RoboRuby
Posts
Ruby AI News - May 6th, 2025

Ruby AI News - May 6th, 2025

Vibe coding... so hot right now

Matt Solt
May 06, 2025

Welcome to the 5th edition of the Ruby AI Newsletter! This edition features discussions on Vibe Coding, a quick start tutorial on RubyLLM, a look at the expanded Ruby AI ecosystem, and much more.

Top Stories
Events
- Upcoming
- Previous
Content
Open Source Updates
Jobs & Opportunities

Events

Upcoming

ArtificialRuby is hosting a meetup in NYC on May 7th. There are still a few spaces left, so be sure to RSVP ASAP if you want to attend.

ArtificialRuby.ai May NYC Meetup · Luma

lu.ma/crm6e9sy

The London Ruby User Group (LRUG) is meeting to discuss Practical AI in Ruby: What LLMs Can (and Can’t) Do For Your Projects Today by Lorenzo Barasti, an exploration of integrating LLM capabilities into Ruby applications using RubyLLM and similar libraries, highlighting real-world use cases without the Silicon Valley hyperbole.

May 2025 Meeting | London Ruby User Group

The May 2025 meeting of LRUG will be on Monday the 12th of May, from 6:00pm to 8:00pm (meeting starts at 6:30pm).

lrug.org/meetings/2025/may

VanRuby is presenting AI Engineering for Ruby Developers by Tom Gladhill on May 13th in Vancouver, offering a tour through emerging tools, libraries, and practices that make Ruby a surprisingly solid platform for working with LLMs.

VanRuby: AI Engineering for Ruby Developers · Luma

AI Engineering for Ruby Developers

lu.ma/90d0vykh

Cincinnati Ruby Brigade posted their April recap with Bill Barnett as he dives into the chat interface of a demo app, discussing workflows, RAGs, MCP servers, and LLM integrations. The source code is available here.

Wroclove.rb released Chris Hasiński’s talk Next Token! that demystifies LLMs and shows how to integrate them into your applications by treating them as token generators.

Content

In this reddit discussion on AI tools that can build Rails apps, Kody Kendall showcased Llama Press, an AI chatbot that can modify the underlying Ruby on Rails application. He also posted a demo video and asked for feedback and suggestions going forward. Let’s encourage continued development!

Llama Press - Build Anything with AI

llamapress.ai/users/sign_up

Clayton Lengel-Zigich launched Cartoonie, an AI-powered Rails app that turns your photos into Cartoons without a ChatGPT subscription.

Photo to Cartoon AI App | Cartoonie

Cartoonize your photos into cartoon art with our AI cartoonizer. Choose from styles like Classic Anime, Animation Studio, Mouse House Animation, Block Movie, and more!

cartoonie.ai

Justin Paulson compared Streaming LLM Responses with Rails: SSE vs. Turbo Streams. The article showcases two techniques to handle LLM responses with specific Rails code examples. The first looks at Server Side Events with ActionController::Live and Stimulus, while the second method uses Hotwire and Turbo Streams. The code samples are also available on Github.

Streaming LLM Responses with Rails: SSE vs. Turbo Streams

In the world of Rails development, integrating large language models (LLMs) like OpenAI's GPT has become increasingly common. One challenge developers face…

www.aha.io/engineering/articles/streaming-llm-responses-rails-sse-turbo-streams

Siva Gollapalli connects agents of different programming languages with Bridging the Gap: Connecting Python AI Agents to Ruby Apps with MCP. In this tutorial, the fast-mcp gem is used to integrate agents written in other languages with your Ruby backend via Model Context Protocol.

Bridging the Gap: Connecting Python AI Agents to Ruby Apps with MCP

The Rise of MCP and Connecting AI to Your Apps:

learnitnow.medium.com/bridging-the-gap-connecting-python-ai-agents-to-ruby-apps-with-mcp-614977012399

Joshua Harding showcased his new gem ollama-struct with Beyond the Text Box: Structured LLM Outputs in Ruby with ollama-struct. The articles demonstrates how to use ollama-struct to define schemas for structured data and submit those definitions alongside prompts to Ollama LLMs, with examples for recipe extraction, product reviews, and game character generation.

Beyond the Text Box: Structured LLM Outputs in Ruby with ollama-struct

In the rapidly evolving landscape of AI development, Large Language Models (LLMs) have become an indispensable tool for developers. These…

joshua-harding.medium.com/beyond-the-text-box-structured-llm-outputs-in-ruby-with-ollama-struct-d94c75494a39

JetBrains opened an early access program for Junie, its Ruby AI coding Agent. If you would like to apply for the program, the signup form is available here.

Junie, the coding agent by JetBrains

Delegate your tasks, focus on the results

www.jetbrains.com/junie

JetBrains also open-sourced Mellum, an LLM built for developers.

Mellum Goes Open Source: A Purpose-Built LLM for Developers, Now on Hugging Face | The JetBrains Blog

Mellum doesn't pretend to know everything, but it does one thing exceptionally well. We call it a focal model – small, efficient, and built with a specialized focus.

blog.jetbrains.com/ai/2025/04/mellum-goes-open-source-a-purpose-built-llm-for-developers-now-on-hugging-face

Radamés Roriz did a deep dive into Scaling GenAI by Engineer Vision. The discussion looks at techniques to handle the complexity of generative AI systems using a layered approach. This philosophy is implemented in ActiveGenie, a Rubygem to streamline LLM integration.

Scaling GenAI by Engineer vision

Scaling Generative AI (GenAI) features presents unique challenges.

roriz.dev/post/scaling_genAI_by_engineer_vision

Grant Petersen-Speelman presented Single Ruby DSL for AI LLMs, Reporting, and No-Code Solutions, asking: what if a single Ruby DSL could handle reporting, no-code automation, and AI integrations with minimal duplication? In the post, he discusses the early steps to achieve this goal and presents an upcoming OdataDuty gem.

Single Ruby DSL for AI LLMs, Reporting, and No-Code Solutions

In my last post, Rails to Power BI with OData: The Journey Begins, I explored how to connect a Ruby on Rails app to reporting tools like…

medium.com/nexl-engineering/single-ruby-dsl-for-ai-llms-reporting-and-no-code-solutions-dc0b0cfde285

Indigo Tech Tutorials streamed How To Generate Embeddings for Rag Search in a Ruby on Rails App with Ollama and Sqlite.

Scott Werner wrote about The Coming Knowledge Work Supply, arguing that we must reimagine knowledge work as a high-velocity decision-making operation rather than a creative production process, and sees knowledge workers remaining in demand.

The Coming Knowledge-Work Supply-Chain Crisis

"Humans. Here they are. Bottleneck, bottleneck. Hi, good to see you. And some of you are terrified. You're going to be even bigger bottlenecks." - Tyler Cowen

worksonmymachine.substack.com/p/the-coming-knowledge-work-supply

Christian Ekrem published Coding as Craft: Going back to the Old Gym, advocating for thoughtful, intentional collaboration with AI that preserves the essence of coding as a craft.

Coding as Craft: Going Back to the Old Gym

Why 'reflexive AI usage' sounds like a diagnosis I don't want next to my name

cekrem.github.io/posts/coding-as-craft-going-back-to-the-old-gym

Irina Nazarova interviewed prominent indie hacker Marc Köhlbrugge, founder of BetaList and StartupJobs, to discuss his playbook for building successful apps: Rails, hacking, and Stripe as a scoreboard. The article features a section on AI with Rails and mentions Obie Fernandez’s Ruby book Patterns of Application Development Using AI.

Rails, hacking, and Stripe as a scoreboard: Marc Köhlbrugge’s playbook—Martian Chronicles, Evil Martians’ team blog

Marc Köhlbrugge on hacking, building startups, shipping early, why Rails powers his experiments—and his wish for the Ruby ecosystem.

evilmartians.com/chronicles/rails-hacking-and-stripe-as-a-scoreboard-marc-kohlbrugge-playbook

DeepWiki recently launched a service that uses AI to analyze Github repositories and generate both human and machine readable documentation. While I do not recommend using it for well-documented repositories like Rails, it can be useful to generate insights into repositories without clear documentation. AnyCable (which has great docs!) highlighted their DeepWiki page:

anycable/anycable | DeepWiki

AnyCable is a high-performance realtime server for two-way reliable communication over WebSockets and Server-Sent Events (SSE). It's designed to be language-agnostic with a focus on performance, relia

deepwiki.com/anycable/anycable

Brad Gessler is looking for feedback from Rails developers that want to integrate Model Context Protocol (MCP) capabilities into their SaaS applications.

Curious how many Rails devs on here are integrating their Rails SaaS with AI tools like Cursor, Claude Code, etc. via model context protocol (aka: MCP)?
Tell me how in the thread, what pain points you’re running into, and what could be better.
— Brad Gessler (@bradgessler)
6:29 PM • Apr 27, 2025

Jose Blanco of Thoughtbot recapped Developing a voice AI app in Rails for drive-through ordering, reviewing the technical details from the development livestream. Thoughtbot also continues to regularly release AI content. Since the last edition they published AI for Business: AI and Cognitive Insight, The Meaning of Work in an AI World, and The Rise of the Intelligent Marketplace: Lessons Learned from Thoughtbot Clients.

Developing a voice AI app in Rails for drive-through ordering

In this post we will explore the code for an AI assistant in our application.

thoughtbot.com/blog/developing-a-voice-ai-app-in-rails-for-drive-through-ordering

Julián Duque of Heroku detailed How I Improved My Productivity with Cursor and the Heroku MCP Server. While not Ruby-specific, the post does provide clear insights into working with Cursor and the Heroku MCP Server.

How I Improved My Productivity with Cursor and the Heroku MCP Server | Heroku

Integrating Cursor and Heroku MCP Server for AI-powered development. Learn how this setup boosts productivity and simplifies app deployment.

www.heroku.com/blog/improved-my-productivity-cursor-and-heroku-mcp-server

Open Source Updates

Code Spotlight

Gumroad has been on a tear lately, open sourcing all of their repositories including their primary Ruby on Rails code base and their typescript-based AI-powered customer support application Helper.

However I wanted to highlight a brief portion of novel AI-powered code from Flexile, their software for managing equity and cap tables. In this example, they use a well-defined prompt, schema, and data processing workflow to extract structured data from an Excel-based cap table:

flexile/apps/rails/app/services/process_cap_table_upload.rb at main · antiwork/flexile

Equity for everyone. Contribute to antiwork/flexile development by creating an account on GitHub.

github.com/antiwork/flexile/blob/main/apps/rails/app/services/process_cap_table_upload.rb

New Versions

torch-rb 0.20.0 - Deep learning for Ruby powered by LibTorch.

raix 0.8.3 - Add large-language model AI components to your Ruby applications.

llm.rb 0.5.0 - Lightweight library that provides a common interface and set of functionality for multple Large Language Models. Brief intro posted to Rubyflow.

leoandruby 0.5.3 - Integrates with the Leonardo.ai API, enabling AI-powered image generation in Ruby and Rails applications.

leva 0.1.9.1 - Rails framework for evaluating Language Models using ActiveRecord datasets. Create experiments, manage datasets, and implement evaluation logic.

lluminary 0.2.1 - Framework for building applications with LLMs that provides a structured way to define tasks, manage prompts, and handle LLM interactions.

llm_fixer 0.0.4 - CLI tool to use LLMs to automatically fix errors detected by static analysis tools such as RuboCop.

active_genie 0.0.24 - A set of Ruby modules to work with Generative AI.

OxAiWorkers 0.9.1 - A powerful state machine with OpenAI generative intelligence integration.

fast-mcp 1.3.1 - Implementation of the MCP with multiple approaches for defining tools.

ruby-mcp-client 0.5.1 - Client library for integrating with Model Context Protocol (MCP) servers to access and invoke tools from AI assistants.

mcp_on_ruby 0.3.0 - Implement Model Context Protocol servers to standardize interactions with AI language models.

chalk_ruby 0.3.1 - Ruby client for Chalk.ai.

openai-chat 0.0.6 - Use OpenAI's Chat Completions endpoint and support structured output.

foobara 0.0.117 - Command-centric and discoverable software framework with a focus on domain concepts and abstracting away integration code.

smart_agent 0.1.8 - Build AI agents with a declarative DSL and Model Context Protocol support.

ruby_conversations 1.1.3 - Manage AI conversations via a remote API.

New Gems

geminize - Interface for the Google Gemini API with support for text generation, chat conversations, embeddings, and multimodal content.

a2a - An open protocol enabling communication and interoperability between opaque agentic applications.

kagi-api - Kagi API client for privacy focused information.

omamori - Scan Ruby code and diffs using Google Gemini to detect security vulnerabilities.

protocol-mcp - Multi-schema client and server Model Context Protocol (MCP) implementation, implementing the full MCP spec through transpilation of the MCP TypeScript definitions.

plasma-mcp - Provides a Rails-inspired, convention-over-configuration approach to building Model Context Protocol servers.

vector_mcp - Minimal server implementation for the Model Context Protocol (MCP) in Ruby.

sentiment_insights - Analyze sentiment from survey responses, feedback, and other text sources.

github-daily-digest - Fetches recent GitHub commits and PR reviews and uses Google's Gemini API to analyze and summarize the activity.

gitlore - CLI tool to create summaries of git commits using LLMs.

rasti-ai - AI for apps.

ai-chat - Use OpenAI's Responses API. Supports structured output and image processing.

obsidian_fetch - MCP server focused on fetching and presenting information from Obsidian vaults.