul><li><strong>“AWS credentials not found”</strong>: Run aws configure to enter your credentials or set the required environment variables manually.</li><li><strong>“Model access denied”</strong>: You must specifically request access for <strong>Claude 4.5 Sonnet</strong> in the Amazon Bedrock console and ensure you are working in the <strong>us-west-2 (Oregon)</strong> region.</li><li><strong>“Permission denied” errors</strong>: Verify that your IAM user has the minimum required permissions: bedrock:InvokeModel, bedrock:Converse, and bedrock:ListFoundationModels.</li><li><strong>“Python command not found”</strong>: Ensure Python is installed and added to your system PATH; on some systems, you may need to use python3 and pip3 instead of python and pip.</li><li><strong>Virtual environment issues</strong>: If you encounter execution errors, ensure your virtual environment is active (indicated by (.venv) in your terminal prompt).</li><li><strong>Package installation fails</strong>: Try re-running the installation command for your dependencies in the terminal.</li><li><strong>Invalid or international locations</strong>: The National Weather Service API used by the agent <strong>only covers US locations</strong>; international cities or misspellings may result in failed data retrieval.</li><li><strong>Network and connectivity issues</strong>: The agent may encounter timeouts or connection problems, which require robust error handling or retry logic.</li><li><strong>Data processing errors</strong>: The agent must be able to handle unexpected, incomplete, or inconsistent JSON data returned from the weather APIs.</li></ul><h3>Understanding Our Agent Architecture</h3><p><strong>How the National Weather Service APIs Work</strong></p><p>The National Weather Service (NWS) provides free weather data through a mandatory <strong>two-step API process</strong> because the US is divided into a grid system where different forecast offices cover specific squares.</p><ul><li><strong>Points API</strong>: This is the first step (https://api.weather.gov/points/{lat},{lon}). It takes latitude and longitude coordinates as input and identifies which NWS forecast office covers that specific location, along with the necessary grid coordinates. For example, coordinates for Seattle map to the &quot;SEW&quot; office at grid 124,67.</li><li><strong>Forecast API</strong>: This is the second step (https://api.weather.gov/gridpoints/{office}/{gridX},{gridY}/forecast). It uses the specific office and grid data obtained from the Points API to retrieve the detailed weather forecast in a technical JSON format.</li></ul><p><strong>How Our Weather Agent Works</strong></p><p>The weather assistant functions as a <strong>reasoning engine</strong> that follows a linear path: <strong>User Input → AI Planning → API Calls → AI Summary → Response</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/827/0*xh30YKIsiKSbM0SW" /></figure><p>Instead of hardcoding every possible location, the agent uses Claude 4.5 Sonnet to dynamically figure out what coordinates and API calls are needed based on what the user types.</p><p><strong>Understanding the Flow.</strong></p><p>The flow of the agent is designed to bridge the gap between human language and technical data:</p><ol><li><strong>Analyze</strong>: The AI (Claude) interprets the user’s location description.</li><li><strong>Determine</strong>: It identifies the required coordinates and API endpoints.</li><li><strong>Execute</strong>: Two sequential API calls are made — first to get the office info, then the weather data.</li><li><strong>Translate</strong>: The AI processes the resulting raw data into a friendly summary for the user.</li></ol><p><strong>The 4 Steps in Detail</strong></p><p><strong>Step 1: User Input</strong></p><p>📝The agent is designed to be highly flexible, accepting various location formats without requiring a specific structure. Users can enter:</p><ul><li><strong>City names</strong> (e.g., “Seattle” or “Seattle, WA”).</li><li><strong>ZIP codes</strong> (e.g., “90210”).</li><li><strong>Informal descriptions</strong> (e.g., “downtown Portland”).</li><li><strong>Coordinates</strong> (e.g., “47.6062, -122.3321”).</li></ul><p><strong>Step 2: AI Planning</strong></p><p>🧠In this phase, Claude 4.5 Sonnet acts as the “brain.” It analyzes the input to create a technical strategy:</p><ul><li>It <strong>determines the precise coordinates</strong> for the user’s location.</li><li>It <strong>identifies the correct NWS API endpoints</strong>.</li><li>It <strong>plans the sequence of calls</strong> (knowing it must hit the Points API before the Forecast API).</li></ul><p><strong>Step 3: API Calls</strong></p><p>🔗The agent executes the plan by making two distinct requests:</p><ol><li><strong>Points API Call</strong>: It sends the coordinates to get the forecast office and grid info.</li><li><strong>Forecast API Call</strong>: It uses the URL returned by the first call to fetch the actual, detailed weather data.</li></ol><p><strong>Step 4: AI Summary</strong></p><p>📊The final step involves converting the <strong>raw JSON data</strong> (which is complex and technical) into a <strong>human-friendly format</strong>. Claude filters through the technical details to provide a natural response, such as: <em>“Today: Partly cloudy with a high of 72°F. Wind: West at 5–10 mph”</em></p><h3>Building the Command-Line Agent</h3><p>Constructing a command-line interface (CLI) agent allows you to experience the “aha moment” of agentic AI by building every component from scratch. This approach is ideal for developers because it provides a <strong>direct, focused interaction</strong> and clearly displays the raw workflow steps</p><p><strong>Step 1: Establishing the Amazon Bedrock Connection</strong></p><p>The foundation of the agent is its connection to the “brain” <strong>Claude 4.5 Sonnet</strong></p><ul><li><strong>Implementation:</strong> Create a file named weather_agent_cli.py within your agentic-ai-workshop directory</li></ul><pre># Import necessary libraries<br>import boto3        # AWS SDK for Python - allows us to interact with AWS services<br>import json         # For handling JSON data<br>import subprocess   # For running system commands like curl<br>import time         # For adding delays and timing operations<br>from datetime import datetime  # For timestamps and date operations<br><br>def call_claude_sonnet(prompt):<br>    &quot;&quot;&quot;<br>    This function sends a prompt to Claude 4.5 Sonnet and gets a response.<br>    This is the &quot;brain&quot; of our agent - where the AI thinking happens.<br>    <br>    Args:<br>        prompt (str): The question or instruction we want to send to Claude<br>    <br>    Returns:<br>        tuple: (success: bool, response: str) - success status and Claude&#39;s response or error message<br>    &quot;&quot;&quot;<br>    # Create a connection to Amazon Bedrock service<br>    # Bedrock is AWS&#39;s service for accessing AI models like Claude<br>    bedrock = boto3.client(<br>        service_name=&#39;bedrock-runtime&#39;,  # Specify we want the runtime version for making AI calls<br>        region_name=&#39;us-west-2&#39;          # AWS region - using us-west-2 as specified<br>    )<br>    <br>    try:<br>        # Send our prompt to Claude and get a response<br>        response = bedrock.converse(<br>            # Specify which version of Claude we want to use<br>            modelId=&#39;us.anthropic.claude-sonnet-4-5-20250929-v1:0&#39;,  # Claude 4.5 Sonnet<br>            <br>            # Format our message - Claude expects messages in a specific structure<br>            messages=[<br>                {<br>                    &quot;role&quot;: &quot;user&quot;,                    # We are the user asking a question<br>                    &quot;content&quot;: [{&quot;text&quot;: prompt}]      # Our actual question/prompt<br>                }<br>            ],<br>            <br>            # Configure how Claude should respond<br>            inferenceConfig={<br>                &quot;maxTokens&quot;: 2000,    # Maximum length of response (tokens ≈ words)<br>                &quot;temperature&quot;: 0.7   # Creativity level (0=very focused, 1=very creative)<br>            }<br>        )<br>        <br>        # Extract the actual text response from Claude&#39;s response structure<br>        # The response comes nested in a complex structure, so we dig down to get the text<br>        return True, response[&#39;output&#39;][&#39;message&#39;][&#39;content&#39;][0][&#39;text&#39;]<br>        <br>    except Exception as e:<br>        # If something goes wrong, return an error message<br>        return False, f&quot;Error calling Claude: {str(e)}&quot;<br><br>def execute_curl_command(url):<br>    &quot;&quot;&quot;<br>    Execute a curl command to fetch data from an API.<br>    This is how our agent &quot;acts&quot; in the real world - making HTTP requests.<br>    <br>    Args:<br>        url (str): The URL to fetch data from<br>    <br>    Returns:<br>        tuple: (success: bool, response: str) - success status and API response or error message<br>    &quot;&quot;&quot;<br>    try:<br>        # Use curl command to make HTTP request<br>        # curl is a command-line tool for making HTTP requests<br>        result = subprocess.run(<br>            [&#39;curl&#39;, &#39;-s&#39;, url],  # -s flag makes curl silent (no progress info)<br>            capture_output=True,   # Capture the output so we can process it<br>            text=True,            # Return output as text (not bytes)<br>            timeout=30            # Give up after 30 seconds<br>        )<br>        <br>        # Check if the command was successful<br>        if result.returncode == 0:<br>            return True, result.stdout<br>        else:<br>            return False, f&quot;Curl command failed: {result.stderr}&quot;<br>            <br>    except subprocess.TimeoutExpired:<br>        return False, &quot;Request timed out after 30 seconds&quot;<br>    except Exception as e:<br>        return False, f&quot;Error executing curl: {str(e)}&quot;<br><br>def generate_weather_api_calls(location):<br>    &quot;&quot;&quot;<br>    Use Claude to intelligently generate National Weather Service API calls for a given location.<br>    This is where the &quot;agentic&quot; magic happens - AI generating the API call URL.<br>    <br>    Args:<br>        location (str): The location provided by the user<br>    <br>    Returns:<br>        tuple: (success: bool, api_calls: list) - success status and list of API URLs or error message<br>    &quot;&quot;&quot;<br>    # Create a detailed prompt that teaches Claude how to generate NWS API calls<br>    prompt = f&quot;&quot;&quot;<br>You are an expert at working with the National Weather Service (NWS) API.<br><br>Your task: Generate the NWS API URL to get weather forecast data for &quot;{location}&quot;.<br><br>Instructions:<br>1. First, determine the approximate latitude and longitude coordinates for this location<br>2. Generate the NWS Points API URL: https://api.weather.gov/points/{{lat}},{{lon}}<br><br>For the coordinates, use your knowledge to estimate:<br>- Major cities: Use well-known coordinates<br>- ZIP codes: Estimate based on the area<br>- States: Use approximate center coordinates<br>- In case a location description is provided instead of a location name, please use the most likely city and state name as the location for the coordinates<br><br>Example for Seattle:<br>https://api.weather.gov/points/47.6062,-122.3321<br><br>Example for largest city in USA:<br>Based on your knowledge, you will establish location is New York City<br>https://api.weather.gov/points/40.7128,-74.0060<br><br>Now generate the API call (Points API) for the established location. <br>Return ONLY the complete Points API URL, nothing else.<br>Format: https://api.weather.gov/points/LAT,LON<br>&quot;&quot;&quot;<br>    <br>    print(f&quot;🧠 AI is analyzing &#39;{location}&#39; and generating weather API calls...&quot;)<br>    success, response = call_claude_sonnet(prompt)<br>    <br>    if success:<br>        # Clean up the response - sometimes Claude adds extra text<br>        api_url = response.strip()<br>        # Make sure we got a valid URL<br>        if api_url.startswith(&#39;https://api.weather.gov/points/&#39;):<br>            return True, [api_url]  # Return as list for consistency<br>        else:<br>            return False, f&quot;AI generated invalid URL: {api_url}&quot;<br>    else:<br>        return False, response<br><br>def get_forecast_url_from_points_response(points_json):<br>    &quot;&quot;&quot;<br>    Extract the forecast URL from the NWS Points API response.<br>    <br>    Args:<br>        points_json (str): JSON response from the Points API<br>    <br>    Returns:<br>        tuple: (success: bool, forecast_url: str) - success status and forecast URL or error message<br>    &quot;&quot;&quot;<br>    try:<br>        data = json.loads(points_json)<br>        forecast_url = data[&#39;properties&#39;][&#39;forecast&#39;]<br>        return True, forecast_url<br>    except (json.JSONDecodeError, KeyError) as e:<br>        return False, f&quot;Error parsing Points API response: {str(e)}&quot;<br><br>def process_weather_response(raw_json, location):<br>    &quot;&quot;&quot;<br>    Use Claude to convert raw NWS API JSON into a human-readable weather summary.<br>    This is where AI processes complex data into useful information.<br>    <br>    Args:<br>        raw_json (str): Raw JSON response from NWS API<br>        location (str): Original location for context<br>    <br>    Returns:<br>        tuple: (success: bool, summary: str) - success status and processed summary or error message<br>    &quot;&quot;&quot;<br>    prompt = f&quot;&quot;&quot;<br>You are a weather information specialist. I have raw National Weather Service forecast data for &quot;{location}&quot; that needs to be converted into a clear, helpful summary for a general audience.<br><br>Raw NWS API Response:<br>{raw_json}<br><br>Please create a weather summary that includes:<br>1. A brief introduction with the location<br>2. Current conditions and today&#39;s forecast<br>3. The next 2-3 days outlook with key details (temperature, precipitation, wind)<br>4. Any notable weather patterns or alerts<br>5. Format the response to be easy to read and understand<br><br>Make it informative and practical for someone planning their activities. Focus on being helpful and clear.<br>&quot;&quot;&quot;<br>    <br>    print(f&quot;📊 AI is processing weather data and creating summary...&quot;)<br>    success, response = call_claude_sonnet(prompt)<br>    <br>    return success, response<br><br>def run_weather_agent():<br>    &quot;&quot;&quot;<br>    Main function that orchestrates our AI agent.<br>    This demonstrates the complete agentic workflow.<br>    &quot;&quot;&quot;<br>    print(&quot;🌤️ Welcome to the Weather AI Agent!&quot;)<br>    print(&quot;This agent uses Claude 4.5 Sonnet to help you get weather forecasts.&quot;)<br>    print(&quot;=&quot; * 60)<br>    <br>    while True:<br>        # Get user input<br>        location = input(&quot;\n🔍 Enter a location name or description (or &#39;quit&#39; to exit): &quot;).strip()<br>        <br>        if location.lower() in [&#39;quit&#39;, &#39;exit&#39;, &#39;q&#39;]:<br>            print(&quot;👋 Thanks for using the Weather Agent!&quot;)<br>            break<br>            <br>        if not location:<br>            print(&quot;❌ Please enter a location name or description.&quot;)<br>            continue<br>            <br>        print(f&quot;\n🚀 Starting weather analysis for &#39;{location}&#39;...&quot;)<br>        print(&quot;-&quot; * 40)<br>        <br>        # Step 1: AI generates the Points API URL<br>        print(&quot;Step 1: 🧠 AI Planning Phase&quot;)<br>        success, api_calls = generate_weather_api_calls(location)<br>        <br>        if not success:<br>            print(f&quot;❌ Failed to generate API calls: {api_calls}&quot;)<br>            continue<br>            <br>        points_url = api_calls[0]<br>        print(f&quot;✅ Generated Points API URL: {points_url}&quot;)<br>        <br>        # Step 2: Execute the Points API call<br>        print(&quot;\nStep 2: 🔗 Points API Execution&quot;)<br>        print(&quot;Fetching location data from National Weather Service...&quot;)<br>        success, points_response = execute_curl_command(points_url)<br>        <br>        if not success:<br>            print(f&quot;❌ Failed to fetch points data: {points_response}&quot;)<br>            continue<br>            <br>        print(f&quot;✅ Received points data&quot;)<br>        <br>        # Step 3: Extract forecast URL from Points response<br>        print(&quot;\nStep 3: 📍 Extracting Forecast URL&quot;)<br>        success, forecast_url = get_forecast_url_from_points_response(points_response)<br>        <br>        if not success:<br>            print(f&quot;❌ Failed to extract forecast URL: {forecast_url}&quot;)<br>            continue<br>            <br>        print(f&quot;✅ Forecast URL: {forecast_url[:60]}...&quot;)<br>        <br>        # Step 4: Execute the Forecast API call<br>        print(&quot;\nStep 4: 🌦️ Forecast API Execution&quot;)<br>        print(&quot;Fetching weather forecast data...&quot;)<br>        success, forecast_response = execute_curl_command(forecast_url)<br>        <br>        if not success:<br>            print(f&quot;❌ Failed to fetch forecast data: {forecast_response}&quot;)<br>            continue<br>            <br>        print(f&quot;✅ Received {len(forecast_response)} characters of forecast data&quot;)<br>        <br>        # Step 5: AI processes the response<br>        print(&quot;\nStep 5: 📊 AI Analysis Phase&quot;)<br>        success, summary = process_weather_response(forecast_response, location)<br>        <br>        if not success:<br>            print(f&quot;❌ Failed to process data: {summary}&quot;)<br>            continue<br>            <br>        # Step 6: Display results<br>        print(&quot;\nStep 6: 💬 Weather Forecast&quot;)<br>        print(&quot;=&quot; * 60)<br>        print(summary)<br>        print(&quot;=&quot; * 60)<br>        <br>        print(f&quot;\n✅ Weather analysis complete for &#39;{location}&#39;!&quot;)<br><br># Run the agent when the script is executed<br>if __name__ == &quot;__main__&quot;:<br>    run_weather_agent()</pre><ul><li><strong>Key Components</strong>: The code establishes a <strong>Bedrock Client</strong> in the us-west-2 region and structures prompts in the specific message format Claude expects</li><li><strong>Configuration</strong>: You must set inference parameters to control the agent’s <strong>response length and creativity</strong>, while implementing error handling to catch connection issues</li></ul><p><strong>Step 2: Developing the Agent’s Logic</strong></p><p>An agent needs both “hands” to act and a “brain” to plan. This is achieved through four primary functions.</p><ul><li><strong>execute_curl_command():</strong> Functions as the agent’s <strong>hands</strong>, executing the actual HTTP requests to external APIs</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*dxYyWel7YXbWt9VV" /></figure><ul><li><strong>generate_weather_api_calls():</strong> Acts as the <strong>planning</strong> <strong>brain</strong>, determining the correct coordinates and API endpoints based on user input</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*O1uhjIMwuuDqvV3Y" /></figure><ul><li><strong>get_forecast_url_from_points_response():</strong> A utility function to extract the specific forecast URL from the National Weather Service (NWS) JSON response</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*CSJsBeBh9U0fqGSD" /></figure><ul><li><strong>process_weather_response():</strong> Serves as the <strong>analysis brain</strong>, translating complex technical data into human-readable insights</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*kkr3BkGwSnNrPR9W" /></figure><p><strong>Step 3: Executing the Main Workflow</strong></p><p>The agent follows a circular pattern that defines the <strong>Agentic AI Workflow</strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*WN51dpypcW80B76x" /></figure><ul><li><strong>Planning</strong>: The AI analyzes the location (e.g., “Seattle” or “90210”) and maps out the necessary API strategy.</li><li><strong>Action</strong>: The agent executes the planned sequence, calling the <strong>Points API</strong> first to get office data, followed by the <strong>Forecast API</strong>.</li><li><strong>Processing</strong>: Claude analyzes the raw, multi-period JSON weather data to identify the most relevant information.</li><li><strong>Response</strong>: The final summary is presented to the user in a friendly format, such as: <em>“Today: Partly cloudy with a high of 72°F”</em></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/684/0*fszadDpkgvz7Ks0C" /></figure><p><strong>Testing and Validation</strong><br>Once your script is complete, you can run it from your terminal using python weather_agent_cli.py To see the agent&#39;s <strong>autonomy and flexibility</strong> in action,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*OX1xNApcvasbUFrj" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*jvpG8UjNgCsIw8KG" /></figure><h3>Running the Web Application</h3><p>To run the web application for your weather agent, you will use <strong>Streamlit</strong>, a Python library that transforms the agentic logic into an interactive, visual experience. While the web version uses the <strong>identical agentic AI logic</strong> as the command-line version, it provides a professional presentation with real-time progress indicators.</p><p><strong>Launching the Application</strong></p><p>Inside your agentic-ai-workshop folder, ensure you have created the weather_agent_web.py file with the complete code provided in the workshop.</p><p>Copy this complete code:</p><pre>import streamlit as st<br>import boto3<br>import subprocess<br>import json<br>import time<br>from datetime import datetime<br>from PIL import Image<br>import os<br><br># Page configuration<br>st.set_page_config(<br>    page_title=&quot;Weather AI Agent&quot;,<br>    page_icon=&quot;🌤️&quot;,<br>    layout=&quot;wide&quot;,<br>    initial_sidebar_state=&quot;expanded&quot;<br>)<br><br># Custom CSS for better styling<br>st.markdown(&quot;&quot;&quot;<br>&lt;style&gt;<br>    .step-container {<br>        border: 2px solid #e0e0e0;<br>        border-radius: 10px;<br>        padding: 20px;<br>        margin: 10px 0;<br>        background-color: #f9f9f9;<br>    }<br>    .step-header {<br>        font-size: 18px;<br>        font-weight: bold;<br>        color: #1f77b4;<br>        margin-bottom: 10px;<br>    }<br>    .success-box {<br>        border-left: 5px solid #28a745;<br>        background-color: #d4edda;<br>        padding: 10px;<br>        margin: 10px 0;<br>        color: #000000;<br>    }<br>    .error-box {<br>        border-left: 5px solid #dc3545;<br>        background-color: #f8d7da;<br>        padding: 10px;<br>        margin: 10px 0;<br>        color: #000000;<br>    }<br>    .info-box {<br>        border-left: 5px solid #17a2b8;<br>        background-color: #d1ecf1;<br>        padding: 10px;<br>        margin: 10px 0;<br>        color: #000000;<br>    }<br>&lt;/style&gt;<br>&quot;&quot;&quot;, unsafe_allow_html=True)<br><br>def call_claude_sonnet(prompt):<br>    &quot;&quot;&quot;<br>    Connect to Claude 4.5 Sonnet via Amazon Bedrock<br>    &quot;&quot;&quot;<br>    bedrock = boto3.client(<br>        service_name=&#39;bedrock-runtime&#39;,<br>        region_name=&#39;us-west-2&#39;<br>    )<br>    <br>    try:<br>        response = bedrock.converse(<br>            modelId=&#39;us.anthropic.claude-sonnet-4-5-20250929-v1:0&#39;,<br>            messages=[<br>                {<br>                    &quot;role&quot;: &quot;user&quot;,<br>                    &quot;content&quot;: [{&quot;text&quot;: prompt}]<br>                }<br>            ],<br>            inferenceConfig={<br>                &quot;maxTokens&quot;: 2000,<br>                &quot;temperature&quot;: 0.7<br>            }<br>        )<br>        <br>        return True, response[&#39;output&#39;][&#39;message&#39;][&#39;content&#39;][0][&#39;text&#39;]<br>        <br>    except Exception as e:<br>        return False, f&quot;Error calling Claude: {str(e)}&quot;<br><br>def execute_curl_command(url):<br>    &quot;&quot;&quot;<br>    Execute curl command to fetch API data<br>    &quot;&quot;&quot;<br>    try:<br>        result = subprocess.run(<br>            [&#39;curl&#39;, &#39;-s&#39;, url],<br>            capture_output=True,<br>            text=True,<br>            timeout=30<br>        )<br>        <br>        if result.returncode == 0:<br>            return True, result.stdout<br>        else:<br>            return False, f&quot;Curl command failed: {result.stderr}&quot;<br>            <br>    except subprocess.TimeoutExpired:<br>        return False, &quot;Request timed out after 30 seconds&quot;<br>    except Exception as e:<br>        return False, f&quot;Error executing curl: {str(e)}&quot;<br><br>def generate_weather_api_calls(location):<br>    &quot;&quot;&quot;<br>    Use Claude to generate NWS API calls<br>    &quot;&quot;&quot;<br>    prompt = f&quot;&quot;&quot;<br>You are an expert at working with the National Weather Service (NWS) API.<br><br>Your task: Generate the NWS API URL to get weather forecast data for &quot;{location}&quot;.<br><br>Instructions:<br>1. First, determine the approximate latitude and longitude coordinates for this location<br>2. Generate the NWS Points API URL: https://api.weather.gov/points/{{lat}},{{lon}}<br><br>For the coordinates, use your knowledge to estimate:<br>- Major cities: Use well-known coordinates<br>- ZIP codes: Estimate based on the area<br>- States: Use approximate center coordinates<br>- In case a location description is provided instead of a location name, please use the most likely city and state name as the location for the coordinates<br><br>Example for Seattle:<br>https://api.weather.gov/points/47.6062,-122.3321<br><br>Example for largest city in USA:<br>Based on your knowledge, you will establish location is New York City<br>https://api.weather.gov/points/40.7128,-74.0060<br><br>Now generate the API call (Points API) for the established location. <br>Return ONLY the complete Points API URL, nothing else.<br>Format: https://api.weather.gov/points/LAT,LON<br>&quot;&quot;&quot;<br>    <br>    success, response = call_claude_sonnet(prompt)<br>    <br>    if success:<br>        api_url = response.strip()<br>        if api_url.startswith(&#39;https://api.weather.gov/points/&#39;):<br>            return True, [api_url]<br>        else:<br>            return False, f&quot;AI generated invalid URL: {api_url}&quot;<br>    else:<br>        return False, response<br><br>def get_forecast_url_from_points_response(points_json):<br>    &quot;&quot;&quot;<br>    Extract forecast URL from Points API response<br>    &quot;&quot;&quot;<br>    try:<br>        data = json.loads(points_json)<br>        forecast_url = data[&#39;properties&#39;][&#39;forecast&#39;]<br>        return True, forecast_url<br>    except (json.JSONDecodeError, KeyError) as e:<br>        return False, f&quot;Error parsing Points API response: {str(e)}&quot;<br><br>def process_weather_response(raw_json, location):<br>    &quot;&quot;&quot;<br>    Use Claude to process NWS API response<br>    &quot;&quot;&quot;<br>    prompt = f&quot;&quot;&quot;<br>You are a weather information specialist. I have raw National Weather Service forecast data for &quot;{location}&quot; that needs to be converted into a clear, helpful summary for a general audience.<br><br>Raw NWS API Response:<br>{raw_json}<br><br>Please create a weather summary that includes:<br>1. A brief introduction with the location<br>2. Current conditions and today&#39;s forecast<br>3. The next 2-3 days outlook with key details (temperature, precipitation, wind)<br>4. Any notable weather patterns or alerts<br>5. Format the response to be easy to read and understand<br><br>Make it informative and practical for someone planning their activities. Focus on being helpful and clear.<br>&quot;&quot;&quot;<br>    <br>    success, response = call_claude_sonnet(prompt)<br>    return success, response<br><br># Sidebar with information<br>st.sidebar.title(&quot;🤖 About This Agent&quot;)<br>st.sidebar.markdown(&quot;&quot;&quot;<br>This AI agent demonstrates **Agentic AI** principles:<br><br>**🧠 Intelligence**: Uses Claude 4.5 Sonnet to understand locations and plan API calls<br><br>**🔗 Action**: Automatically calls the National Weather Service API<br><br>**📊 Processing**: Converts complex weather data into readable forecasts<br><br>**💬 Response**: Provides helpful, practical weather information<br>&quot;&quot;&quot;)<br><br>st.sidebar.markdown(&quot;---&quot;)<br>st.sidebar.markdown(&quot;### 🏗️ Architecture&quot;)<br>st.sidebar.markdown(&quot;&quot;&quot;<br>1. **User Input** → Location name<br>2. **AI Planning** → Generate API calls<br>3. **Points API** → Get forecast office  <br>4. **Forecast API** → Get weather data<br>5. **AI Processing** → Create summary<br>6. **Display Results** → Show to user<br>&quot;&quot;&quot;)<br><br># Main application<br>st.title(&quot;🌤️ Weather AI Agent&quot;)<br>st.markdown(&quot;### Powered by Claude 4.5 Sonnet on Amazon Bedrock&quot;)<br><br>st.markdown(&quot;&quot;&quot;<br>This intelligent agent helps you get weather forecasts using the National Weather Service API. <br>Enter any location below and watch the AI agent work through its reasoning process!<br>&quot;&quot;&quot;)<br><br># Initialize session state for results<br>if &#39;show_results&#39; not in st.session_state:<br>    st.session_state.show_results = False<br><br># Input section<br>st.markdown(&quot;---&quot;)<br>location = st.text_input(<br>    &quot;🔍 Enter a location name or description:&quot;,<br>    placeholder=&quot;e.g., Seattle, 90210, New York City, National park near Homestead in Florida&quot;,<br>    help=&quot;You can enter city names, ZIP codes, state names, or location descriptions&quot;<br>)<br><br># Create columns for the buttons<br>button_col1, button_col2 = st.columns([2, 1])<br><br>with button_col1:<br>    get_forecast = st.button(&quot;🚀 Get Weather Forecast&quot;, type=&quot;primary&quot;)<br><br>with button_col2:<br>    clear_results = st.button(&quot;🗑️ Clear Results&quot;, type=&quot;secondary&quot;)<br><br># Clear results functionality<br>if clear_results:<br>    st.session_state.show_results = False<br>    st.success(&quot;🗑️ Results cleared! Enter a new location to get a fresh forecast.&quot;)<br><br>if get_forecast:<br>    st.session_state.show_results = True<br><br>if st.session_state.show_results and get_forecast:<br>    if not location:<br>        st.error(&quot;❌ Please enter a location name or description.&quot;)<br>    else:<br>        # Create columns for better layout<br>        col1, col2 = st.columns([2, 1])<br>        <br>        with col1:<br>            st.markdown(f&quot;## Weather Analysis for: **{location}**&quot;)<br>            <br>            # Step 1: AI Planning<br>            with st.container():<br>                st.markdown(&#39;&lt;div class=&quot;step-container&quot;&gt;&#39;, unsafe_allow_html=True)<br>                st.markdown(&#39;&lt;div class=&quot;step-header&quot;&gt;🧠 Step 1: AI Planning Phase&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                <br>                with st.spinner(&quot;Claude is analyzing the location and planning the API calls...&quot;):<br>                    success, api_calls = generate_weather_api_calls(location)<br>                <br>                if success:<br>                    points_url = api_calls[0]<br>                    st.markdown(&#39;&lt;div class=&quot;success-box&quot;&gt;✅ Points API URL generated successfully!&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.code(points_url, language=&quot;text&quot;)<br>                else:<br>                    st.markdown(f&#39;&lt;div class=&quot;error-box&quot;&gt;❌ Failed to generate API calls: {api_calls}&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.stop()<br>                <br>                st.markdown(&#39;&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>            <br>            # Step 2: Points API Execution<br>            with st.container():<br>                st.markdown(&#39;&lt;div class=&quot;step-container&quot;&gt;&#39;, unsafe_allow_html=True)<br>                st.markdown(&#39;&lt;div class=&quot;step-header&quot;&gt;🔗 Step 2: Points API Execution&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                <br>                with st.spinner(&quot;Fetching location data from National Weather Service...&quot;):<br>                    success, points_response = execute_curl_command(points_url)<br>                <br>                if success:<br>                    st.markdown(&#39;&lt;div class=&quot;success-box&quot;&gt;✅ Received location data from NWS&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    <br>                    # Show a preview of the raw data<br>                    with st.expander(&quot;🔍 View Raw Points API Response (first 500 characters)&quot;):<br>                        st.code(points_response[:500] + &quot;...&quot; if len(points_response) &gt; 500 else points_response, language=&quot;json&quot;)<br>                else:<br>                    st.markdown(f&#39;&lt;div class=&quot;error-box&quot;&gt;❌ Failed to fetch points data: {points_response}&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.stop()<br>                <br>                st.markdown(&#39;&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>            <br>            # Step 3: Extract Forecast URL<br>            with st.container():<br>                st.markdown(&#39;&lt;div class=&quot;step-container&quot;&gt;&#39;, unsafe_allow_html=True)<br>                st.markdown(&#39;&lt;div class=&quot;step-header&quot;&gt;📍 Step 3: Extracting Forecast URL&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                <br>                success, forecast_url = get_forecast_url_from_points_response(points_response)<br>                <br>                if success:<br>                    st.markdown(&#39;&lt;div class=&quot;success-box&quot;&gt;✅ Forecast URL extracted successfully!&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.code(forecast_url, language=&quot;text&quot;)<br>                else:<br>                    st.markdown(f&#39;&lt;div class=&quot;error-box&quot;&gt;❌ Failed to extract forecast URL: {forecast_url}&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.stop()<br>                <br>                st.markdown(&#39;&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>            <br>            # Step 4: Forecast API Execution<br>            with st.container():<br>                st.markdown(&#39;&lt;div class=&quot;step-container&quot;&gt;&#39;, unsafe_allow_html=True)<br>                st.markdown(&#39;&lt;div class=&quot;step-header&quot;&gt;🌦️ Step 4: Forecast API Execution&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                <br>                with st.spinner(&quot;Fetching weather forecast data...&quot;):<br>                    success, forecast_response = execute_curl_command(forecast_url)<br>                <br>                if success:<br>                    st.markdown(f&#39;&lt;div class=&quot;success-box&quot;&gt;✅ Received {len(forecast_response):,} characters of forecast data&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    <br>                    # Show a preview of the raw data<br>                    with st.expander(&quot;🔍 View Raw Forecast API Response (first 500 characters)&quot;):<br>                        st.code(forecast_response[:500] + &quot;...&quot; if len(forecast_response) &gt; 500 else forecast_response, language=&quot;json&quot;)<br>                else:<br>                    st.markdown(f&#39;&lt;div class=&quot;error-box&quot;&gt;❌ Failed to fetch forecast data: {forecast_response}&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.stop()<br>                <br>                st.markdown(&#39;&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>            <br>            # Step 5: AI Processing<br>            with st.container():<br>                st.markdown(&#39;&lt;div class=&quot;step-container&quot;&gt;&#39;, unsafe_allow_html=True)<br>                st.markdown(&#39;&lt;div class=&quot;step-header&quot;&gt;📊 Step 5: AI Analysis Phase&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                <br>                with st.spinner(&quot;Claude is processing the weather data and creating a summary...&quot;):<br>                    success, summary = process_weather_response(forecast_response, location)<br>                <br>                if success:<br>                    st.markdown(&#39;&lt;div class=&quot;success-box&quot;&gt;✅ Weather analysis complete!&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                else:<br>                    st.markdown(f&#39;&lt;div class=&quot;error-box&quot;&gt;❌ Failed to process data: {summary}&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>                    st.stop()<br>                <br>                st.markdown(&#39;&lt;/div&gt;&#39;, unsafe_allow_html=True)<br>            <br>            # Step 6: Results<br>            st.markdown(&quot;---&quot;)<br>            st.markdown(&quot;## 🌤️ Weather Forecast&quot;)<br>            st.markdown(summary)<br>            <br>        with col2:<br>            # Real-time status updates<br>            st.markdown(&quot;### 📊 Process Status&quot;)<br>            <br>            status_container = st.container()<br>            with status_container:<br>                st.markdown(&quot;&quot;&quot;<br>                &lt;div class=&quot;info-box&quot;&gt;<br>                &lt;strong&gt;🔄 Agent Workflow:&lt;/strong&gt;&lt;br&gt;<br>                ✅ Planning Phase&lt;br&gt;<br>                ✅ Points API Call&lt;br&gt;<br>                ✅ URL Extraction&lt;br&gt;<br>                ✅ Forecast API Call&lt;br&gt;<br>                ✅ Data Processing&lt;br&gt;<br>                ✅ Results Generated<br>                &lt;/div&gt;<br>                &quot;&quot;&quot;, unsafe_allow_html=True)<br>            <br>            st.markdown(&quot;### 🎯 What Makes This Agentic?&quot;)<br>            st.markdown(&quot;&quot;&quot;<br>            - **🧠 Reasoning**: AI understands location formats<br>            - **📋 Planning**: Generates appropriate API call sequences<br>            - **🔧 Action**: Executes real-world API requests<br>            - **📊 Processing**: Converts raw data to insights<br>            - **🔄 Adaptation**: Handles different location types<br>            &quot;&quot;&quot;)<br><br># Footer<br>st.markdown(&quot;---&quot;)<br>st.markdown(&quot;&quot;&quot;<br>### 🔬 About This Demo<br><br>This application demonstrates **Agentic AI** principles using:<br>- **Amazon Bedrock** with Claude 4.5 Sonnet for intelligent reasoning<br>- **National Weather Service API** for real-time weather data<br>- **Streamlit** for interactive web interface<br><br>**⚠️ Important**: This uses official NWS data for educational purposes. For critical weather decisions, consult official sources.<br>&quot;&quot;&quot;)<br><br># Add some example queries<br>st.markdown(&quot;### 💡 Try These Examples:&quot;)<br>st.markdown(&quot;&quot;&quot;<br>**Suggested locations to test:**<br>- **Seattle** - Major city (tests city name recognition)<br>- **90210** - ZIP code (tests postal code handling)  <br>- **New York City** - Multi-word city (tests complex location parsing)<br>- **Miami, FL** - City with state (tests state abbreviations)<br>- **Chicago** - Another major city (tests different coordinates)<br>- **National park near Homestead in Florida** - Location description (tests AI reasoning)<br>- **Largest City in California** - Descriptive query (tests knowledge-based location finding)<br><br>Simply copy any of these into the location input above and click &quot;Get Weather Forecast&quot;!<br>&quot;&quot;&quot;)</pre><p>To start the application, run the following command in your terminal: streamlit run weather_agent_web.py</p><p>Once executed, you will see output in your terminal indicating the local URL where you can view the interface in your web browser.</p><p>You should see output like:</p><pre>  You can now view your Streamlit app in your browser.<br><br>  Local URL: http://localhost:8501<br>  Network URL: http://192.168.1.100:8501</pre><p><strong>Interactive Features to Explore</strong></p><p>The web interface is designed to make the “agentic” nature of the AI visible to the user through several key features:</p><ul><li><strong>Step-by-Step Visualization</strong>: Unlike the CLI, the web app allows you to watch each phase of the workflow as it happens, including the <strong>AI Planning Phase</strong>, <strong>API Execution</strong>, and <strong>AI Analysis</strong>.</li><li><strong>Raw Data Inspection</strong>: You can use expandable sections to view the <strong>complex JSON data</strong> the agent is processing behind the scenes.</li><li><strong>Real-time Status</strong>: A sidebar allows you to monitor the agent’s progress as it moves from understanding your request to fetching and summarizing data</li></ul><p><strong>Observations and Testing</strong></p><p>When running the web application, you should observe how the AI demonstrates <strong>Planning Intelligence</strong> and <strong>Data Processing</strong>:</p><ul><li><strong>Input Flexibility</strong>: Test how it recognizes various inputs like “94129” or “Times Square” and maps them to coordinates.</li><li><strong>Data Translation</strong>: Observe how Claude 4.5 Sonnet identifies the most relevant information from multiple forecast periods to create a human-readable summary.</li><li><strong>Error Handling</strong>: Try entering misspelled city names or international locations (remembering that the NWS API only covers the US) to see how the agent provides helpful guidance rather than crashing</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Y9Zmp8gx3ldP54n_" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*cupKkpWO0LvpvzkG" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*d5a2cAXjthnNypbG" /></figure><p><strong>Source Code &amp; Repository</strong></p><p>You can access the full implementation of the weather_agent_cli.py and weather_agent_web.py files, along with the environment setup requirements, at the following GitHub repository here : <a href="https://github.com/dineshrajdhanapathyDD/agentic-ai-amazon-bedrock.git">Agentic AI Building Blocks using Amazon Bedrock</a></p><p><strong>Why Use the Web Version?</strong></p><p>The web application is particularly useful for <strong>non-technical users</strong> or for <strong>professional presentations</strong>. It provides a visual confirmation of the agent’s autonomy and reactivity, allowing you to see exactly how the “brain” (Claude) is directing the “hands” (API calls) to solve the problem</p><p><strong>What to Observe</strong></p><p>During the execution of the agent, you should pay close attention to how the AI handles the transition from natural language to technical execution:</p><ul><li><strong>Planning Intelligence:</strong> Notice how the AI recognizes various input types, such as mapping “Seattle” to coordinates or handling ZIP codes like “90210”. Observe how it dynamically generates different API calls based on the specific location type provided.</li><li><strong>Data Processing:</strong> Watch how the AI converts complex JSON data often containing multiple forecast periods into a human-readable summary. It must identify the most relevant information and add appropriate context or recommendations.</li><li><strong>Error Handling:</strong> Test the agent’s resilience by entering invalid locations, misspelled city names, or international locations. Observe how it provides helpful guidance rather than failing, noting that the National Weather Service API is limited to US locations</li></ul><p><strong>Comparing CLI vs. Web Experience</strong></p><p>While both versions use the <strong>identical agentic AI logic</strong> — relying on the same Claude 4.5 Sonnet reasoning and NWS API integration — they offer distinct advantages depending on the user:</p><pre>| Feature | Command-Line Version (CLI)| Web Version (Streamlit) |<br>| --------| ----------------------------| --------------|<br>|Primary Audience |Developers and technical users. |Non-technical users or professional presentations.|<br>|Key Strength|Direct, focused interaction ideal for automation and scripting.|Visual, interactive experience with a polished presentation.|<br>|Visibility|Shows the raw workflow steps clearly in the terminal.|Provides real-time progress indicators and a sidebar to monitor status.|<br>|Visualization|Text-based output of the process.|Step-by-step visualization of the planning, execution, and analysis phases.|<br>|Inspection|Best for seeing the immediate code execution.|Includes expandable sections to inspect raw API responses.|</pre><p>Ultimately, the choice between them depends on whether you prioritize <strong>developer efficiency and automation</strong> (CLI) or <strong>visual clarity and user-friendly status tracking</strong> (Web).</p><p><strong>Key Takeaways</strong></p><ul><li><strong>Definition of Agentic AI:</strong> The fundamental shift in AI is moving from a “Question → Answer” model (smart encyclopedias) to a <strong>“Problem → Plan → Action → Result”</strong> workflow. Truly agentic systems are defined by <strong>autonomy</strong> (making independent decisions), <strong>reactivity</strong> (responding to environmental changes and errors), and <strong>proactivity</strong> (taking initiative through multi-step planning).</li><li><strong>The AI as a Reasoning Engine:</strong> Instead of just generating text, the AI (specifically Claude 4.5 Sonnet) acts as a <strong>reasoning engine</strong>. It dynamically interprets descriptive user input, such as “largest city in California,” maps it to precise coordinates, and constructs the necessary API URLs rather than relying on hardcoded logic.</li><li><strong>The Agentic Workflow:</strong> A successful agent follows a reusable four-step pattern: <strong>Planning</strong> (analyzing input and strategizing), <strong>Action</strong> (executing technical tasks like API calls), <strong>Processing</strong> (analyzing raw data for insights), and <strong>Response</strong> (presenting results in a human-friendly format).</li><li><strong>Building without Abstractions:</strong> Sophisticated agents can be built using <strong>pure Python and the AWS SDK (boto3)</strong> without the need for complex frameworks. This “from-scratch” approach helps developers understand every component, from establishing a Bedrock connection to creating an <strong>error-resilient architecture</strong> that handles network timeouts and invalid data.</li><li><strong>Versatile Real-World Applications:</strong> The patterns used to build a weather assistant such as <strong>coordinate resolution and sequential API execution</strong> are directly applicable to other fields like travel planning, emergency management, market analysis, and agriculture</li></ul><p><strong>Conclusion</strong></p><p>The workshop concludes that the future of artificial intelligence is not just about chatbots, but about <strong>agents that can take meaningful actions in the world</strong>. By moving beyond static responses and integrating real-time data with AI reasoning, developers can create systems that provide genuine value across any domain. The core principle for success in this new landscape is to <strong>start simple and iterate quickly</strong>, using foundational building blocks to scale toward more complex multi-agent systems and advanced production environments.</p><h3>Resources</h3><ul><li><a href="https://docs.aws.amazon.com/bedrock/">Amazon Bedrock Documentation</a></li><li><a href="https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-claude.html">Claude Model Parameters</a></li><li><a href="https://www.weather.gov/documentation/services-web-api">National Weather Service API</a></li><li><a href="https://docs.streamlit.io/">Streamlit Documentation</a></li></ul><p>A special thanks to <strong>BeSA (Become a Solutions Architect)</strong> for providing a free mentoring platform that enabled me to gain <strong>practical, hands-on experience</strong> with Generative AI through this workshop. BeSA’s guidance and community support played a key role in helping me strengthen my GenAI and cloud skills.</p><p>If you’d like to explore another hands-on agentic AI example using Amazon Bedrock, check out the AWS workshop 1 article. You can read it here:🔗 <a href="https://builder.aws.com/content/3A2RZg37vuIPXHmXmPVAhZGkIUm/building-an-intelligent-shopping-assistant-with-amazon-bedrock-agents"><strong>Building an Intelligent Shopping Assistant with Amazon Bedrock Agents</strong></a><strong>.</strong></p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments; I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud, AI and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️<a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4f496da35dcb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automating AWS S3 with Terraform]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/automating-aws-s3-with-terraform-1108665bfe97?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/1108665bfe97</guid>
            <category><![CDATA[iac]]></category>
            <category><![CDATA[s3]]></category>
            <category><![CDATA[terraform]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sun, 04 Jan 2026 19:40:43 GMT</pubDate>
            <atom:updated>2026-01-04T20:20:27.578Z</atom:updated>
            <content:encoded><![CDATA[<p>Over the past months in my cloud and DevOps journey, I’ve constantly been looking for ways to move beyond theory and gain <em>real, practical, hands-on experience</em>. As an <strong>AWS Community Builder</strong>, I’ve had the opportunity to learn from industry experts, collaborate with passionate builders, and explore how real-world cloud environments are automated and managed at scale.</p><p>While working with AWS services, I realized that manually creating resources in the console is fine for learning but it doesn’t scale in real DevOps environments. That’s when I decided to start my <strong>Terraform journey</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J6-HkhwOYba244ApaihWYA.png" /></figure><p>I wanted to understand how automation to build, deploy, and manage infrastructure efficiently. Instead of creating cloud resources manually in the AWS Console, Infrastructure as Code (IaC) allows us to define infrastructure in reusable configuration files – making deployments faster, repeatable, and less error-prone.</p><p>In this blog, I used <strong>Terraform</strong> to automate the deployment of an <strong>AWS S3 bucket</strong> and upload an object to it. I will demonstrate how to install and set up Terraform, configure AWS credentials securely via the terminal, and use Terraform to create and manage S3 buckets. The goal is to automate the provisioning of AWS S3 resources and upload files directly to S3 using Terraform configurations, ensuring efficient and repeatable infrastructure deployment. This was a great hands-on experience that strengthened my understanding of cloud automation and DevOps workflows.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/887/1*RVm8UGtaRHNL5N8oUe4rdw.png" /></figure><p><strong>Download and Install Terraform</strong></p><ul><li>Head to the official Terraform download page: <a href="https://developer.hashicorp.com/terraform/downloads">https://developer.hashicorp.com/terraform/downloads</a></li></ul><blockquote><strong>🍎 MacOS:</strong> Select the Apple icon from the top left hand corner of your computer’s menu bar. Select <strong>About this Mac</strong>, and note whether your <strong>Chip</strong> says <strong>Apple</strong> (pick ARM64) or <strong>Intel</strong> (pick AMD64).</blockquote><blockquote><strong>🖼️ Windows:</strong> Select the <strong>Start</strong> button and search for <strong>System Information</strong>. Note whether your <strong>System Type</strong> says x64-based or ARM-based PC.</blockquote><blockquote><strong>🐧 Linux:</strong> Open a terminal and run <strong>uname -m</strong>. Match the output to the correct Linux package option.</blockquote><h4>What I Set Out to Do</h4><p><strong>I demonstrated how to:</strong></p><ul><li>Install and configure Terraform</li><li>Configure AWS credentials securely using AWS CLI</li><li>Create and manage an S3 bucket through Terraform</li><li>Upload a file to S3 using Terraform</li><li>Destroy resources safely after use</li></ul><p>The goal was to automate AWS resource provisioning instead of doing it manually, following real-world DevOps practices.</p><h3>What is Terraform?</h3><p>Terraform is an <strong>Infrastructure as Code (IaC) </strong>tool that lets you build and manage cloud infrastructure using configuration files instead of creating resources manually. You define what you want, such as S3 buckets, servers, networks, or databases, and Terraform automatically provisions them for you based on that blueprint.</p><h3>What is Infrastructure as Code (IaC)?</h3><p>Infrastructure as Code means managing infrastructure using code rather than manual setup in the cloud console. Because everything is defined in files, deployments are:</p><ul><li>consistent</li><li>repeatable</li><li>easy to rebuild or modify</li><li>less prone to human error</li></ul><p>Terraform is one of the most popular IaC tools because it supports multiple cloud providers like <strong>AWS, Azure, and Google Cloud</strong>, making it ideal for scalable and multi-cloud environments.</p><p><strong>What is a Terraform binary?</strong><br>A Terraform binary is the main <strong>executable</strong> file that lets you run Terraform commands on your system. It’s the core program your operating system uses to execute Terraform.</p><p><strong>What is PATH?</strong></p><p>PATH is a list of system folders where your computer looks for programs when you run a command in the terminal. It lets you run tools like terraform without typing their full file location.</p><h4>Installing Terraform</h4><p>Terraform is a powerful IaC tool that allows you to define infrastructure using configuration files.</p><h4>🍎🐧 MacOS / Linux — Add Terraform to PATH</h4><ol><li>Open Terminal and navigate to the unzipped Terraform folder, e.g.:</li></ol><pre>cd ~/Downloads/terraform_1.10.x_xxx</pre><p>2. (Optional) Confirm location:</p><pre>pwd</pre><p>3. Move the Terraform binary to /usr/local/bin:</p><pre>sudo mv terraform /usr/local/bin/</pre><p>4. Enter your system password if prompted.</p><p>You’ve just installed Terraform on your local computer. Now we’ll need to verify we’ve installed it correctly.</p><h4>🖥️ Windows — Add Terraform to PATH</h4><ol><li>Open <strong>System Properties → Advanced → Environment Variables</strong></li><li>Under <strong>System Variables</strong>, select <strong>PATH → Edit</strong></li><li>Click <strong>New</strong> and add:</li></ol><pre>C:\terraform</pre><p>4. Save changes and open a <strong>new terminal session.</strong></p><p>After installation, <strong>verify Terraform Installation</strong></p><ul><li>Run the following command:</li></ul><pre>terraform version</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/491/1*TwunTqZZ99ZfSGDoI0AhMw.png" /><figcaption><em>Terraform version output</em></figcaption></figure><p>✔️ Terraform was successfully installed.</p><h4>📁 Setting Up the Terraform Project</h4><p>I created my project directory and configuration file:</p><pre>mkdir ~/Desktop/DD_terraform<br>cd ~/Desktop/DD_terraform<br>touch main.tf</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/786/1*TOWhADRarnyn3rA1LNkvZA.png" /></figure><h4>What is main.tf?</h4><p>main.tf is the primary Terraform configuration file where you define how your infrastructure should be built. The main.tf file acts as the blueprint of the infrastructure Terraform will build.</p><h4>How does Terraform use configuration files?</h4><p>Terraform uses configuration files (like main.tf) to describe the desired state of your infrastructure and then automatically creates or updates resources to match that state.</p><p>I will create a project directory and set up the main Terraform file (main.tf) because it organizes the infrastructure code and defines the resources to be provisioned. This structure makes the project clear, reusable, and easier for Terraform to manage and apply changes.</p><h4>🧱 Defining the AWS S3 Bucket in Terraform</h4><p>Inside main.tfI defined the AWS provider and S3 bucket:</p><ul><li>Open main.tf in a text editor like <strong>VS Code, NotePad++</strong> or <strong>TextEdit.</strong></li><li>Copy and paste the following code into main.tf</li></ul><pre>provider &quot;aws&quot; {<br>  region = &quot;us-east-2     #update the region closest to you<br>}<br>resource &quot;aws_s3_bucket&quot; &quot;my_bucket&quot; {<br>  bucket = &quot;unique-bucket-{yourname}-{random number}&quot; #Ensure the bucket name is unique<br>}<br>resource &quot;aws_s3_bucket_public_access_block&quot; &quot;my_bucket_public_access_block&quot; {<br>  bucket = aws_s3_bucket.my_bucket.id<br>  block_public_acls       = true<br>  ignore_public_acls      = true<br>  block_public_policy     = true<br>  restrict_public_buckets = true<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2YpTU8V-Q4gsnSc37Qkvgg.png" /><figcaption>main.tf configuration</figcaption></figure><p>Bucket names must be <strong>globally unique</strong>, so I customized the bucket name accordingly.</p><h4><strong>What’s in this Terraform configuration?</strong></h4><ul><li><strong>provider “aws”</strong> — Tells Terraform to use AWS as the cloud provider and enables it to interact with AWS services.</li><li><strong>resource “aws_s3_bucket” “my_bucket”</strong> — Creates an S3 bucket, with my_bucket used as a reference name inside the configuration.</li><li><strong>resource “aws_s3_bucket_public_access_block”</strong> — Blocks public access to the bucket by enabling security settings that prevent public ACLs and policies.</li></ul><p><strong>How are configuration files structured in Terraform?</strong></p><p><strong>Terraform configuration files are structured in blocks</strong>, where each block defines a specific part of the infrastructure.</p><ul><li><strong>Provider blocks</strong> specify which cloud or platform (AWS, Azure, GCP, etc.) Terraform will use.</li><li><strong>Resource blocks</strong> define the actual resources to create or manage, like S3 buckets, servers, or networks.</li></ul><p>The configuration is structured in provider and resource blocks, where each block defines a specific part of the infrastructure. This keeps the code modular, readable, and easy to maintain. For example, the provider block connects to AWS, while the S3 resource block defines the bucket.</p><ul><li>Visit the Terraform registry’s documentation on <a href="https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/s3_bucket">S3 buckets.</a></li></ul><h4><strong>What is the Terraform registry?</strong></h4><p>The Terraform Registry is an online library of Terraform providers and modules. It contains documentation, examples, best practices, and configuration options. Developers can also publish modules there, allowing others to reuse them instead of writing everything from scratch.</p><h4><strong>What can I find in this aws_s3_bucket page?</strong></h4><p>The aws_s3_bucket page in the Terraform documentation explains how to create and manage S3 buckets. It includes configuration options, properties, examples, and details on how the bucket integrates with other AWS services.</p><p>I visited the Terraform documentation to understand how to use the aws_s3_bucket resource. It explains how to configure S3 buckets using arguments like bucket name, ACLs, and tags, and provides examples and best-practice guidance, helping me customize and extend my project confidently.</p><ul><li>From the documentation, choose an additional configuration that you might want to add to your S3 bucket or the object you’ve uploaded.</li><li>Write your configuration in main.tf</li><li>Save your changes to main.tf</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/798/1*8cP4mGrFEPBdTbgm1LN0CA.png" /><figcaption>Tags</figcaption></figure><p>I customized my S3 bucket by adding tags to help identify and organize resources. I used tags Environment = &quot;Dev&quot; to label its purpose and environment. I verified this customization in the AWS S3 console by checking the Tags section, where these labels appear, making the bucket easier to manage and filter in larger environments.</p><h4>⚙️ Initializing and Planning the Terraform Configuration</h4><p>I will initialize Terraform and run a plan because initialization sets up the project and downloads the required provider plugins, while the plan previews the changes Terraform will make in AWS. This helps verify the configuration before applying it and reduces the risk of errors or unexpected changes.</p><p>I initialized Terraform using:</p><pre>terraform init</pre><p>This downloaded the AWS provider plugin and prepared the project.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y8OkO3l9S35Ej6dtcdoECA.png" /><figcaption>terraform init output</figcaption></figure><p>Next, I previewed the changes Terraform would make:</p><pre>terraform plan</pre><p>This execution plan showed which resources would be created before anything was applied.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S0Z5w68C8yJPdbvVCNlLqQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i_zKqwgQH0TtEnHO9EZHkA.png" /><figcaption>terraform plan output</figcaption></figure><p>I ran terraform plan to generate an execution plan that previews the changes Terraform will make to my AWS infrastructure based on the configuration in main.tf. It shows which resources will be created, updated, or deleted, allowing me to review the configuration safely before applying any real changes.</p><h4><strong>you get an error?</strong></h4><blockquote>When I tried to plan my Terraform configuration, I received an error message that says the AWS provider couldn’t authenticate because my credentials were missing or incorrect. This happened because I hadn’t configured the AWS CLI properly or used invalid access keys. Terraform needs valid credentials to connect to AWS and manage resources, so the plan command fails without proper authentication setup.</blockquote><h3>🔑 Configuring AWS CLI and Credentials</h3><p>To allow Terraform to authenticate with AWS, I installed AWS CLI and verified it:</p><pre>aws --version</pre><p>I will install the <strong>AWS CLI</strong>, generate access keys, and configure the CLI so Terraform can securely authenticate with my AWS account. This ensures Terraform has the required permissions to create and manage resources like S3 buckets during deployment.</p><p>Then configured my credentials:</p><pre>aws configure</pre><p>I entered:</p><ul><li>Access Key ID</li><li>Secret Access Key</li><li>AWS Region</li><li>Output format (JSON)</li></ul><p>This enabled <strong>secure programmatic access</strong> for Terraform.</p><h3>🪣 Creating the S3 Bucket with Terraform</h3><p>Once everything was configured, I deployed the infrastructure:</p><pre>terraform apply</pre><p>I will apply my Terraform configuration because this command takes the planned changes and actually creates the defined infrastructure in this case, an S3 bucket on AWS. This is the step where Terraform provisions resources, after which I will verify the bucket in the S3 console to ensure it was created as defined in main.tf.</p><p>After reviewing the plan, I confirmed with <strong>&#39;yes&#39;,</strong> and Terraform created the S3 bucket successfully.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a2ENPVA6v_xHfo5m_mfSOg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YVEPGNo3SJ2k_6LhHl5S2w.png" /><figcaption>terraform apply output</figcaption></figure><p>I ran terraform apply to execute the changes in my configuration and provision resources like the S3 bucket. This command creates, updates, or deletes resources in my AWS account based on main.tf, and it confirms the plan before applying changes turning the code into real infrastructure.</p><p>Seeing the bucket appear in AWS was a very satisfying moment 🙂</p><h3>💎 Uploading an Object to S3 Using Terraform</h3><p>For the project’s, I uploaded a file to the bucket using Terraform.</p><p>I will update my Terraform configuration, reapply it, and document the changes to demonstrate how Terraform detects and applies infrastructure updates. By modifying settings like tags or versioning, I can show how changes are managed safely and reliably using Infrastructure as Code.</p><p><strong>Add a New Image File</strong></p><ul><li><a href="https://web-unified-docs-hashicorp.vercel.app/api/assets/terraform/latest/img/docs/intro-terraform-workflow.png">Download this image.</a> To download it as an image, you can right click on this link and select <strong>Save Link As…</strong></li></ul><p>I placed image.png inside my project folder and added:</p><pre>resource &quot;aws_s3_object&quot; &quot;image&quot; {<br>  bucket = aws_s3_bucket.my_bucket.id<br>  key    = &quot;image.png&quot;<br>  source = &quot;image.png&quot;<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2CgJYDK5308ObNQHhe5lGg.png" /><figcaption>Add image resource</figcaption></figure><p>Then I reapplied the configuration:</p><pre>terraform plan<br>terraform apply</pre><p>We need to run terraform apply again because any changes made to the configuration must be applied to update the actual infrastructure. Terraform compares the current state with the new configuration and applies only the detected changes ensuring the deployed environment stays consistent with the code. Terraform detected the new resource and uploaded the file automatically.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VAR3xRxTCI1V5VHLS9ld1Q.png" /><figcaption>S3 bucket in AWS console</figcaption></figure><p>To validate my update, I confirmed that terraform apply completed successfully, then checked the S3 console and verified that the image.jpg file appeared in my bucket. This confirmed that the Terraform changes were applied correctly and reflected in my AWS environment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xQwofQiFAhsZvK-pPLfwOQ.png" /><figcaption>uploaded image in S3 bucket</figcaption></figure><p>I validated the upload by downloading and viewing the file inside AWS S3.</p><h3>🧹 Cleaning Up — Destroying Resources</h3><p>To avoid keeping unused resources, I removed the infrastructure using:</p><pre>terraform destroy</pre><p>Terraform deleted:</p><ul><li>The S3 bucket</li><li>The uploaded object</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/837/1*g62KB5gWuGZ2fbGQW0eYEQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/828/1*s6vdG4SQ_2iwK7c78C8X9Q.png" /><figcaption>terraform destroy output</figcaption></figure><p>I also deleted the IAM access key created for the project following security best practices.</p><h3>📚 Key Concepts I Learned</h3><p>This project helped me strengthen my understanding of:</p><p>✔️ Infrastructure as Code (IaC)<br> ✔️ Terraform providers and resource blocks<br> ✔️ Terraform lifecycle<br> init → plan → apply → destroy<br> ✔️ State files and execution plans<br> ✔️ Secure AWS authentication via CLI<br> ✔️ S3 storage automation</p><p>I also experienced how Terraform enables:</p><ul><li>repeatable deployments</li><li>reduced manual effort</li><li>version-controlled infrastructure</li><li>safer cloud operations</li></ul><p>GitHub link: <a href="https://github.com/dineshrajdhanapathyDD/Terraform-series.git">https://github.com/dineshrajdhanapathyDD/Terraform-series.git</a></p><h3>Conclusion</h3><p>Terraform is a highly practical and in-demand skill for roles such as DevOps Engineer, DevSecOps Engineer, Platform Engineer, and Cloud Engineer. Through this project, I gained hands-on experience in real-world cloud automation from configuring AWS credentials to provisioning and managing S3 buckets using Infrastructure as Code.</p><p>I worked with AWS S3 for storage and AWS CLI for authentication, and strengthened key Terraform concepts including providers, resource blocks, and state files. I also practiced the full Terraform workflow initialization, planning, applying changes, and updating infrastructure safely and consistently.</p><p>This experience has motivated me to continue expanding my Terraform journey into areas like EC2 provisioning, VPC and networking automation, bucket versioning and lifecycle rules, and remote Terraform state management.</p><p>Overall, this project was a great learning experience and a meaningful step forward in my cloud and DevOps journey</p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments, I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌ </strong><a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻 </strong><a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p><strong>🐙 </strong><a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1108665bfe97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AWS Rift Rewind Story: Learning Cloud, Building Projects]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/aws-rift-rewind-story-learning-cloud-building-projects-0485bf2c46cb?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/0485bf2c46cb</guid>
            <category><![CDATA[s3]]></category>
            <category><![CDATA[cloudfront]]></category>
            <category><![CDATA[amplify]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[aws-lambda]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sun, 05 Oct 2025 14:34:01 GMT</pubDate>
            <atom:updated>2025-10-05T14:34:01.007Z</atom:updated>
            <content:encoded><![CDATA[<p>AWS beginner to deploying a global project via Rift Rewind: secure storage, CloudFront, automated pipelines, live APIs — my journey &amp; tips for builders.</p><h3>🌐 Live Project</h3><p>Check out my live project here: <a href="https://main.d1y7eioc5275a7.amplifyapp.com/">https://main.d1y7eioc5275a7.amplifyapp.com/</a></p><h3>✨ Adding My Personal Touch</h3><p>I didn’t just follow the templates, I made the site my own:</p><ul><li>Customized the <strong>design, colors, and layout</strong> to match my style</li><li>Added <strong>Riot Games API integration</strong> to fetch live League of Legends data</li><li>Included <strong>personal content and visuals</strong> to make the site unique</li></ul><p>This wasn’t just a website — it became my personal playground for learning AWS.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*7hil0frocMgQ_LZH" /></figure><h3>Power Spike Prep</h3><p>I started by setting up my AWS account safely:</p><ul><li>Enabled <strong>MFA</strong></li><li>Created an <strong>admin IAM user</strong></li><li>Set up <strong>budget alerts</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*CUJye7v61AJ3l4-b" /></figure><p>💡 <em>Lesson learned:</em> Security is the foundation. Even small mistakes can be costly in the cloud.</p><h3>Ward Your Data</h3><p>I created my first <strong>S3 bucket</strong> and uploaded my project files securely.</p><h3>✅ What I Accomplished</h3><ul><li>Created a <strong>private Amazon S3 bucket</strong> (with “Block Public Access” enabled)</li><li>Verified that files are <strong>inaccessible publicly (Access Denied = ✅ security)</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*gKG-fo4MLrUivW-p" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*v1wK8OAuuNu_NiK6" /></figure><p>💡 <em>Aha moment:</em> Default privacy settings in S3 aren’t just a suggestion — they’re essential for protecting your data.</p><h3>📝 Level Up Your Code</h3><p>Time to structure my project:</p><ul><li>Organized <strong>HTML, CSS, and assets</strong></li><li>Applied <strong>best practices for project structure</strong></li><li>Uploaded everything to S3</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*lbn8AoGc87ZbK0Lo" /></figure><ul><li>🖥️ Personal website customized</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/885/0*I9bcAZ89bJd0fKKR" /></figure><p>💡 <em>Lesson learned:</em> Clean, organized files make everything else easier — from deployment to debugging.</p><h3>🌍 Global Distribution</h3><p>I set up <strong>CloudFront</strong> to make my website fast and accessible worldwide.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*UlR3OCHP4pbxcHS-" /></figure><h3>✅ What I Accomplished</h3><ul><li>Understood why <strong>distance = latency</strong> (ping problem in real life 🌐)</li><li>Learned how <strong>Content Delivery Networks (CDNs)</strong> solve the speed problem</li><li>Created a <strong>CloudFront distribution</strong> connected to my private S3 bucket</li><li>Configured <strong>index.html as default root object</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*_L95Eqh45bS1DHoN" /></figure><p>💡 <em>Aha moment:</em> CloudFront doesn’t just speed up content — it also secures access to your S3 bucket.</p><h3>🚀 Pro Deploy Mode</h3><p>I linked <strong>GitHub → AWS Amplify</strong> for automated deployments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*C_FZPM0P_bWCv5oj" /></figure><p>💡 <em>Lesson learned:</em> Automation is a game-changer. Every change in GitHub now instantly updates my live site.</p><h3>🎯 Tap Into the Rift</h3><p>I built my first <strong>serverless Lambda function</strong> to fetch real-time Riot Games data:</p><ul><li>Secured my <strong>API key with Parameter Store</strong></li></ul><p><strong>Store API Key Securely</strong></p><ul><li>Use <strong>AWS Systems Manager → Parameter Store</strong></li><li>Store as /rift-rewind/riot-api-key (type: SecureString)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*qIL_SvYllXwLpUSW" /></figure><ul><li>Displayed <strong>summoner stats &amp; top champions</strong> on my site</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*HIqP7i0lYKY_-Kdz" /></figure><p><strong>Create Function URL</strong></p><ul><li>Go to Lambda → Function URL → Create</li><li>Auth: NONE (so frontend can call it)</li><li>Configure <strong>CORS</strong>:</li><li>Allowed origin = your Amplify site domain (must match exactly with <a href="https://...)/">https://...)</a></li><li>Allow headers: Content-Type</li><li>Allow methods: POST</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*L9h83wbXFNXuxfzr" /></figure><p><strong>Test with a small event</strong>:</p><ul><li>In Lambda Console → Test → Use event:</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*2Qt1Rs7cMM_-rPkx" /></figure><ul><li>Use <strong>CloudWatch logs</strong> for debugging Lambda errors</li><li>Riot dev keys expire in 24h → regenerate if broken</li><li>If you see it 403 Forbidden, it’s usually a <strong>CORS or expired API key issue</strong></li></ul><p>💡 <em>Aha moment:</em> Serverless computing feels like magic, you get full backend functionality without maintaining a server.</p><ul><li>I levelled up my cloud and development skills: I started by setting up AWS securely with MFA and budget alerts, then stored files safely in S3. I organised my project like a pro, made it globally accessible with CloudFront, and automated deployment using GitHub and Amplify. Finally, I integrated the Riot Games API with AWS Lambda to display live League of Legends data — all while following best practices for security, structure, and performance.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/889/0*J1zK29URU7hQEiZL" /></figure><h3>🏆 Victory Lap</h3><p>I reflected on my journey and published my story:</p><ul><li>Shared screenshots of AWS services in action</li><li>Documented lessons, tips, and personal experiences.</li></ul><h3>🎯 Goal for Today</h3><p>Celebrate completing the challenge and <strong>share my journey</strong> with the AWS Builder community by writing an article on <strong>AWS Builder Center</strong>.</p><h3>🔑 Key Takeaways</h3><ol><li>Security first: MFA &amp; IAM aren’t optional.</li><li>Organization matters: structure saves headaches.</li><li>Automation rocks: GitHub + Amplify = effortless deployment.</li><li>Global distribution matters: CloudFront = fast, secure, worldwide.</li><li>Serverless is powerful: Lambda + APIs = endless possibilities.</li></ol><h3>💡 Tips for Future Builders</h3><ul><li>Keep AWS security top of mind</li><li>Organize files from the start</li><li>Automate deployments with Amplify &amp; GitHub</li><li>Test API integrations in small steps</li><li>Experiment boldly — serverless lets you fail fast and learn faster</li></ul><h3>🌟 Final Thoughts</h3><p>In just 7 days, I transformed from an AWS beginner to a cloud builder:</p><ul><li>Deployed globally distributed apps</li><li>Integrated real-time APIs with serverless functions</li><li>Implemented professional security practices</li><li>Built a portfolio-ready project</li></ul><p>The Rift Rewind Challenge wasn’t just about building a website — it was about <strong>learning cloud architecture, modern workflows, and practical AWS skills</strong>. And the best part? My live project now stands as proof of what I can build.</p><ul><li>Riot ID in the format GameName#TAG correct region</li><li>Should display <strong>Summoner Level + Top 3 Champs</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*R03N0Bbg-Ttz209d" /></figure><p><em>In this article, I share my 7-day journey building a cloud-powered Rift Rewind project using AWS, integrating live Riot Games data, and adding my personal design twists to create a fully interactive gaming platform.</em> Check out the full project on the <a href="https://github.com/dineshrajdhanapathyDD/rift-rewind-project">GitHub Repo </a>.</p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments — I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌</strong> <a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻</strong> <a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p>🛡️<a href="https://medium.com/@dineshrajdhanapathy"><strong>Medium</strong></a></p><p><strong>🐙</strong> <a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><p>🌍<a href="https://builder.aws.com/community/@dd1"> <strong>AWS Builder Profile</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0485bf2c46cb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Getting Started with AWS Developer Tools for Modern DevOps & CI/CD]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/getting-started-with-aws-developer-tools-for-modern-devops-ci-cd-8a87f28a71ea?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/8a87f28a71ea</guid>
            <category><![CDATA[devtools]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[mcp-server]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sat, 23 Aug 2025 12:00:41 GMT</pubDate>
            <atom:updated>2025-09-09T07:03:04.539Z</atom:updated>
            <content:encoded><![CDATA[<p>In today’s cloud-native world, <strong>DevOps</strong> and <strong>CI/CD (Continuous Integration and Continuous Delivery)</strong> have become essential for faster and more reliable software delivery. <a href="https://docs.aws.amazon.com/dtconsole/"><strong>AWS Developer Tools</strong></a> offer a robust set of managed services designed to help developers and DevOps teams <strong>build, test, and deploy applications securely and at scale</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-pXm6d0JMCVqte9WGlP--Q.png" /></figure><p>Whether you’re a beginner in the cloud or looking to streamline your software lifecycle, this guide introduces the core <a href="https://docs.aws.amazon.com/dtconsole/">AWS Dev Tools</a> and related services that support efficient application delivery.</p><p>As part of the <a href="https://aws.amazon.com/developer/community/community-builders/">AWS Community Builders program</a>, I had the exciting opportunity to dive deep into the <a href="https://docs.aws.amazon.com/dtconsole/"><strong>AWS Developer Tools</strong></a><strong> (DevTools)</strong> track. This experience has allowed me to gain <strong>hands-on exposure</strong> to some of the most powerful tools AWS provides to build, test, and deploy software in a scalable, secure, and automated way.</p><p>In AWS, <a href="https://aws.amazon.com/blogs/developer/"><strong>DevTools</strong></a><strong> blog (Developer</strong> <strong>Tools</strong>) refer to services that help developers and DevOps teams <strong>build, test, and deploy</strong> applications quickly and securely. These tools are especially useful for automating CI/CD pipelines and managing infrastructure as code. In this article, I’d like to share my <strong>key takeaways, real-world examples, and insights</strong> from working with these services.</p><h4>What are Developer Tools?</h4><p><strong>Developer tools</strong> are technologies that simplify and speed up software development. They act as a bridge between real-world problems and computer systems by using programming languages, frameworks, and platforms to abstract complexity. Instead of low-level coding, developers use tools like libraries, APIs, and software services to build solutions more efficiently.</p><p><strong>Benefits of Developer Tools :</strong></p><p>Developer tools help software teams code, test, deploy, and monitor applications more efficiently. Key benefits include:</p><ul><li><strong>Better Code Quality</strong>: Standardized libraries and frameworks help write clean, high-performance code.</li><li><strong>Fewer Errors</strong>: Debugging and testing tools catch bugs early, reducing costly production issues.</li><li><strong>Cross-Platform Development</strong>: Tools enable building apps that run across different devices using a single codebase.</li><li><strong>Faster Delivery</strong>: Automation and CI/CD tools speed up development and release cycles.</li><li><strong>Improved Collaboration</strong>: Version control and code review tools support teamwork on complex projects.</li><li><strong>Lower Costs</strong>: Reusable components and DevOps practices reduce development time and expenses.</li></ul><h3>What are some types of developer tools?</h3><p>You use different types of tools to support development workflow. Each tool is often specific to a programming language, platform, or purpose. The following are several common development tools.</p><h4>Coding Tools</h4><p>These tools help you write, compile, and build code efficiently. Examples include compilers, builders, libraries, frameworks, and SDKs that accelerate development.</p><h4>Programming Support Tools</h4><p>Support tools like IDEs and version control systems streamline workflows, manage projects, and enable collaboration among developers working on the same codebase.</p><h4>DevOps Tools</h4><p>DevOps tools connect development and operations, enabling faster releases and continuous feedback. They support CI/CD pipelines, infrastructure as code (IaC), and real-time monitoring.</p><h4>Software Testing Tools</h4><p>Testing tools identify bugs, performance issues, and security risks. They include debuggers, profilers, bug trackers, and security scanners like SAST and DAST.</p><p><strong>How AWS Supports Developer Tool Needs :</strong></p><p>AWS offers a suite of developer tools to help you code, build, test, and deploy applications faster in the cloud. You can:</p><ul><li>Use IDEs, SDKs, and CLI tools for seamless AWS integration</li><li>Automate DevOps workflows with CI/CD pipelines</li><li>Debug and manage code directly from your editor or browser</li><li>Get ML-powered code insights and best-practice suggestions</li><li>Monitor app performance with built-in observability tools</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gu9LnXkIQAyhr40J_ZZlmg.png" /></figure><h3>🔧 Core AWS Developer Tools Services</h3><p><strong>1. AWS CodeCommit</strong> — A secure, scalable, and managed <strong>Git-based source control service</strong>, similar to GitHub or GitLab. It enables teams to collaborate on code securely within AWS.</p><p>🛠 <strong>Example</strong>: A fintech startup replaces GitHub with <strong>CodeCommit</strong> to meet security compliance. Developers push code to a private Git repo directly inside AWS, integrating it with CodePipeline for secure deployment.</p><blockquote>We needed our code to stay within AWS due to internal audit rules — CodeCommit made that easy and seamless.</blockquote><p><strong><em>Since AWS CodeCommit is no longer available to new customers, you might want to consider alternative Git repository hosting services like GitHub, GitLab, Bitbucket, or AWS CodeStar Connections with these providers.</em></strong></p><p><strong>2. AWS CodeBuild — A</strong> fully managed <strong>build service</strong> that compiles source code, runs tests, and produces build artifacts ready for deployment. You don’t need to manage your own build servers.</p><p>🛠 <strong>Example</strong>: A media company builds its video processing Lambda functions using <strong>CodeBuild</strong>, which compiles Python code, runs unit tests, and packages it — all automatically when code is pushed.</p><blockquote>Instead of managing Jenkins build agents, we just let CodeBuild handle parallel builds for all our Lambda microservices.</blockquote><p><strong>3. AWS CodeDeploy</strong> — Automates the <strong>deployment</strong> of applications to EC2 instances, Lambda functions, or on-prem servers, with built-in strategies like blue/green and rolling updates.</p><p>🛠 <strong>Example</strong>: A logistics firm uses <strong>CodeDeploy</strong> to update EC2 instances running tracking software, using <strong>blue/green deployments</strong> to minimize downtime.</p><blockquote>We reduced downtime from 10 minutes to under 1 minute by using CodeDeploy’s blue/green strategy.</blockquote><p><strong>4. AWS CodePipeline</strong> — CI/CD orchestration tool that automates the build, test, and deploy phases.A continuous delivery service that helps you <strong>automate the entire release process</strong> — from build and test to deployment.</p><p>🛠 <strong>Example</strong>: A SaaS product team configures a full CI/CD pipeline using <strong>CodePipeline</strong> with CodeCommit, CodeBuild, and CodeDeploy to automate production releases multiple times a day.</p><blockquote>No more manual releases — CodePipeline gave us a reliable way to go from pull request to production in 15 minutes.</blockquote><p><strong>5. AWS CodeStar</strong> — Unified UI to manage software development activities with integrated CI/CD. An integrated development environment for managing software projects on AWS with <strong>built-in CI/CD</strong>, issue tracking, and collaboration tools.</p><p>🛠 <strong>Example</strong>: A startup bootstraps a new microservice in under 5 minutes using <strong>CodeStar</strong>, which sets up Git repo, CI/CD pipeline, and a sample Lambda app with monitoring.</p><blockquote>For a hackathon MVP, CodeStar helped us skip all the setup and just focus on coding.</blockquote><p><strong>6. AWS Cloud9</strong> — Cloud-based IDE for writing, running, and debugging code directly in the browser.A <strong>cloud-based IDE</strong> accessible from a browser. It includes a code editor, terminal, and debugger — all pre-configured for AWS development.</p><p>🛠 <strong>Example</strong>: A remote DevOps team uses <strong>Cloud9 IDE</strong> to onboard new hires — developers log in via browser and start coding with preconfigured AWS CLI and credentials.</p><blockquote>No more messy laptop setup. We give new devs a Cloud9 URL and they’re ready in minutes.</blockquote><p><strong><em>Since AWS Cloud9 is no longer available to new customers, you might want to consider alternatives like GitHub Codespaces, Visual Studio Code with remote development, or Kiro IDEs with cloud integrations.</em></strong></p><p><strong>7. AWS CodeGuru</strong> — ML-powered code reviewer and performance profiler for applications.A machine learning-powered tool that provides <strong>automated code reviews</strong> and <strong>performance profiling</strong> to optimize applications.</p><p>🛠 <strong>Example</strong>: An enterprise uses <strong>CodeGuru Reviewer</strong> to scan pull requests and detect costly Java code inefficiencies — like unnecessary database calls in loops.</p><blockquote>It flagged a function making 1000 DB calls in a loop — we saved hours of debugging thanks to CodeGuru.</blockquote><p><strong><em>Recent updates: AWS announced that support for Amazon CodeGuru Security will end on November 20, 2025. After this date, you won’t be able to access the console, service resources, existing scans, or documentation.</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7iY9npqER78FPInqUjWfZQ.png" /></figure><h3>⚙️ Related Topics / Tools Supporting DevOps</h3><p>To support infrastructure automation and deployment, AWS also offers powerful <strong>DevOps-related services</strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bjNOoEshyTw9zwERRoRwbg.png" /></figure><p><strong>1. AWS CloudFormation</strong> — Infrastructure as Code (IaC) tool to define AWS resources in YAML/JSON.An <strong>Infrastructure as Code (IaC)</strong> service to define and provision AWS resources using JSON or YAML templates.</p><p>🛠 <strong>Example</strong>: A DevOps team uses <strong>CloudFormation templates</strong> to deploy identical infrastructure across dev, staging, and production in different AWS accounts.</p><blockquote>With one template, we set up VPCs, EC2, RDS, and IAM roles consistently — no more manual config drift.</blockquote><p><strong>2. AWS CDK (Cloud Development Kit)</strong> — Define cloud infrastructure using familiar programming languages.Define cloud infrastructure using programming languages like TypeScript, Python, Java, or C#.</p><p>🛠 <strong>Example</strong>: A Python developer uses <strong>AWS CDK</strong> to define S3, Lambda, and DynamoDB in TypeScript instead of YAML.</p><blockquote>We built infrastructure just like writing application code. CDK made it readable, reusable, and testable.</blockquote><p><strong><em>Recent updates: End of Support for Node.js 18.x: Effective November 30, 2025, AWS CDK drops support for Node.js 18.x (which reached EOL on April 30, 2025). This impacts the CLI, Construct Library, and ecosystem tools like JSII and Projen. Upgrade to an active LTS version like Node.js 22.x for security patches and new features.</em></strong></p><p><strong>3. AWS Elastic Beanstalk</strong> — Platform as a Service (PaaS) to deploy and manage applications automatically.A <strong>Platform-as-a-Service (PaaS)</strong> offering that lets you deploy and manage applications without managing infrastructure.</p><p>🛠 <strong>Example</strong>: A team deploys a Django web app using <strong>Elastic Beanstalk</strong>, letting AWS manage load balancing, autoscaling, and health checks.</p><blockquote>We focused on the app, not the infrastructure — Beanstalk auto-managed the rest.</blockquote><p><strong>4. AWS SAM (Serverless Application Model)</strong> — Framework for building serverless applications.A framework to build, test, and deploy <strong>serverless applications</strong> using simplified syntax.</p><p>🛠 <strong>Example</strong>: A team creates and tests a serverless application (API Gateway + Lambda + DynamoDB) locally using the <strong>AWS SAM CLI</strong>.</p><blockquote>We simulated API calls locally and deployed serverless stacks with one command using SAM.</blockquote><p><strong>5. AWS Amplify</strong> — Full-stack development platform for frontend + backend apps.A full-stack platform for developing web and mobile applications with built-in <strong>backend support</strong> (Auth, API, storage).</p><p>🛠 <strong>Example</strong>: A solo developer builds a React app with Amplify Hosting, Auth, GraphQL API, and file storage — without managing backend servers.</p><blockquote>Amplify let me go from idea to full-stack app in 2 days — all hosted and integrated with CI/CD.</blockquote><p>Recently, I completed a project using AWS Amplify — a powerful service to build, deploy, and host full-stack applications with ease. With just a few steps, Amplify lets you connect your frontend, backend, and CI/CD for seamless app hosting.</p><p>👉 Check out my project and explore how you can host your own apps with Amplify: <a href="https://builder.aws.com/content/31idxVxBul5kfR34ms8D0ru4LGM/building-my-serverless-portfolio-website-on-aws"><strong><em>building-my-serverless-portfolio-website-on-aws</em></strong></a></p><p><strong>6. AWS X-Ray</strong> — Distributed tracing system for debugging and analyzing microservices-based apps.A distributed <strong>tracing tool</strong> for debugging microservices and understanding application performance bottlenecks.</p><p>🛠 <strong>Example</strong>: A payment service team traces user requests through 10+ microservices using <strong>X-Ray</strong>, pinpointing a latency issue in an SQS queue handler.</p><blockquote>We visualized the request path and found the bottleneck in under 10 minutes with X-Ray.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OexpCmH9GliIZAcQq2jlfA.png" /></figure><h3>📊 Monitoring and Logging Tools for Dev</h3><p>To ensure application health and compliance, AWS provides powerful observability tools:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*z_EMM7ruywzjbiK-tutKzA.png" /></figure><p><strong>1. Amazon CloudWatch</strong> — Monitoring and logging service.Monitors AWS resources and applications, collects logs and metrics, and triggers alerts based on thresholds.</p><p>🛠 <strong>Example</strong>: An e-commerce site tracks EC2 CPU usage and sets up <strong>CloudWatch Alarms</strong> to scale out when CPU exceeds 70%.</p><blockquote>CloudWatch and autoscaling kept our Black Friday site alive without manual intervention.</blockquote><p><strong><em>Amazon CloudWatch Evidently, a service for safely launching features and running experiments with A/B testing, is scheduled for discontinuation on October 17, 2025, after which access to the console, resources, and capabilities will end. Active users can continue using it normally until that date, with critical security patches provided but no support for limit increase requests or new features.</em></strong></p><p><strong>2. AWS CloudTrail</strong> — Tracks user activity and API usage. Tracks <strong>user activity and API calls</strong>, useful for security audits and troubleshooting.</p><p>🛠 <strong>Example</strong>: After a security incident, the IT team uses <strong>CloudTrail logs</strong> to trace who changed IAM policies and when.</p><blockquote>We identified the root cause of a permission error by inspecting the API call logs in CloudTrail.</blockquote><p><strong>3. AWS Config</strong> — Tracks AWS resource configurations and compliance over time.Monitors <strong>configuration changes</strong> to AWS resources and checks for compliance with best practices.</p><p>🛠 <strong>Example</strong>: A regulated company uses <strong>AWS Config</strong> to ensure encryption is always enabled on S3 buckets and logs violations.</p><blockquote>With AWS Config, we automatically audit every bucket for compliance and get alerts if policies change.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qSdiY5PfWwrUF5pKPenYzw.png" /></figure><h3>🧩 3rd-Party Integrations Support</h3><p>AWS Dev Tools work seamlessly with popular tools used by developers:</p><p><strong>1. GitHub / GitLab / Bitbucket integration — </strong>Easily integrate external Git repositories into CodePipeline or CodeBuild.</p><p>🛠 <strong>Example</strong>: A developer team connects <strong>GitHub PRs</strong> to <strong>CodePipeline</strong>, so every merge triggers a new build and deployment automatically.</p><blockquote>No manual steps — every GitHub merge deploys our app in minutes via CodePipeline.</blockquote><p><strong>2. Jenkins integration with AWS Dev Tools — </strong>Connect Jenkins pipelines to AWS services for hybrid workflows.</p><p>🛠 <strong>Example</strong>: A company uses <strong>Jenkins</strong> for job orchestration but offloads heavy builds to <strong>AWS CodeBuild</strong>, reducing build agent maintenance.</p><blockquote>We let Jenkins coordinate the workflow, but let AWS handle the compute for builds.</blockquote><p><strong>3. Terraform &amp; AWS</strong> — Though not AWS-native, commonly used for IaC with AWS.Not AWS-native, but widely used for Infrastructure as Code alongside AWS.</p><p>🛠 <strong>Example</strong>: A DevOps team provisions complete VPC, ECS, and RDS stacks across multiple environments using <strong>Terraform modules</strong>.</p><blockquote>Terraform gave us reproducible, version-controlled infrastructure across regions and accounts.</blockquote><p><strong>4. Docker &amp; Container Registries</strong>: Integrate with Amazon ECR for storing and pulling container images.</p><p>🛠 <strong>Example</strong>: A SaaS company packages its microservices as <strong>Docker containers</strong> and pushes them to <strong>Amazon Elastic Container Registry (ECR)</strong>. These images are then pulled automatically by <strong>ECS and EKS clusters</strong> during deployment.</p><blockquote>“Every time we push a new Docker image to ECR, our CI/CD pipeline deploys it to production in minutes — no more manual uploads or broken versions.”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HYhmKCzCz_PGaWk_uYDpaw.png" /></figure><h3>Conclusion</h3><h4>Why AWS DevTools Matter</h4><p>Whether you’re setting up your first CI/CD pipeline or optimising an enterprise-grade DevOps workflow, <strong>AWS Developer Tools offer the building blocks</strong> you need to succeed. With native integrations, scalability, and security, you can automate your entire development lifecycle on the cloud.</p><p><strong><em>Explore MCP (Model Context Protocol) can be used to access and work with AWS documentation of </em></strong><a href="https://github.com/dineshrajdhanapathyDD/AWSCB-DD/blob/main/Devtools%20Article%20ref/Devtool%20overview.md"><strong><em>Dev Tools overview</em></strong></a><strong><em> using kiro IDE, making developer tools more efficient and improving the way engineers discover, integrate, and use cloud resources .</em></strong></p><p><strong><em>Along with accessing documentation, MCP can also integrate with tools like AWS Diagrams to automatically generate architecture visuals. This allows developers to pull AWS service icons directly from documentation, quickly create diagrams without leaving the IDE, and keep their documentation, configurations, and visuals in sync. By combining Amazon Q, MCP, and AWS Diagrams, teams can code, reference AWS docs, and generate architecture diagrams in one place check my blog </em></strong><a href="https://medium.com/@dineshrajdhanapathy/architecture-diagram-local-development-with-amazon-q-wsl-ubuntu-mcp-server-for-application-ddd424b0c14c"><strong><em>here</em></strong></a></p><p>Working with AWS DevTools helped me understand how <strong>modern CI/CD, DevOps, and Infrastructure as Code</strong> can come together in a unified, secure, and scalable way.</p><blockquote>The more I explored AWS Developer Tools, the more I realized they aren’t just for big teams — they’re perfect for anyone who wants to build reliable software delivery systems.</blockquote><p><strong><em>Stay tuned for my upcoming project on this topic — I’ll be sharing a hands-on article soon.</em></strong></p><p>Being part of the <strong>AWS Community Builder program</strong> gives me the opportunity to explore cutting-edge <strong>developer tools</strong> and deepen my knowledge in <strong>cloud and DevOps</strong>. It’s a great way to learn, build, and grow with support from the AWS community.</p><p>AWS Community Builders program gave me:</p><ul><li>🎓 Learning resources and mentorship</li><li>🌍 Global community to share and grow</li></ul><p>All <strong>AWS</strong> builders are welcome and encouraged to apply 👇</p><p><strong>Add your name to waitlist for upcoming years</strong> — <a href="https://aws.amazon.com/developer/community/community-builders/"><strong><em>AWS Community Builders</em></strong></a></p><p>Check before article here : <a href="https://medium.com/@dineshrajdhanapathy/how-i-started-with-aws-and-became-an-aws-community-builder-89d894e75306"><strong>How I Started with AWS and Became an AWS Community Builder</strong></a></p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments — I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌ </strong><a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻 </strong><a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p><strong>🐙 </strong><a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8a87f28a71ea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an Interactive Office Scene with Pure CSS]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/building-an-interactive-office-scene-with-pure-css-659693271d7d?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/659693271d7d</guid>
            <category><![CDATA[js]]></category>
            <category><![CDATA[challenge]]></category>
            <category><![CDATA[css]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[front-end-development]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Sat, 09 Aug 2025 05:07:44 GMT</pubDate>
            <atom:updated>2025-08-09T05:07:44.303Z</atom:updated>
            <content:encoded><![CDATA[<h3>Exploring Creativity, Accessibility, and Performance—Without Relying on JavaScript</h3><h3>👋 Introduction</h3><p>I set out on a creative challenge: <strong>Can you build a fully interactive, visually engaging office environment using just HTML and CSS—with minimal JavaScript?</strong> The answer turned into one of the most fun and rewarding projects I’ve worked on.</p><h3>🌟 Inspiration</h3><p>The idea sparked during a quiet evening while working remotely. I glanced around my real desk — sticky notes, a flickering terminal, a half-full coffee mug, and a wall clock ticking past midnight — and thought, <em>“What if I could bring this physical workspace into code?”</em><br> That blend of physical presence and digital interaction inspired me to recreate the essence of a modern workplace using nothing but frontend code.</p><p>This interactive CSS art piece simulates a <strong>modern digital workspace</strong> and demonstrates how far CSS has come—from simple styling to complex interaction and animation. Inspired by <strong>human-centred design</strong> and <strong>digital collaboration tools</strong>, this project blends <strong>visual storytelling</strong> with <strong>technical innovation</strong>.</p><h3>What I Built</h3><p>I created an interactive CSS art piece that depicts a complete office environment using only HTML, CSS, and minimal JavaScript. The project showcases advanced CSS techniques while telling the story of modern office culture through visual elements and animations. Just click the pop-up to start exploring.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1j-uRkduWrm_4-gAUiMVFw.png" /></figure><p><strong>Key Features:</strong></p><ul><li><strong>Interactive Office Elements</strong>: Coffee mug with steam animation, clickable monitor with cycling code snippets, water cooler conversations</li><li><strong>Time-Based Simulation</strong>: Automatic day/night cycle that affects lighting and office activity levels</li><li><strong>Full Accessibility</strong>: Keyboard navigation, screen reader support, and motion preference handling</li><li><strong>Responsive Design</strong>: Optimized for desktop, tablet, and mobile with touch-friendly interactions</li><li><strong>Advanced CSS Architecture</strong>: Custom properties system, hardware acceleration, and modular organization</li></ul><h3>Demo</h3><p>The interactive office scene includes:</p><p>🖥️ <strong>Dynamic Monitor</strong>: Click to cycle through different code snippets and terminal outputs<br>☕ <strong>Smart Coffee Mug</strong>: Interactive steam animation with temperature states that change on click<br>👥 <strong>Character Conversations</strong>: Water cooler characters with rotating dialogue systems<br>🕐 <strong>Office Clock</strong>: Click to advance time and trigger day/night transitions<br>📝 <strong>Interactive Notes</strong>: Sticky notes and notice board items with hover effects<br>🌱 <strong>Office Plants</strong>: Animated leaves that respond to user interaction</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*es0shdgPNlNiOZ-n96Q6Jw.png" /></figure><p><strong>Try it yourself</strong>: Hover over elements, use Tab to navigate with the keyboard, or click the help button (?) for technical details.</p><p><em>Want to explore the code? Check out the </em><a href="https://github.com/dineshrajdhanapathyDD/CSS-Art---challenge.git"><em>GitHub repository</em></a><em> or click the help button (?) in the demo to see the technical implementation details.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gecVi6Zy-mD-XfhpkykPpg.png" /></figure><h3>Journey</h3><h3>The Challenge</h3><p>I wanted to push the boundaries of what’s possible with pure CSS while creating something that felt alive and interactive. The goal was to build a complete office ecosystem that responds to user interaction without relying on heavy JavaScript frameworks.</p><h3>Technical Approach</h3><p><strong>1. CSS Architecture Revolution</strong></p><pre>:root {<br>  /* Theme System */<br>  --primary-hue: 210;<br>  --saturation: 50%;<br>  --theme-primary: hsl(var(--primary-hue), var(--saturation), var(--lightness));<br>  <br>  /* Animation System */<br>  --animation-speed-multiplier: 1;<br>  --animation-duration-base: 0.3s;<br>  <br>  /* Time-Based Variables */<br>  --time-of-day: 1; /* 0 = night, 1 = day */<br>  --ambient-light: calc(0.7 + (var(--time-of-day) * 0.3));<br>}</pre><p>I built a comprehensive custom properties system that acts like a design system, allowing for dynamic theming and consistent animations across all components.</p><p><strong>2. Advanced Animation Techniques</strong><br>The coffee mug steam uses multiple pseudo-elements with staggered animations:</p><pre>.coffee-mug__steam::before,<br>.coffee-mug__steam::after {<br>  animation: steam var(--coffee-steam-duration) infinite ease-out;<br>  box-shadow: <br>    -3px -10px 0 -2px rgba(255, 255, 255, 0.3),<br>    3px -15px 0 -1px rgba(255, 255, 255, 0.2);<br>}</pre><p><strong>3. Accessibility-First Design</strong><br>Every interactive element includes proper ARIA labels and keyboard navigation:</p><pre>&lt;div class=&quot;coffee-mug interactive-element&quot; <br>     tabindex=&quot;0&quot; <br>     role=&quot;button&quot; <br>     aria-label=&quot;Coffee mug - click to interact&quot;&gt;</pre><p><strong>4. Performance Optimization</strong><br>Hardware acceleration and CSS containment ensure smooth 60fps animations:</p><pre>.animated-element {<br>  will-change: transform;<br>  transform: translateZ(0);<br>  contain: layout style paint;<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-ElD8cxXcCWfUxPLowGXow.png" /></figure><h3>Key Learnings</h3><p><strong>CSS Can Do More Than You Think</strong>: By combining custom properties, advanced selectors, and modern CSS features, you can create complex interactive systems without JavaScript.</p><p><strong>Accessibility Enhances Creativity</strong>: Adding proper keyboard navigation and screen reader support didn’t limit the design — it made it more robust and usable for everyone.</p><p><strong>Performance Matters</strong>: Using CSS containment and hardware acceleration made the difference between choppy animations and smooth 60fps interactions.</p><p><strong>Modular Architecture Scales</strong>: Organizing CSS with clear sections and comprehensive custom properties made the 1000+ line stylesheet maintainable and extensible.</p><h3>The Result</h3><p>The final project demonstrates that CSS has evolved into a powerful tool for creating interactive experiences. It’s not just about styling anymore — it’s about building complete user interfaces that are performant, accessible, and engaging.</p><p><strong>Technical Stats:</strong></p><ul><li>1000+ lines of organized CSS</li><li>14 interactive elements</li><li>4 time-based states</li><li>Full keyboard navigation</li><li>Mobile-responsive design</li><li>90+ Lighthouse performance score</li></ul><p>This project proves that with modern CSS techniques, you can create rich, interactive experiences that rival JavaScript-heavy applications while maintaining better performance and accessibility.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cFH01llFkV_MQfbQaEhoIQ.png" /></figure><p><strong>Special thanks to Axero</strong> for inspiring this project. Their commitment to <strong>human-centered design</strong> and <strong>digital collaboration</strong> continues to shape the way we envision connected, <strong>thriving workplaces</strong>.</p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments — I’d love to hear from you.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=659693271d7d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Idea to Infrastructure: How Kiro AI Agents Empower My Portfolio Websites Build Smarter]]></title>
            <link>https://medium.com/@dineshrajdhanapathy/from-idea-to-infrastructure-how-kiro-ai-agents-empower-my-portfolio-websites-build-smarter-2b906c193c2b?source=rss-2d8438cdd725------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b906c193c2b</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[kiro]]></category>
            <category><![CDATA[website]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Dineshrajdhanapathy]]></dc:creator>
            <pubDate>Mon, 28 Jul 2025 15:15:49 GMT</pubDate>
            <atom:updated>2025-08-09T18:41:57.762Z</atom:updated>
            <content:encoded><![CDATA[<p>One of the most rewarding aspects of being part of the <a href="https://medium.com/@dineshrajdhanapathy/how-i-started-with-aws-and-became-an-aws-community-builder-89d894e75306"><strong>AWS Community Builders program</strong></a> is the chance to gain early access to innovative tools , and today, I’m excited to share one that’s now available in public preview: <a href="https://kiro.dev/"><strong>Kiro</strong></a>.</p><p>Through this program, I’ve had the opportunity to attend exclusive webinars, interact with AWS professionals, and learn directly from builders across the globe. These experiences have not only expanded my technical knowledge but also given me a broader perspective on how products are built, used, and evolved.</p><p><a href="https://kiro.dev/"><strong>Kiro</strong></a> is a perfect example of that journey, an agentic development environment that pushes the boundaries of productivity and redefines how we interact with code.</p><p>I’ve been testing <a href="https://kiro.dev/"><strong>Kiro</strong></a> extensively, and while it’s still in its early stages, there’s no doubt that it’s already <strong>supercharged my productivity</strong> as a cloud and DevOps engineer portfolio. Compared to traditional IDEs and even other agentic IDEs I’ve explored, Kiro stands out—not just as a <strong>VS Code fork</strong>, but as a <strong>whole new development experience</strong>.</p><p>Then I found <a href="https://kiro.dev/"><strong>Kiro</strong></a>—a new kind of AI development agent that didn’t just write code. It collaborated. It adapted. And it helped me turn my idea into a working website using <strong>intelligent agent architecture and task list</strong>, powered by hooks and steering inputs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VeOFruq99OQY_BUkpa1-yg.png" /></figure><p>One month ago, I found myself staring at a blank terminal. I had a clear vision: build my own website to showcase my skills as a <strong>Cloud and DevOps engineer</strong> — from automated pipelines to smart infrastructure demos. I planned it all out: a sleek, professional design, responsive layout, project filtering, SEO, accessibility, and high performance. But turning that idea into a <strong>production-ready, secure, and scalable site</strong> quickly became overwhelming. Despite my backend expertise, I found myself spending more time tweaking config files and solving deployment quirks than actually shipping features. What started as a portfolio turned into a real-world challenge — one that pushed me to rethink my own workflow.</p><p>Enter <strong>Kiro</strong>, an agentic AI IDE that doesn’t just code — it <strong>understands your intent</strong>, <strong>steers toward your goals</strong>, and <strong>hooks into your logic</strong>. That’s when things changed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/277/1*uX-lOFsMgEqS2OyATTvMDg.png" /><figcaption>kiro functions</figcaption></figure><h3>What is Kiro?</h3><p>Kiro is an AI assistant and IDE built specifically for developers. It combines the power of artificial intelligence with practical development tools to streamline your coding process. Unlike traditional IDEs, Kiro can understand natural language instructions, recommend code changes, and help troubleshoot issues in real-time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3Vg_DTFcgpIsIRyTiHY8bQ.png" /></figure><h4>Complexity with Spec-Driven Development</h4><p>Modern development often drowns in chaos, unclear requirements, scattered documentation, and disconnected tasks. <strong>Kiro changes that narrative</strong> with a <strong>spec-driven development model</strong> that brings structure to your workflow. It transforms your initial prompt into <strong>clear requirements</strong>, <strong>system designs</strong>, and <strong>actionable tasks, </strong>all in one place.</p><p>Whether you’re planning a microservice or a full-stack app, Kiro lets you <strong>collaborate directly within the spec pane</strong>, aligning on architecture and goals before writing a line of code. Once finalized, <strong>Kiro’s intelligent agents</strong> can implement the spec autonomously, while you stay in control of every decision.</p><h4>Automate Smarter with Agent Hooks</h4><p>With <strong>agent hooks</strong>, Kiro allows you to automate repetitive and context-aware tasks. You can <strong>delegate operations to background AI agents</strong> that trigger on specific events, like saving a file. These agents can generate <strong>unit tests</strong>, improve <strong>code efficiency</strong>, or even <strong>update documentation, </strong>all without breaking your flow.</p><p>By using agent hooks, you scale your development process without scaling your mental overhead. The agents work behind the scenes while you stay focused on what matters most, <strong>shipping great features</strong>.</p><h3>Kiro’s Spec-Driven Workflow</h3><p>Kiro bridges the gap between <strong>product ideas</strong> and <strong>technical execution</strong> with a structured, spec-driven approach. Each feature begins with a dedicated spec folder containing three essential files:</p><ul><li><strong>requirements.md</strong>: Captures clear, testable user stories using EARS (Easy Approach to Requirements Syntax), This structured approach brings clarity, testability, and traceability to your workflow. By turning vague ideas into clear, testable requirements, <strong>Kiro reduces miscommunication and accelerates development</strong> — ensuring alignment between product and engineering from day one.</li><li><strong>design.md</strong>: Outlines system architecture, data flow, interfaces, and testing strategy—providing a collaborative space for visualizing complex interactions.</li><li><strong>tasks.md</strong>: Breaks the work into discrete, trackable tasks with real-time status updates, ensuring smooth progress tracking and accountability.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bFVazL8F1nWU7Qs53SkIuA.png" /></figure><p>This structured format reduces ambiguity, improves alignment between product and engineering, and streamlines the entire development lifecycle — from feature idea to production-ready code.</p><h3>Automate with Precision: Agent Hooks in Kiro</h3><p>In a world where efficiency defines velocity, <strong>Kiro’s Agent Hooks</strong> bring automation to your IDE in a seamless, intelligent way. Rather than manually initiating routine tasks like linting, generating tests, or updating documentation, <strong>Agent Hooks respond to events like file saves or new file creation, executing predefined AI-powered actions behind the scenes</strong>.</p><h4>What Are Agent Hooks?</h4><p>Agent Hooks are <strong>event-driven triggers</strong> in your development workflow. When an event occurs, such as saving, creating, or deleting a file, Kiro automatically activates an agent that follows your prompt and performs the associated task. Imagine never having to remind yourself to generate unit tests again. With hooks, that’s your new normal.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4Fqt0DXg7_N9VWD2CPfw5w.png" /><figcaption>Hook created</figcaption></figure><h4>Why Use Agent Hooks?</h4><p>By automating repetitive actions, Agent Hooks help you:</p><ul><li>🧹 <strong>Maintain consistent code quality</strong></li><li>🛡️ <strong>Prevent security vulnerabilities</strong></li><li>⏱️ <strong>Reduce manual overhead</strong></li><li>🧑‍🤝‍🧑 <strong>Standardize team workflows</strong></li><li>⚡ <strong>Accelerate development cycles</strong></li></ul><h3>How Agent Hooks Work</h3><p>Kiro’s automation system runs in three simple steps:</p><ol><li><strong>Event Detection</strong>: Monitors IDE activity for triggers (e.g. file save)</li><li><strong>Prompt Execution</strong>: Sends your predefined natural language prompt to the agent</li><li><strong>Automated Action</strong>: The agent executes the task autonomously</li></ol><h3>Guide the AI with Confidence: Steering in Kiro</h3><p>One of Kiro’s most powerful features isn’t flashy—it’s smart, subtle, and scalable. Meet <strong>Steering</strong>, Kiro’s mechanism for persistent project context. Instead of repeating yourself in every conversation, <strong>Steering ensures Kiro remembers your conventions, standards, and architecture </strong>automatically.</p><h4>What Is Steering?</h4><p>Steering is Kiro’s way of building long-term memory around your project using markdown files located in .kiro/steering/. These files act as blueprints that teach Kiro <strong>how your project works</strong>, what libraries you use, and how your team writes code. It’s like having a senior engineer embedded in your IDE who never forgets the rules of your codebase.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VZrR-GAp6u88bjgKwnRT8w.png" /></figure><h3>Why Use Steering in Kiro?</h3><p><strong>Steering</strong> in Kiro isn’t just a configuration — it’s your project’s memory, blueprint, and quality enforcer all in one. Here’s why it’s a game-changer:</p><p>✅ <strong>Consistent Code Generation</strong><br> Every API, test, or component adheres to your team’s conventions — automatically.</p><p>🔁 <strong>Reduced Repetition</strong><br> You don’t need to keep explaining your preferences. Kiro learns and remembers.</p><p>🤝 <strong>Team Alignment</strong><br> Onboard new developers effortlessly. Kiro enforces your standards from day one.</p><p>📚 <strong>Scalable Project Knowledge</strong><br> As your project evolves, your documentation grows with it — structured, searchable, and agent-aware.</p><p>With <strong>Steering</strong>, your codebase becomes cleaner, your team becomes faster, and your decisions scale with confidence.</p><h3>🌐 Model Context Protocol (MCP): Supercharging Kiro with Specialized Context</h3><p>Modern development doesn’t happen in isolation — it thrives on context, references, and real-time insight. That’s where <strong>Model Context Protocol (MCP)</strong> comes in. By extending Kiro’s AI with MCP, you’re not just coding — you’re tapping into a live stream of domain-specific intelligence.</p><h4>🔎 What is MCP?</h4><p>The <strong>Model Context Protocol (MCP)</strong> is a communication standard that allows <strong>Kiro</strong> to connect with external servers packed with <strong>specialized knowledge, tools, and APIs</strong>. These MCP servers act like power-ups for your AI workflow — feeding it relevant data and making it context-aware.</p><p>Imagine having <strong>AWS documentation</strong>, <strong>API specs</strong>, or even <strong>design system rules</strong> instantly accessible by your assistant — MCP makes that possible.</p><h4>What Can You Do With MCP?</h4><p>When paired with Kiro, MCP unlocks the following superpowers:</p><ul><li><strong>🔍 Live Documentation Access</strong><br> Browse and retrieve AWS Docs, OpenAPI specs, and developer references <em>within the IDE.</em></li><li><strong>🤖 AI with Domain Awareness</strong><br> Instead of general answers, Kiro can respond using <em>accurate, official sources</em> from your custom MCP server.</li><li><strong>💡 Contextual Code Recommendations</strong><br> Based on the connected service, Kiro can auto-suggest configurations, cloud commands, best practices, and more.</li><li><strong>🧰 External Tool Integration</strong><br> You can hook in documentation engines, dev portals, internal wikis, or any server that supports MCP’s JSON protocol.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S2c3BeGdoauh3fP7iW4nLw.png" /><figcaption>mcp server -aws docs</figcaption></figure><h4>🧱 Real-World Example: AWS MCP Server</h4><p>Let’s say you’re building an app on AWS. With the <strong>AWS Documentation MCP server</strong>, you can:</p><ul><li>Ask Kiro “How does S3 versioning work?”</li><li>Get code snippets with AWS SDK best practices</li><li>Auto-complete IAM policy JSON</li><li>Generate Terraform or CloudFormation templates, instantly</li></ul><p>All without leaving your IDE or flipping tabs.</p><h4>Why MCP Matters</h4><p>MCP transforms Kiro from a <strong>smart code assistant</strong> into a <strong>context-aware development ally</strong>. Whether you’re learning cloud tech, building internal tools, or scaling infrastructure, you get AI help <em>informed by your real-world environment.</em></p><h3><strong>Personal Portfolio Built</strong></h3><p>This is a modern, responsive personal portfolio website built for GitHub Pages hosting. The project showcases professional work, skills, and contact information using a clean, performance-focused architecture.</p><ul><li>Create a professional online presence for career opportunities</li><li>Showcase technical projects with detailed case studies</li><li>Demonstrate modern web development skills and best practices</li><li>Provide multiple contact channels for potential employers/collaborators</li><li>Achieve excellent performance and accessibility scores</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bFVazL8F1nWU7Qs53SkIuA.png" /><figcaption>Task list</figcaption></figure><p><strong>How It’s Built<br>Core Technologies:</strong></p><p><strong>Static HTML5/CSS3/JavaScript </strong>— Pure web technologies for GitHub Pages compatibility<br><strong>Jekyll</strong> — Optional static site generator for GitHub Pages integration<br><strong>Modern CSS</strong> — CSS Grid, Flexbox, Custom Properties for responsive design<br><strong>Vanilla JavaScript</strong> — ES6+ modules for interactive features</p><p><strong>Key Features:</strong></p><p><strong>Responsive Design</strong> — Mobile-first approach with CSS Grid layouts<br><strong>Project Showcase</strong> — Dynamic filtering and search functionality<br><strong>Contact Integration</strong> — Forms with third-party service handling (Formspree)<br><strong>Performance Optimized</strong> — Lazy loading, image optimization, minified assets<br><strong>Accessibility First </strong>— WCAG AA compliance with keyboard navigation and screen reader support<br><strong>SEO Optimized </strong>— Structured data, meta tags, and sitemap generation</p><p><strong>Architecture:</strong></p><p><strong>Component</strong>-based CSS using BEM methodology<br><strong>Modular JavaScript</strong> with ES6 modules<br><strong>File-based content management</strong> through YAML data files<br><strong>Automated build pipeline</strong> with GitHub Actions and Lighthouse CI</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vsKO2iw8_nBfUjP7VPrceA.png" /><figcaption>My Profile — Dashboard</figcaption></figure><p>The project demonstrates modern web development practices while maintaining the simplicity required for static hosting, making it both performant and maintainable. check the live website here : <a href="https://dineshrajdhanapathydd.github.io/dineshrajdhanapathy/"><strong>DD-website</strong></a></p><h3>Conclusion: From Complexity to Clarity — The Kiro Way</h3><p>In a world where shipping software often feels like solving a puzzle with missing pieces, <strong>Kiro</strong> provides the structure, automation, and contextual intelligence developers have long craved. From <strong>spec-driven clarity</strong> to <strong>automated agent hooks</strong>, <strong>steering-based consistency</strong>, and <strong>MCP-powered insight</strong>, Kiro transforms development from a manual grind into an intelligent, collaborative workflow.</p><p>For me, what started as a personal project to showcase my Cloud and DevOps skills evolved into a deep dive into the future of AI-assisted development. Kiro didn’t just help me build faster — it helped me <strong>build smarter</strong>. And it’s just the beginning.</p><h3>🌐 Next Up in the Series</h3><h3>“From Spec to Site: Building a Frontend Portfolio with Kiro”</h3><p>In the next part of this series, I’ll show you how I transformed my Kiro-powered specs into a live, responsive frontend website — complete with automated code generation, component scaffolding, and spec-driven UI logic.</p><p>After weeks of refining specs, collaborating with intelligent agents, and streamlining my workflow inside Kiro, I’m excited to share the result: <a href="https://dineshrajdhanapathydd.github.io/dineshrajdhanapathy/"><strong>my live cloud and DevOps portfolio website</strong></a>, designed and built using Kiro’s agentic IDE. From layout to logic, every component was generated through spec-driven development, hooks, and steering files — allowing me to focus on what matters most: showcasing real skills, real projects, and real impact.</p><p>In the next part of this series, I’ll walk you through how this site came to life — how I turned natural language specs into production-ready frontend code, automated build flows, and shipped a polished site without touching endless config files. Whether you’re building a personal brand or launching a product, this next article will show how Kiro can turn your idea into infrastructure — fast.</p><p><strong>Stay tuned.</strong> ✨</p><p><strong>Thank you for taking the time to read my article! If you found it helpful, feel free to like, share, and drop your thoughts in the comments — I’d love to hear from you.</strong></p><p>If you want to connect or dive deeper into cloud and DevOps, feel free to follow me on my socials:</p><p>💼 <a href="https://www.linkedin.com/in/dineshrajdhanapathydd"><strong>LinkedIn</strong></a></p><p><strong>❌ </strong><a href="https://x.com/DD_Dineshraj"><strong>X (formerly Twitter)</strong></a></p><p><strong>👨‍💻 </strong><a href="https://dev.to/dineshrajdhanapathydd"><strong>DEV Community</strong></a></p><p><strong>🐙 </strong><a href="https://github.com/dineshrajdhanapathydd"><strong>GitHub</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b906c193c2b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>