Over the last years I have done a lot of automation with LLM’s. In this blog post I’ll show you how I use it to make release management easier.

Release management is a critical part of the software development lifecycle, where precision, speed, and clarity are essential. At MeldApp—an ticket management system built with SvelteKit for managing maintenance tickets, anonymous reporting, and building/location management—I’ve built an automated deployment assistant powered by Claude AI to simplify my release process. MeldApp is already used in production by several customers and companies, and I, working in a small team, needed a reliable way to manage releases without the usual manual overhead.

By leveraging AI subagents, I can analyze repository changes, manage versioning, prepare Progressive Web App (PWA) for updates (yes, Meldapp can also be installed on your phone or desktop), and validate releases—ensuring a smooth deployment every time.

In this post, I’ll walk through how my Deployment Assistant automates release management to save time and reduce errors, including a step-by-step breakdown of its workflow. You can find the complete prompt I use for this automation HERE.

Why Automate Release Management?

I found that managing releases manually was error-prone and time-intensive. From analyzing repository changes to deciding on a version bump and ensuring PWA updates are handled correctly, there’s plenty of room for oversight. But honestly, one of the main reasons I built this was because I always forget to update version numbers :-).

Now I can simply run /deploy in my terminal and let the automation handle it. This has helped me by:

Providing structure: Automating repetitive tasks ensures consistency across my releases.

Reducing time-to-deploy: AI subagents analyze and execute tasks in seconds.

Minimizing errors: Automated processes streamline complex tasks like CHANGELOG generation and service worker updates.

Never forgetting version bumps: The assistant always handles versioning, so I never have to remember this step again.

With my Deployment Assistant, I can focus on product innovation while the assistant handles the release cycle.

Step 1: Pre-Deployment Analysis Using Subagents

Before a release, I need to understand what has changed since the last version. My assistant uses Claude AI’s general-purpose subagents to analyze repository changes.

Change Review: The assistant starts by reading my CHANGELOG.md file to identify the latest version. It uses git to list commits and categorizes changes into:

  • New Features
  • Bug Fixes
  • Refactoring
  • Dependencies
  • Documentation
  • Infrastructure

Code Inspection: If commit messages are unclear, the subagent examines modified code files for more details.

Categorization: The subagent prepares a structured summary of changes, including commit hashes and specific file modifications.

This upfront analysis ensures I have full clarity on what’s included in the release.

Step 2: Version Management and Changelog Automation

After identifying repository changes, the assistant handles version updates and changelog generation.

Version Recommendations: Based on the nature of changes, the assistant recommends whether I should update the version following Semantic Versioning as:

  • Patch: For bug fixes or security patches
  • Minor: For new, non-breaking features
  • Major: For breaking changes or major rewrites

Changelog Updates: Following the Keep a Changelog standard, the assistant creates or updates my CHANGELOG.md file. Entries are categorized as Added, Changed, Fixed, Security, Deprecated, or Removed.

Confirmation Loop: The assistant allows me to confirm or adjust the version update type before finalizing the version number.

This ensures my versioning and documentation are accurate and consistent with best practices.

Step 3: Updating the PWA Service Worker

For Progressive Web Apps, ensuring users receive the latest updates is essential. As part of my release process, the Claude assistant automates updates to the PWA service worker:

FORCE_UPDATE Timestamp: The assistant reads the current FORCE_UPDATE timestamp from my src/service-worker.js and updates it to the current date and time.

PWA Manifest Validation: It verifies that my site.webmanifest file for the PWA is valid.

Cache Invalidation Confirmation: By updating the service worker timestamp, the assistant ensures client-side caches are refreshed after deployment.

Automating this step ensures seamless updates for my end-users without manual intervention.

Step 4: Final Validation and Deployment Preparation

The final pre-deployment step is to ensure my dev branch is ready for a manual merge into main.

Branch Validation: The assistant checks that all my changes are committed, and the current branch is dev.

Build Verification: Before preparing the branch, the assistant ensures the build is successful. Testing is handled elsewhere in my development process, so this step focuses purely on deployment readiness.

Deployment Summary: A comprehensive summary is generated, including:

  • The new version number
  • Key changes in the release
  • Updated service worker timestamp
  • Confirmation of build success

With this, my dev branch is fully prepared for a final manual merge (which for now I prefer, let’s say my trust in LLM’s and prompts has a very healthy dose of scepticism. This is the final step before deployment to production is triggered. I want to be sure I am the one who has the final say. The LLM does give me a example of what commands to use, including tagging the version number to the production branch, so that is as easy as it gets.)

The Complete Deployment Workflow

The workflow in the end consists of these steps, some of them are done in parallel to save time.

  1. Analyze: Subagents analyze repository changes since my last release
  2. Version: They handle version updates and generate a new changelog based on changes
  3. Update PWA: Service worker updates are automated to ensure cache invalidation
  4. Validate: My dev branch is verified for deployment readiness
  5. Prepare: Instructions for merging dev into main are provided

Conclusion

By leveraging Claude AI and its subagent functionality, the Deployment Assistant makes release management faster, safer, and more efficient. Whether it’s analyzing repository changes, managing versions, or ensuring PWAs are ready for deployment, the assistant handles it all with precision. The combination of AI-powered analysis and structured workflows creates a deployment process that’s both reliable and scalable.

Want to learn more about MeldApp? Visit https://meldapp.nl (landing page with more information coming soon).

Want to implement similar automation in your workflow? Use the exact prompt I created as a starting point - find it HERE. Start by identifying your most repetitive release tasks, then consider how AI agents can systematically handle each step. The time investment in setting up automation pays dividends in reduced errors and faster deployments.