<?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 Viktoriia Kholodenko on Medium]]></title>
        <description><![CDATA[Stories by Viktoriia Kholodenko on Medium]]></description>
        <link>https://medium.com/@viktoriya.kholodenko?source=rss-7af12538b108------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*oyuqvVbJJ48oLTywebZhIA.jpeg</url>
            <title>Stories by Viktoriia Kholodenko on Medium</title>
            <link>https://medium.com/@viktoriya.kholodenko?source=rss-7af12538b108------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 10:35:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@viktoriya.kholodenko/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[AI Learning Buddy: Helping Kids Turn Daily Routine into Magical Stories]]></title>
            <link>https://medium.com/@viktoriya.kholodenko/ai-learning-buddy-helping-kids-turn-daily-routine-into-magical-stories-659ca910b059?source=rss-7af12538b108------2</link>
            <guid isPermaLink="false">https://medium.com/p/659ca910b059</guid>
            <dc:creator><![CDATA[Viktoriia Kholodenko]]></dc:creator>
            <pubDate>Mon, 21 Apr 2025 01:01:02 GMT</pubDate>
            <atom:updated>2025-04-21T01:01:02.620Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BhXHMKR2ixevyGmixO0YTw.png" /></figure><p>AI Learning Buddy project was created as part of the <strong>Capstone for the 5-Day Gen AI Intensive Course</strong> by <strong>Google and Kaggle</strong>, where we explored how to apply generative AI to real-world use cases.</p><p>While taking the course, I thought about a challenge I faced at home: my daughter’s teacher recommended keeping a summer journal to improve storytelling skills. She struggled with what to write — and honestly, so did I. As a parent, it was hard to come up with new ideas every day and help her keep doing this in a fun way, not forced.</p><p>That’s when I realized: this was the perfect chance to test the power of generative AI in a real, everyday parenting situation.</p><p>And just like that, the <strong>AI Learning Buddy</strong> was born — a fun, interactive assistant that helps kids aged 5–10 turn their daily reflections into magical, personalized stories.</p><h3>The Problem</h3><p>Designing a journaling experience that’s consistent, creative, and enjoyable for both kids and parents is harder than it seems. This project explores how Gen AI can support that process in a playful, personalized way.</p><h3>How Gen AI Solves This</h3><p>To bring this idea to life, I used <strong>Google’s Gemini models</strong> via the langchain-google-genai SDK.</p><p>The assistant takes a short journal entry and turns it into an imaginative story, using the child’s name, age, and favorite characters to make it more engaging and personalised. The tone is gentle, playful, and always age-appropriate.</p><p>But the real power comes from how the assistant interacts.</p><p>Instead of just answering prompts, the AI behaves like a buddy. I used <strong>LangGraph</strong> to build an agentic flow that gives the assistant memory and structure. It remembers the child’s preferences, guides the conversation, and encourages reflection with supportive feedback.</p><h3>How I Built It</h3><p>This project isn’t just a prompt-in, story-out setup. It uses <strong>LangGraph</strong> to build a dynamic agent flow, where each interaction depends on memory, tools, and conversation history.</p><p>The system is built as a graph of three key nodes:</p><ul><li>&quot;agent&quot; – handles language model generation</li><li>&quot;tools&quot; – handles tool use like memory read/write</li><li>&quot;user&quot; – simulates user input (or uses real input if not in test mode)</li></ul><p>Here’s how I constructed the graph:</p><pre>graph_builder = StateGraph(AgentState)<br><br>graph_builder.add_node(&quot;agent&quot;, agent)<br>graph_builder.add_node(&quot;user&quot;, user)<br>graph_builder.add_node(&quot;tools&quot;, agent_tools)<br><br>graph_builder.add_conditional_edges(&quot;agent&quot;, agent_tool_condition)<br>graph_builder.add_conditional_edges(&quot;user&quot;, user_exit_condition)<br>graph_builder.add_edge(&quot;tools&quot;, &quot;agent&quot;)<br>graph_builder.add_edge(START, &quot;agent&quot;)<br><br>graph = graph_builder.compile()</pre><p>The state machine ensures every interaction flows logically, like collecting missing info, storing it in memory, and responding based on saved preferences. More details are in my <a href="https://www.kaggle.com/code/viktoriiakholodenko/gen-ai-2025-q2-capstone-project-learning-buddy">Capstone project</a></p><h3>Simulation Mode</h3><p>To test this easily, I built a <strong>simulation mode</strong>. It mimics the whole back-and-forth loop by feeding in queued user inputs and skipping the need for manual typing during development.</p><pre>if SIMULATION_CONTEXT[&quot;simulate&quot;]:<br>    user_input = SIMULATION_CONTEXT[&quot;input_queue&quot;].pop(0)<br>    print(&quot;\nUser:&quot;, user_input)<br>else:<br>    user_input = input(&quot;\nUser: &quot;)</pre><p>This made it super fast to validate logic and test different edge cases — like restarting a story, trying unknown names, or skipping the summary step.</p><h3>Prompt Design: Behaving Like a Real Buddy</h3><p>The assistant runs on a carefully designed <strong>system prompt</strong> that tells it exactly how to behave like a magical storytelling guide.</p><p>It outlines:</p><ul><li>When to ask for a name</li><li>How to look up memory silently</li><li>What info to collect (age, favorite characters)</li><li>When to generate the story</li><li>And how to encourage story summaries and give kind feedback</li></ul><p>A small piece of that prompt:</p><pre>SYSTEM_PROMPT = &quot;&quot;&quot;<br>You are StoryBuddy, a magical storytelling assistant for children aged 5 to 10.<br>...<br>After the story:<br>- Ask the child to summarize the story in up to 3 sentences.<br>- Then provide encouraging feedback based on what they said.<br>&quot;&quot;&quot;</pre><p>This prompt handles structure and tone, making the assistant feel warm, curious, and consistent.</p><h3>Tech Stack Overview</h3><ul><li><strong>Google Generative AI API</strong> (Gemini Pro and Flash)</li><li><strong>LangGraph</strong> to structure agent behavior</li><li><strong>LangChain</strong> for prompts and memory</li><li><strong>JSON Mode + Retry Logic</strong> to ensure smooth interactions</li><li><strong>Kaggle Notebooks + API Secrets</strong> for safe, serverless prototyping</li></ul><h3>Possible Improvements &amp; Next Steps</h3><p>Even though the core experience works, there’s still a lot of room to grow. Here are some features that are in progress or planned:</p><ol><li><strong>Interactive Story Loop: </strong>After the child shares their story summary, the AI buddy will:</li></ol><ul><li>Cheer them on with encouraging feedback</li><li>Ask if they’d like to hear another story using the same favorite characters</li><li>End the session with a warm goodbye if they’re done for the day</li></ul><p>This helps build routine and keeps the interaction playful and consistent.</p><p><strong>2. Drawing Practice Prompts: </strong>After each story, the agent will suggest fun drawing ideas based on the adventure, helping kids visualize what they just imagined and encouraging creative expression.</p><p>3. <strong>Character Flexibility: </strong>The next version will give kids a choice:</p><ul><li>Continue a new story with the same saved characters</li><li>Or pick a brand new character for a fresh adventure</li></ul><h3>Final Thoughts</h3><p>What started as a capstone project quickly became a personal one. I got to apply new Gen AI tools in a real-life parenting situation with creating something my daughter enjoys.</p><p>Big thanks to <strong>Google and Kaggle</strong> for the structure and inspiration during the Gen AI Intensive. As a result, I have new skills and a real working project that can grow from here.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=659ca910b059" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Git Commit Best Practices: A QA Automation Engineer’s Cheat Sheet]]></title>
            <link>https://medium.com/@viktoriya.kholodenko/git-commit-best-practices-a-qa-automation-engineers-cheat-sheet-fd0430ad7376?source=rss-7af12538b108------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd0430ad7376</guid>
            <dc:creator><![CDATA[Viktoriia Kholodenko]]></dc:creator>
            <pubDate>Wed, 19 Mar 2025 03:16:48 GMT</pubDate>
            <atom:updated>2025-03-19T03:16:48.227Z</atom:updated>
            <content:encoded><![CDATA[<p>Effective version control is very important for any software development process, test automation isn’t an exception. As QA automation engineers navigate complex test suites, framework updates, and environment configurations, having a well-structured Git commit history becomes a great strength. In this guide, I want to share my thoughts about optimizing Git commit practices specifically for test automation workflows.</p><h4>Why Git Commit Practices Matter for Test Automation</h4><p>Test automation code changes rapidly — new test cases are added, flaky tests are fixed, and frameworks are continuously improved. Without proper commit practices, your Git history can quickly become a chaotic mess that makes it difficult to:</p><ul><li>Understand the evolution of your test framework</li><li>Trace when and why a specific test started to fail</li><li>Collaborate effectively with team members</li><li>Revert changes without breaking other functionality</li></ul><h4>The Commit Formula for QA Automation</h4><h4>1. Make Atomic Commits</h4><p>The principle is simple: one commit should represent one logical change. For QA automation it means:</p><ul><li>One test case or test suite</li><li>One bug fix in your automation framework</li><li>One configuration update</li></ul><p>This approach simplifies troubleshooting when tests start failing after a specific commit and makes it easier to revert without affecting other parts of your code.</p><h4>2. Use a Consistent Commit Message Format</h4><p>Following a standardized format helps everyone understand the purpose of each commit:</p><pre>&lt;type&gt;(&lt;scope&gt;): &lt;subject&gt;<br><br>&lt;body&gt;<br><br>&lt;footer&gt;</pre><p><strong>For QA Automation context:</strong></p><ul><li><strong>Type</strong>: test, fix, feat, refactor, config</li><li><strong>Scope</strong>: ui-tests, api-tests, e2e, framework</li><li><strong>Subject</strong>: Clear summary of what changed</li></ul><h4>3. QA-Specific Commit Types</h4><p>While traditional commit types work well, consider the QA-focused alternatives:</p><ul><li>test: Adding or modifying test cases</li><li>fix: Fixing flaky tests or framework issues</li><li>refactor: Improving test structure without changing behavior</li><li>config: Updates to test environments or configurations</li><li>docs: Test documentation updates</li></ul><h4>4. Examples of How to Apply</h4><p>Here’s how effective commit messages might look in practice:</p><pre><strong>test(login-flow):</strong> Add user role permission verification tests<br><br>Added tests that verify different user roles can access appropriate features.<br>Includes admin, editor, and viewer role test cases.<br><br><strong>Resolves</strong>: QA-1234</pre><pre><strong>fix(api-tests):</strong> Fix intermittent timeout in payment API tests<br><br>Increased wait time and added proper assertion handling to prevent<br>false negatives during high server load.<br><br><strong>Fixes</strong>: QA-5678</pre><p>Also, I think this list of best practices with examples will help to see in practice a huge variety of cases where communication through simple commit messages could be short, but informative.</p><ul><li><strong>Link to Test Cases</strong>. Such small change, such as adding a test ID to commit messages, helps to create traceability between your automated tests and your test management system.</li></ul><pre><strong>test(checkout)</strong>: Implement automated tests for discount codes<br><br><strong>Covers test cases:</strong> TC-452, TC-453, TC-454</pre><ul><li><strong>Tag Environment-Specific Changes: </strong>Marked changes that affect specific environments help prevent confusion when tests pass locally, but fall on the specific environment</li></ul><pre><strong>config(staging):</strong> Update API endpoints for new staging environment</pre><ul><li><strong>Highlight Test Data Changes: </strong>test data modifications can cause unexpected test failures, so make them explicit:</li></ul><pre><strong>refactor(test-data):</strong> Update mock user profiles for auth testing</pre><ul><li><strong>Indicate Test Stability Improvements: </strong>Highlight changes aimed at improving test reliability:</li></ul><pre><strong>fix(framework):</strong> Implement retry mechanism for flaky network tests</pre><ul><li><strong>Document Test Coverage Changes:</strong></li></ul><pre><strong>test(payment):</strong> Add credit card validation negative test cases<br><br>Improves coverage of payment validation edge cases by 15%</pre><h4>The Long-Term Benefits</h4><p>Implementing these Git commit practices offers significant advantages:</p><ul><li><strong>Easier Debugging</strong>: Quickly identify which commit introduced a test failure</li><li><strong>Better Onboarding</strong>: New team members can understand the test suite’s evolution</li><li><strong>Improved Collaboration</strong>: Clear communication between QA and development teams</li><li><strong>Efficient Code Reviews</strong>: Reviewers understand the purpose and scope of each change</li><li><strong>Streamlined Maintenance</strong>: Simpler to update and maintain tests over time</li></ul><h4>Conclusion</h4><p>Taking the time to establish and follow proper Git commit practices will benefit your automation project. While it might be some extra work at the beginning, these practices quickly become a part of daily activities and dramatically improve the maintainability of your test code.</p><p>I believe that applying even some of these Git commit practices can significantly improve automation workflows and collaboration between QA and development teams. Also, the test maintenance becomes considerably easier. A well-structured Git history becomes an invaluable resource when investigating test failures or understanding the evolution of a test suite over time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd0430ad7376" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>