Techalicious ~ Casual Tech Hobbyist Social Group

🧙‍♂️ Vibe Coding from Your Terminal with Claude Code + Local Models

Ever wanted to just describe what you need and have working code appear? That's vibe coding, and now you can do it locally.

Claude Code is Anthropic's agentic coding CLI. By default it calls their cloud API, but Ollama now supports Anthropic API compatibility, so you can point Claude Code at your local models instead. All the polish of Anthropic's CLI, none of the cloud dependency.

🧙‍♂️ What We'll Cover

• Installing Claude Code (just needs Node.js)
• Running qwen3-coder locally through Ollama
• Pointing Claude Code at localhost instead of the cloud
• Basic workflow for generating, refactoring, and explaining code
• When vibe coding saves time vs when to just write it yourself

🧙‍♂️ Who's This For?

Anyone comfortable with the terminal who wants AI-assisted coding without sending their code to someone else's servers. No cloud subscriptions required. Local AI doing useful things.

☕ What This Is (And Isn't)

This is a casual hobbyist demo, not a professional workshop. Think of it like showing a friend something cool you found over coffee. I'll walk through how I set this up on my own machine, share what worked, and we'll chat about it.

No curriculum. No certificates. No polished slides. Just a tech enthusiast sharing something interesting with other tech enthusiasts.

🌐 Tech Stuff Worth Knowing

I'll be demonstrating on a Mac M4. This works on Linux too. I'm not a Windows user so I can't speak to that setup.

If you want to follow along, you'll need:

• macOS or Linux (Windows untested)
• Node.js installed
• Ollama running locally
• Enough RAM to run a coding model (16GB minimum, 32GB+ recommended)

We won't be troubleshooting individual setups during the session. This is a walkthrough, not hands-on tech support.

Tired of paying $20/month for ChatGPT Plus? What if you could run the same thing on your own computer, for free, forever?

Tonight we're setting up a full ChatGPT-style chat interface that runs entirely on your hardware. No cloud. No subscriptions. No one reading your conversations. Just you and your own private AI.

⚠️ System Requirements

• RAM: 16GB minimum (32GB+ recommended)
• Storage: 15GB free disk space
• OS: macOS 10.15+ or Linux (Ubuntu recommended)

🤖 What We'll Cover

• Installing Ollama (the engine that runs AI models locally)
• Downloading and managing AI models
• Setting up OpenWebUI (a slick ChatGPT-style interface)
• Basic configuration and settings
• Testing your first conversation

By the end, you'll have a fully functional chat interface running in your browser, powered by AI on your own machine.

No coding experience required!

Pattern matching is one of those skills that separates "I can code" from "I can wrangle anything." This session goes beyond the basics into the powerful features most people don't know exist.

🧠 What We're Covering:

• The (?x) extended modifier that lets you write regex with comments and whitespace. Finally, patterns that humans can actually read.
• Regex subroutines using (?&name) so you can define a pattern once and reuse it. DRY principle applied to regex.
• Lookaheads and lookbehinds for zero-width assertions. Check for something without consuming it.
• Non-greedy quantifiers when you want the shortest match, not the longest.
• Named capture groups with (?...) for readable, maintainable extractions.
• Conditional patterns with (?(condition)yes|no) for context-aware matching.

🧠 Why This Matters for AI Work:

If you're running local LLMs through Ollama or OpenWebUI, you're dealing with text. Lots of it. Model outputs wrapped in markdown code fences. Log files full of timestamps and metrics. User requests that need parsing into structured prompts.

We'll work through real examples:

• Extracting JSON from chatty LLM responses (they love to add "Sure! Here's the data...")
• Parsing Ollama logs for debugging and monitoring
• Cleaning AI-generated text before downstream processing
• Building dynamic prompts from captured user input

🧠 Who Should Join:

Anyone comfortable with basic regex who wants to level up. If you know .* and [a-z]+ but (?x) and (?&name) look like hieroglyphics, this session is for you.