Keeping Up with Azure Updates – The Challenge

As an Azure architect/developer, staying up-to-date with the latest Azure changes is crucial—but let’s be honest, it’s also challenging. While Microsoft and the community do a great job sharing major updates, I often found myself missing smaller, impactful changes—updates that weren’t highlighted in announcements but could still influence project decisions.

Microsoft typically announces major updates on azure.microsoft.com, yet many smaller updates with large impacts go unnoticed, including other useful tutorials. I wanted to eliminate those “Aha! Has this changed?” moments that sometimes occur months later.

To address the inefficiency of manually refreshing every page to track updates, I developed AzureWatcher.com. Initially a personal console application, AzureWatcher.com has evolved into a public tool for the Azure community. This blog post looks into the inner workings, explaining how it leverages AI to monitor and report changes in Azure documentation.

Why Tracking Azure Documentation is Hard

Microsoft primarily announces major updates on the Azure Updates page. However, smaller yet impactful changes often go unnoticed. This gap can lead to unexpected surprises, such as deprecated services or critical feature modifications, catching developers and architects off guard months after the changes occur.

Azure’s documentation is hosted on GitHub and updated daily. Anyone can contribute via pull requests, which is great for keeping the platform dynamic. If you spot an error, I highly recommend fixing it—contributing to improving the ecosystem for everyone. Here’s a useful guide on contributing.

However, I quickly realised that manually refreshing documentation pages or GitHub repositories was not scalable. That’s when I built AzureWatcher.com—initially as a personal console app, but now a free, AI-powered tool available for the community.

How AzureWatcher.com Works

Step 1: Fetching Documentation Updates

The first step is straightforward. Using GitHub’s REST API, AzureWatcher.com performs hourly checks to identify updates in Microsoft’s Azure documentation repository. This automated process ensures that no changes go unnoticed.

Step 2: Data Cleansing

This step is the most challenging. Most documentation updates involve minor edits such as typos, markup adjustments, name updates, URL clarifications, and other non-substantive corrections. For example, renaming “Active Directory” to “Entra ID” or correcting “Oauth2.0” to “OAuth2.0” are changes that do not impact the core content.

To manage this, I developed an engine that cleans and categorises the changes, filtering out non-essential updates. The cleansing process involves:

• URL Corrections: Ignoring updates related to page links and URLs, index tables, or navigations.
• Metadata Changes: Ignoring updates related to page metadata, such as last update timestamps or reviewer names.
• Non-Content Documents: Excluding changes to images, menu items, and other non-content elements.
• Markdown Adjustments: Filtering out insignificant markdown modifications, such as meta data updates, link updates, video etc..
• Duplicate Titles and Commit Details: Removing redundant titles and irrelevant commit information.
• Typography Detection: Normalizing sentences by removing cases, punctuation, spacing, and diacritic differences, then using Damerau–Levenshtein distance to match their words and decide if changes are minor.
• Ignore Less Popular Docs: To limit consumption, we ignore less frequently used products such as “Private 5G” or “Healthcare API”.

This engine addresses duplications and enhances data quality for later processing. As we all know: garbage in garbage out.

Step 3: Data Enrichment & Aggregation

Once the data is cleansed, AzureWatcher.com aggregates modifications for the same document and enriches the documents with additional metadata. This includes product names, document descriptions, and other relevant details that enhance the information available for analysis and reporting.

Step 4: AI-Powered Summarisation

Using Azure OpenAI’s GPT-4, AzureWatcher.com processes the enriched data to further refine the information. This AI-driven step serves two primary purposes:

• Generating Clear Summaries: It produces detailed summaries of the changes, making it easier for users to understand the updates at a glance.
• Filtering Unimportant Changes: The AI removes other insignificant modifications like minor typos or markup fixes.
• Data Regrouping: Similar changes or modifications around the same impact are categorized together.

Step 5: Storing and Reporting Data

The final step involves storing the refined data in a SQL database and presenting it in an accessible format. Currently, AzureWatcher.com delivers this information via a weekly newsletter sent out every Sunday, allowing users to start their week informed about the latest Azure documentation updates.

Technical Insights

AzureWatcher.com is built with a robust technical stack to ensure reliability and scalability:

• CloudFlare CDN: Enhances website performance and security by distributing content globally.
• Blazor .NET 9: Powers the front-end, providing a responsive and interactive user interface.
• Entity Framework: Manages data access within the application.
• SQL Database: Stores all processed data securely and efficiently.
• Azure Functions .NET 9: Handles scheduled tasks and automation with timer triggers.
• Azure OpenAI (GPT-4): Drives the AI components for data summarisation and filtering.

Interesting Statistics

• In the second week of January 2025, Microsoft updated over 10.000 files in its Azure documentation repository.
• 97% of changes were automatically discarded as non-important based on rules in the cleansing engine.
• From those data is aggregated and 30% of changes were filtered out by AI as minor edits after deeper analysis.
• Only 2% of updates were significant enough to summarise and report to users.

Example Output

Future Developments

AzureWatcher.com is currently in its beta phase, and I’m looking forward to receiving feedback: I encourage users to share their experiences, suggest improvements, and report any bugs. Future enhancements include:

• Improving Report Quality: Identifying additional trivial changes that currently slip through the filters.
• Improved duplicate detection: For the same changes across multiple files.
• Extending the Products List: Tracking changes from SQL Server and Azure OpenAI, which are currently hosted in a separate repository.
• Dashboard: A visual interface displaying changes per product or page over specific time periods.
• Advanced Search Functionality: Enabling users to explore historical changes and navigate the archives in a more detailed matter with AI search capabilities.

The goal is to provide Azure practitioners with a reliable tool to stay informed and make informed decisions based on the latest documentation updates.

Final Thoughts

AzureWatcher.com started as a personal project to address a real pain point—tracking Azure documentation changes. Now, it’s a community tool, and I hope it helps others stay informed and prepared.

I’d love to hear your thoughts and suggestions—what works, what doesn’t, and what’s missing? Visit AzureWatcher.com, sign up, and let me know what you think.

Thank you for being part of this journey!