Local AI Agent in your Terminal

Today’s top AI Highlights:

1. AI agent in your terminal that can run code, edit files, browse the web, and more

2. Opensource framework that gives AI agents their own Linux environment

3. AI model generates entire gameplay in real-time without a game engine

4. Anthropic hikes Claude 3.5 Haiku’s price by 4x

5. Opensource all-in-one AI gateway and API developer portal

& so much more!

Read time: 3 mins

AI Tutorials

We’re always looking for ways to automate complex workflows. Building tools that can search, synthesize, and summarize information is a key part of this, especially when dealing with ever-changing data like news.

For this tutorial, we’ll create a multi-agent AI news assistant using OpenAI’s Swarm framework along with Llama 3.2. You’ll be able to run everything locally, using multiple agents to break down the task into manageable, specialized roles—all without cost.

We will use:

• Swarm to manage the interactions between agents,

• DuckDuckGo for real-time news search, and

• Llama 3.2 for processing and summarizing news.

Each agent will handle a specific part of the workflow, resulting in a modular and flexible app that’s easy to adapt or expand.

We share hands-on tutorials like this 2-3 times a week, designed to help you stay ahead in the world of AI. If you're serious about levelling up your AI skills and staying ahead of the curve, subscribe now and be the first to access our latest tutorials.

Don’t forget to share this newsletter on your social channels and tag Unwind AI (X, LinkedIn, Threads, Facebook) to support us!

Latest Developments

AI Agent in your Terminal Equipped with Local Tools 🧰

Here’s an opensource command-line AI assistant that brings the power of LLMs directly to your terminal. gptme integrates seamlessly with local tools, allowing it to execute code, manipulate files, browse the web, and even utilize computer vision. Unlike cloud-based alternatives, gptme operates locally, giving you full control over your data and avoiding the limitations of internet access and timeouts.

It supports multiple LLM providers including OpenAI, Anthropic, and local models through Ollama. You have full control - every action requires confirmation and every change is traceable.

Key Highlights:

1. Direct Terminal Integration - Work directly within your terminal with gptme's AI assistance without disrupting your workflow. Execute shell commands, run Python scripts, and manage files with AI-powered suggestions and automation.

2. Powerful Toolset - gptme offers tools to edit files using patches, browse the web and take screenshots within the chat, and even control your computer (experimental). This allows for complex tasks like web scraping, automated testing, and more.

3. Extensible Architecture - Easily add new tools and functionalities to gptme, thanks to its modular design. Extend its capabilities to match your specific project needs and create personalized AI-powered workflows.

4. Practical Safety - Includes confirmation prompts for file operations, isolation of code execution, and the ability to undo actions - giving you confidence when using AI for system-level tasks while maintaining full control over what gets executed.

5. Quick Start - Install with pipx install gptme and start with simple commands like gptme 'write an impressive particle effect using three.js' or gptme 'suggest improvements to my vimrc'. Check the full documentation here.

Run LLMs in Isolated Linux Environments 🤖

Monadic Chat brings a full Linux environment to LLMs so you can create chatbots that can go beyond just text-based conversations. They can directly execute system commands, run analyses, and handle complex computations. This opensource framework packages GPT-4, Claude, and other LLMs within Docker containers, providing them with access to external tools for tasks like web scraping, data processing, and visualization.

Beyond code execution, it supports voice interaction, image recognition, and AI-to-AI communication, making it particularly useful for building advanced AI apps that need direct system access.

Key Highlights:

1. Ready-to-use Docker Environment - Ships with pre-configured containers for Ruby, Python, Selenium, and PGVector - you can add custom containers and tools through a standardized process. Each container is accessible via SSH, and new tools can be installed using standard package managers.

2. API Integration - Built-in support for OpenAI, Anthropic, Google, Cohere, and Mistral APIs, with a unified interface for handling responses. Includes local model support through Ollama integration, allowing you to run models like Llama and Gemma directly within the containerized environment.

3. File System Architecture - Implements a shared folder system between host and containers, enabling seamless file operations. Supports automatic handling of various file formats including PDFs, CSVs, and Office documents, with built-in conversion utilities and vector storage using PGVector.

4. Development Tools Integration - Full JupyterLab environment with access to the shared file system, WebSocket support for real-time updates, and built-in debugging tools. Includes helper modules for common tasks like text embedding generation, image processing, and voice synthesis.

Quick Bites

AI startups Etched and Decart have developed Oasis, the first real-time, open-world AI model that generates an interactive video game experience from scratch frame-by-frame using a transformer model—no game engine needed. This model uses user input to create live gameplay. The release includes code, a playable demo, and a 500M parameter model that can run locally.

AMD has released its first series of opensource language models AMD OLMo, featuring 1B models trained on 1.3 trillion tokens using AMD Instinct™ MI250 GPUs. The release includes three model variants - a base pre-trained model, a supervised fine-tuned version, and an aligned version optimized for human preferences - along with complete training details and model checkpoints.

Apple is reportedly developing Vision Pro successor with an M5 chip for a potential 2025 launch, while also exploring iPhone-connected smart glasses as a more accessible entry into its visionOS ecosystem. The Vision Pro SE, or whatever it ends up being called, is now reportedly delayed and won’t be arriving until beyond 2027. Apple doesn’t seem to be giving up on mixed reality just yet.

Anthropic’s latest Claude 3.5 Haiku introduced two weeks back is now available via API. This time however the price has been hiked from $0.25 to $1 per million input tokens and from $1.25 to $5 per million output tokens “to reflect its increase in intelligence.” Claude 3.5 Haiku will currently not support image input capabilities. If you want maximum cost-efficiency and image processing, you can still use 3 Haiku.

Tools of the Trade

1. APIPark: Opensource AI gateway and API developer portal for developers and enterprises to manage, integrate, and deploy AI services easily. It connects to over 100 AI models, standardizes API calls, and provides tools for monitoring and securing API usage.

2. Alex: Cursor-like sidebar for Xcode bringing commands like Cmd+L and Cmd+K to streamline code suggestions, error fixes, and codebase search directly within Xcode.

3. gptel: A simple LLM chat client for Emacs, with support for multiple models and backends. It works in the spirit of Emacs, available at any time and uniformly in any buffer.

4. Awesome LLM Apps: Build awesome LLM apps using RAG to interact with data sources like GitHub, Gmail, PDFs, and YouTube videos through simple text. These apps will let you retrieve information, engage in chat, and extract insights directly from content on these platforms.

Hot Takes

1. Frontier LLMs are very good at professional business writing, but they are not really close to the best humans writing original work with style or satisfying narrative.
   The gap here has not closed as fast as other abilities. I wonder whether this will be a hard barrier for LLMs. ~
   Ethan Mollick

   
   

2. I was pretty freaking close when I called it. AGI this year. We just blasted past human intelligence.
   Even if models couldn't get any smarter, just integrating this tech will change the world.
   But it's just going to keep getting smarter, faster, and cheaper.
   That's why I quit AI, to focus on what comes after. Post Labor Economics, Meaning Economy, a new social contract, decentralized future, postnihilism, transhumanism and everything that's coming.
   Always ahead of the curve. ~
   David Shapiro

Meme of the Day

“python script for exporting postgres as a csv”

That’s all for today! See you tomorrow with more such AI-filled content.

Don’t forget to share this newsletter on your social channels and tag Unwind AI to support us!

Unwind AI - X | LinkedIn | Threads | Facebook

Awesome LLM Apps | Sponsor Us

PS: We curate this AI newsletter every day for FREE, your support is what keeps us going. If you find value in what you read, share it with at least one, two (or 20) of your friends 😉