<?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 Therese Sario on Medium]]></title>
        <description><![CDATA[Stories by Therese Sario on Medium]]></description>
        <link>https://medium.com/@77mssario?source=rss-8fbbc408746b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*NknGMUVNybq2KEE4</url>
            <title>Stories by Therese Sario on Medium</title>
            <link>https://medium.com/@77mssario?source=rss-8fbbc408746b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 21:57:33 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@77mssario/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[CI/CD in Microsoft Fabric with GitHub]]></title>
            <link>https://medium.com/@77mssario/ci-cd-in-microsoft-fabric-with-github-07d0f6814a33?source=rss-8fbbc408746b------2</link>
            <guid isPermaLink="false">https://medium.com/p/07d0f6814a33</guid>
            <category><![CDATA[pipeline]]></category>
            <category><![CDATA[cicd]]></category>
            <category><![CDATA[microsoft-fabric]]></category>
            <category><![CDATA[github]]></category>
            <dc:creator><![CDATA[Therese Sario]]></dc:creator>
            <pubDate>Mon, 10 Aug 2026 10:53:56 GMT</pubDate>
            <atom:updated>2026-08-10T10:53:56.141Z</atom:updated>
            <content:encoded><![CDATA[<blockquote><em>A step-by-step implementation guide using Fabric Git integration and deployment pipelines</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uUl9U7y2yMRs4Pz6FCKYzQ.png" /></figure><p>Building your data platform is one thing; maintaining your artifacts and having good application or data lifecycle management is one of the key components of development that is mostly overlooked or treated as an afterthought.</p><p>In this article, I’m going to show you a step-by-step guide on how to implement CI/CD in Microsoft Fabric with GitHub.</p><p>This guide applies whether your environments sit on separate Fabric capacities or share the same capacity — as long as the workspaces are in the same tenant.</p><h3>Pre-requisites in Fabric</h3><ol><li>You need a Fabric capacity, and you must be a workspace admin to connect a workspace to a repository.</li></ol><p>2. Enable the following tenant settings in the Admin portal:</p><ul><li>Users can sync workspace items with GitHub repositories</li><li>Users can synchronize workspace items with their Git repositories</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/934/1*NsJP1bqVBFBPLX6UzcM62A.png" /></figure><ul><li>Enable Users can create Fabric items</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*bmqfhbXCipzcl9frRfH91w.png" /></figure><p>These switches can be enabled by the tenant admin, and where delegation is turned on by the capacity admin or workspace admin.</p><p>3. Create workspaces for your environments</p><p>This is where you separate your activities. Some of our clients provision a separate capacity for Dev, UAT, and Prod just to isolate the capacity units and compute power, so that development activity doesn’t affect the production environment.</p><ul><li>Development Workspace</li><li>UAT / Test Workspace</li><li>Production Workspace</li></ul><h3>Pre-requisite in Github</h3><p>Git integration supports cloud-based Azure DevOps, GitHub, and GitHub Enterprise. I’d recommend GitHub, since Microsoft has been constantly improving this product — especially the integration with GitHub Copilot.</p><p>With the Github path, you need the following:</p><ol><li>Create a Github Repository, with a branch you are allowed to commit to.</li></ol><p>2. A GitHub personal access token (classic or fine-grained)</p><ul><li>If your organization enforces SAML SSO, authorize the token for that organization. In my case, I’ll authorize Microsoft</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gSHWI6QpDuHTcJ6AD-xz-Q.png" /></figure><p>3. The Repository URL</p><h3>Configure in Fabric</h3><ol><li>Go to the workspace &gt; Workspace settings &gt; Git integration (visible once the tenant setting is enabled), and select GitHub as your provider.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x2TN19JuSKf6C7C4imLOMw.png" /></figure><p>2. Add your account: enter a unique display name, your personal access token, and (optionally) the repository URL. Click Connect.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/583/1*HcJ-iyp-u1G0wA-LIaxxjQ.png" /></figure><p>3. Once added, Click Connect</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/944/1*g_E4aOdVed0kAVxkaz8znA.png" /></figure><p>4. Choose the branch you want to reference your working branch and enter the Git folder. If you leave the folder blank, content is created in the repository root. Note that you can only connect to one branch and one folder at a time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yjzLfkkQziCYLjK_mAJYFA.png" /></figure><p>Click Connect and sync.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qQILxe5-LC8WWcitz5PzyQ.png" /></figure><h3>Github Repo</h3><p>In the GitHub repository, you can create a pull request to promote your feature/dev branch to your UAT branch.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FYn3efN-WevBlu6nhz685w.png" /></figure><h3>Working in branches</h3><p>There are multiple ways to create a new branch:</p><ul><li>Option 1: Create the branch directly in your GitHub repository, it will then appear in Fabric.</li><li>Option 2: Use New branch inside Fabric’s Git integration panel.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*lx3fwbHvPC7fbVavy2eZFA.png" /></figure><ul><li>Option 3: Work in your main branch, and when it’s time to commit, select Commit to new branch , Fabric creates the branch in your repository for you.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*ekVteNcoDoFIj3ZKpA4RVA.png" /></figure><p>By default, branch operations such as Switch branch and Checkout to new branch require the workspace Admin role. A workspace admin can enable Allow users with at least Contributor role to change Git branch to let Contributors and Members do this too.</p><p>Note: If you’re following the standard CI/CD pattern with dev, UAT, and prod environments, be mindful of which branch you’re committing to.</p><h4>GREAT! Now that the Fabric items are in our repository, we can deploy</h4><h3>Deployment Pipelines</h3><p>To use deployment pipelines you need a Fabric subscription, and you must be an admin of the workspaces involved.</p><ol><li>Go to your dev workspace and select Create deployment pipeline (you can also create one from Workspaces &gt; Deployment pipelines).</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6t2-TRjPiCMDD3R_S5R36Q.png" /></figure><p>2. Name the pipeline and define your stages. By default there are three: Development, Test, and Production.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LISj-AT58u7YIv-pqLbNYQ.png" /></figure><p>3. Assign a workspace to each stage. One workspace is assigned per stage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5vcEvF_jyYavV4W2_rP6EQ.png" /></figure><p>4. Select the Fabric items you want to deploy — you can do a full deployment or a selective one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zIturJ1FKY2n3vKgqrzyyg.png" /></figure><p>5. Click deploy</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CTsqzav8x_X-BlUHSRUp8g.png" /></figure><p><strong>Your Fabric items are now deployed to the UAT environment. Repeat the same exercise for the production deployment.</strong></p><h3>Best Practices in Deployment pipeline</h3><p><a href="https://learn.microsoft.com/en-us/fabric/cicd/best-practices-cicd">Best practices for lifecycle management in Fabric — Microsoft Fabric | Microsoft Learn</a></p><h3>Closing</h3><p>A proper CI/CD setup pays off twice: it makes day-to-day development smoother for the team, and it gives you a clear history of every change.</p><p>You now have that working in Fabric — GitHub for source control, deployment pipelines for promotion, and deployment rules to keep each environment pointing where it should.</p><p>Implemented this differently? I’d love to hear what worked for your team.</p><h3>About the Author</h3><blockquote><em>Hi, I’m Therese, a Solution Engineer for Data Platforms at Microsoft, sharing real-world tips from the projects I build.</em></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=07d0f6814a33" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Now Here: OneLake Storage Tiers]]></title>
            <link>https://medium.com/@77mssario/now-here-onelake-storage-tiers-84449e8308b5?source=rss-8fbbc408746b------2</link>
            <guid isPermaLink="false">https://medium.com/p/84449e8308b5</guid>
            <category><![CDATA[storage]]></category>
            <category><![CDATA[onelake]]></category>
            <category><![CDATA[data-platforms]]></category>
            <category><![CDATA[cost-savings]]></category>
            <category><![CDATA[microsoft-fabric]]></category>
            <dc:creator><![CDATA[Therese Sario]]></dc:creator>
            <pubDate>Mon, 03 Aug 2026 17:27:08 GMT</pubDate>
            <atom:updated>2026-08-03T17:37:19.274Z</atom:updated>
            <content:encoded><![CDATA[<p>If you’ve worked with ADLS (Azure Data Lake Storage) tiers: hot, cool, cold — this is the same idea, now available in OneLake.</p><h4>This article is for you if:</h4><ul><li>You have data that isn’t accessed often (say, once a year by your audit team)</li><li>That data is large enough that storage cost actually matters</li><li>You’re working in Microsoft Fabric</li></ul><h4>Content:</h4><ul><li>What is a OneLake storage tier</li><li>Current tiers</li><li>A sample business use case</li><li>Watch-outs</li><li>How to implement it — on a Delta table, a folder, and a specific file</li></ul><h3>What is a OneLake storage tier?</h3><p>A OneLake storage tier is a cost-optimization feature that matches your storage cost to how often data is actually used. There are three tiers: Hot, Cool, and Cold.</p><p>All three tiers are <em>online</em>, so your data stays directly accessible. There’s no restore step before you can read cool or cold data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KbPvX6daRyPsOAie2x1C9g.png" /></figure><h4>Current tiers</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/978/1*CCcX5667AbBy2jyd_SnvVQ.png" /><figcaption><em>As you move from hot to cool to cold, storage gets cheaper, but retrieval gets more expensive</em></figcaption></figure><h4>Sample Business Use Case</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nBI6jJHQePCRCa6M0DcPEA.png" /></figure><h4>Example: saving on storage costs through tiering</h4><p>Assume you have 10 TB of data that must be retained for five years (60 months). Once a year you run a compliance audit that reads roughly 10% (1 TB) of it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/884/1*u-0_zmPWV7eJQbJuHk0h_w.png" /><figcaption>For data this cold, the cold tier saves you $12,000 over five years versus keeping it hot , just remember the trade-off is higher retrieval cost the few times you <em>do</em> read it.</figcaption></figure><h4>Watch-outs</h4><p>The cool and cold tiers each have a minimum number of days a file must be stored before you move or delete it:</p><ul><li>30 days for the cool tier</li><li>90 days for the cold tier</li></ul><p>If a file is deleted, overwritten, or moved to a different tier before that minimum, it’s subject to an early-deletion penalty.</p><h4>Early-deletion penalty</h4><p>Example: If a blob is moved to the cool tier and deleted after 21 days, you’re charged an early-deletion fee equivalent to 9 days of storage (30-day minimum − 21 days stored). The charge is prorated on the storage price of that tier.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/595/1*-Vowqh-u11D4n4dD0--mhg.png" /><figcaption>You completed 21 of the required 30 days, so you’re billed the early-deletion fee for the 9 days you fell short → $4.14.</figcaption></figure><h3>How to implement</h3><p>The easiest way is through Lifecycle Management in your Fabric workspace:</p><p>Fabric UI → Workspace settings → OneLake → Lifecycle management</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*scVUl992_40u4ww4XqMZOw.png" /></figure><p>You can target a tier at three levels of granularity:</p><ol><li>A Delta table</li></ol><p>Format: &lt;LakehouseName&gt;.Lakehouse/Tables/&lt;tableName&gt;</p><p>Example: agri_lakehouse.Lakehouse/Tables/farmer_dim</p><p>2. A whole folder</p><p>Format: &lt;LakehouseName&gt;.Lakehouse/Files/&lt;folder_name&gt;</p><p>Example: agri_lakehouse.Lakehouse/Files/testing_cold</p><p>3. A specific file (e.g. a CSV)</p><p>Format: &lt;LakehouseName&gt;.Lakehouse/Files/&lt;FolderName&gt;/&lt;file_name&gt;.&lt;format&gt;</p><p>Example: agri_lakehouse.Lakehouse/Files/testing_cold/farmers.csv</p><p>Reference: <a href="https://learn.microsoft.com/en-us/fabric/onelake/onelake-storage-tiers">Storage tiers in OneLake — Microsoft Fabric | Microsoft Learn</a></p><h3>Wrapping up</h3><p>Storage tiers won’t change how you build in Fabric but they can quietly cut your storage bill without touching a single pipeline. Know your access patterns, respect the retention minimums, and let lifecycle rules do the rest.</p><h3>About the Author</h3><p>I’m Therese Sario, a Solutions Engineer for Data Platforms at Microsoft. I share practical tips and best practices with the Microsoft community.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=84449e8308b5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Hours to Seconds: Automating Power BI Metadata in Fabric]]></title>
            <link>https://medium.com/@77mssario/from-hours-to-seconds-automating-power-bi-metadata-in-fabric-35010ea8f7b2?source=rss-8fbbc408746b------2</link>
            <guid isPermaLink="false">https://medium.com/p/35010ea8f7b2</guid>
            <category><![CDATA[data-dictionary]]></category>
            <category><![CDATA[power-bi]]></category>
            <category><![CDATA[microsoft-fabric]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data-agents]]></category>
            <dc:creator><![CDATA[Therese Sario]]></dc:creator>
            <pubDate>Tue, 16 Jun 2026 04:46:28 GMT</pubDate>
            <atom:updated>2026-08-17T03:17:23.624Z</atom:updated>
            <content:encoded><![CDATA[<p>In a 4-minute read, learn how to centralize your Power BI semantic model metadata in one file and auto-populate every description in seconds using a Fabric notebook.</p><h3>The Background</h3><h4>In the era of AI, data agents are changing how we work with Power BI — letting people ask questions in plain language instead of digging through reports.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/913/1*NNT2W0Jsg--HO4NFPLgEHg.png" /><figcaption>But here’s the catch: an agent is only as good as the business context built into your semantic model.</figcaption></figure><h3>The Problem</h3><p>Before your agent can answer questions about your data, you first need to <strong>onboard</strong> it to that context. It has to understand what each table holds and what each column actually means.</p><p>That context lives in your Power BI semantic model’s descriptions. And when your model has 50+ tables and 100+ columns, filling all of those in by hand is <strong>slow, repetitive work.</strong></p><p>When I build models for my clients, this one step alone used to take me hours. <strong>So let me show you a faster way.</strong></p><h3><strong>The Solution</strong></h3><p>Populate every table and column description automatically in minutes — using a PySpark notebook in Microsoft Fabric.</p><p>You maintain your business definitions in one simple CSV, and the notebook pushes them straight into your semantic model</p><h3><strong>Before you start</strong></h3><p>You need these things:</p><ol><li>A data dictionary (CSV) — this is the single source of truth for all your descriptions. It needs three columns:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BzFNDqGBqfam8JSMlKB4Pg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*FnTLpcIGMlWNY9QAok_kAA.png" /><figcaption>Sample CSV file</figcaption></figure><p><em>Tip: Store this CSV in your Lakehouse / Files so the notebook can read it directly.</em></p><p>2. A Fabric workspace containing your raw data.</p><p>3. A semantic model already created and published in that workspace.</p><h3><strong>Implementation</strong></h3><ol><li>Install the library</li></ol><pre>%pip install semantic-link-labs</pre><p>2. Step 2 — Run the script</p><p>Update the three variables at the top (DATASET, WORKSPACE, DESC_FILE) to match your environment, then run:</p><pre>import pandas as pd<br>from sempy_labs.tom import connect_semantic_model<br><br>DATASET   = &quot;cruz-support ticket semantic model&quot;   # as it appears in the workspace<br>WORKSPACE = &quot;smec-dataplatform-e5&quot;<br>DESC_FILE = &quot;Files/landing-zone/2_Data Dictionary.csv&quot;   # lakehouse path, or use abfss://...<br><br># --- load the external file (CSV here; use pd.read_excel for .xlsx) ---<br>meta = pd.read_csv(f&quot;/lakehouse/default/{DESC_FILE}&quot;)<br>meta[&quot;column_name&quot;] = meta[&quot;column_name&quot;].fillna(&quot;&quot;).astype(str).str.strip()<br>meta[&quot;table_name&quot;]  = meta[&quot;table_name&quot;].astype(str).str.strip()<br>meta[&quot;description&quot;] = meta[&quot;description&quot;].fillna(&quot;&quot;).astype(str)<br><br># index for quick lookup<br>tbl_desc = {r.table_name: r.description<br>            for r in meta.itertuples() if r.column_name == &quot;&quot; and r.description}<br>col_desc = {(r.table_name, r.column_name): r.description<br>            for r in meta.itertuples() if r.column_name and r.description}<br><br>applied, missing = 0, []<br><br>with connect_semantic_model(dataset=DATASET, workspace=WORKSPACE, readonly=False) as tom:<br>    for t in tom.model.Tables:<br>        # table-level description<br>        if t.Name in tbl_desc:<br>            t.Description = tbl_desc[t.Name]<br>            applied += 1<br>        # column-level descriptions<br>        for c in t.Columns:<br>            key = (t.Name, c.Name)<br>            if key in col_desc:<br>                c.Description = col_desc[key]<br>                applied += 1<br><br>    # report anything in the file that didn&#39;t match a real object<br>    model_tables = {t.Name for t in tom.model.Tables}<br>    model_cols   = {(t.Name, c.Name) for t in tom.model.Tables for c in t.Columns}<br>    missing  = [k for k in tbl_desc if k not in model_tables]<br>    missing += [k for k in col_desc if k not in model_cols]<br><br>print(f&quot;Applied {applied} descriptions.&quot;)<br>if missing:<br>    print(&quot;No matching object for:&quot;, missing)<br># Changes are committed automatically when the `with` block exits.</pre><h3>Output</h3><h4>What the script does, in plain terms:</h4><ul><li>Reads your CSV and cleans it up.</li><li>Builds two lookups — one for table descriptions, one for column descriptions.</li><li>Connects to your semantic model and writes each description onto the matching object.</li><li>Flags any dictionary entry that didn’t match a real table or column, so you can catch typos.</li></ul><h4>The Result<br>The notebook populates your entire model in one pass. In my case, it applied 118 descriptions in about 30 seconds</h4><p><strong>Replacing what used to be hours of manual work.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/619/1*vtHz2vjHTeA7DYG-avHY0g.png" /></figure><p>Sample Output from my semantic model: Table Name</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ppsPeeO3owRP02d-8MSepg.png" /></figure><p>Sample Output from my semantic model: Column Name</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eZqtjbuLd7Lo-9CzqQQfbw.png" /></figure><p>And that’s it. What used to take me hours now runs in the time it takes to grab a coffee. Your descriptions stay version-controlled in one CSV, your semantic model stays documented, and your AI agents get the context they need to answer real business questions.</p><p>If you found this helpful, give it a clap 👏 and follow along — I share practical Fabric and Power BI workflows like this regularly. Got a question or a use case of your own? Drop it in the comments — I’d love to hear how you’re using it.</p><p><em>About Me!</em></p><p><em>Hi, I’m Therese — a Solution Engineer for Data Platforms at Microsoft, sharing real-world tips from the projects I build.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=35010ea8f7b2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>