- unwind ai
- Posts
- Opensource Memory Layer for AI Agents
Opensource Memory Layer for AI Agents
PLUS: Eleven Labs text-to-voice tool, New Apple Intelligence APIs
Shubham Saboo & Gargi Gupta
October 24, 2024
Today’s top AI Highlights:
Opensource memory layer and knowledge graphs for AI agents
Build AI apps with real-time agent interaction and state sync
Eleven Labs now lets you generate a unique voice with text prompts
Apple rolls out new Apple Intelligence features and three APIs
Free opensource app that lets Claude's new Computer Use take over your computer
& so much more!
Read time: 3 mins
AI Tutorials
RAG is becoming a game-changer for applications that need accurate information from large datasets. As developers, we know the value of building tools that can search documents and provide relevant answers quickly. Today, we’ll take that one step further.
In this tutorial, we’ll walk you through building a production-ready RAG service using Claude 3.5 Sonnet and Ragie.ai, integrated into a clean, user-friendly Streamlit interface. With less than 50 lines of Python code, you’ll create a system that retrieves and queries documents—ready for real-world use.
What is Ragie.ai?
Ragie.ai is a fully managed RAG-as-a-Service for developers. It offers connectors for services like Google Drive, Notion, and Confluence, along with APIs for document upload and retrieval. It handles the entire pipeline—from chunking to hybrid keyword and semantic searches—so you can start with minimal setup.
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.
🎁 Bonus worth $50 💵
Latest Developments
memary is an opensource memory layer to equip AI agents with human-like memory and enhance their ability to store and recall information across interactions. It uses knowledge graphs to store entities and events, which helps AI agents maintain context and continuity. This system supports smooth integration with locally running models, defaulting to those available via Ollama, like Llama 3 and LLaVA. memary also supports multi-agent setups through graph-based memory management.
Key Highlights:
Knowledge Graph - Memary uses knowledge graphs to store knowledge for the agent to organize information contextually. It supports multi-hop reasoning, allowing the agent to retrieve related nodes and efficiently search through knowledge subgraphs instead of the entire graph.
Memory Management - Memary automates memory updates as your agent interacts. It uses two key memory modules: the Memory Stream, which logs all interactions with timestamps, and the Entity Knowledge Store, which tracks how frequently and recently each entity is referenced. This ensures the agent understands both the breadth and depth of user interactions over time.
Local Model Preference - It defaults to local LLMs like Llama 3 and vision models like LLaVA. You can switch between models like GPT-3.5-turbo and GPT-4-vision if cloud-based functionality is needed.
Multi-Agent Setup - You can create multiple agents with distinct memory contexts using FalkorDB. Each agent maintains separate knowledge graphs, allowing seamless switching between them based on tasks or personas.
Building AI-agent-native applications just got easier with CoAgents, a new tool that bridges the gap between your application front-end and LangGraph AI agents. CoAgents simplifies integrating custom AI agents directly into your app's UI/UX. It allows for dynamic, real-time interactions, creating a more responsive and intuitive user experience. You can now build applications with complex AI functionalities, similar to Replit's AI agent, with significantly less effort, letting LangGraph handle the agent infrastructure while CoAgents manages the connection to the application interface.
Key Highlights:
Shared State (Agent ↔ App) - You can synchronize state between the app and the AI agent in real-time, giving the agent continuous access to the app's status and sharing agent progress with users through intermediate state streaming.
Agentic Generative UI - CoAgents provides a state-based generative UI that reflects the agent’s real-time actions, ensuring users see exactly what the agent is doing. This feature prevents idle waiting and builds user trust by keeping agents' tasks visible.
Human-in-the-Loop - You can define points in the agent’s process that require human approval or input. This helps with better control and minimizes risks, especially for workflows where automated decisions could have significant impacts.
Agent Steering and Checkpoints (Coming Soon) - This will enable you to set breakpoints and save agent states at key points. Users will be able to rewind actions mid-process and correct the agent’s trajectory if needed.
Quick Bites
ElevenLabs introduces Voice Design to generate custom unique voices from text prompts by specifying age, accent, tone, or character. It is great for creating character voices like “An angry old pirate, shouting” or “A massive evil ogre, troll.” The API will be available in 1 week.
Here’s a new opensource speech-to-text model Moonshine which is 1.7x faster performance than the current SOTA OpenAI’s Whisper while matching its accuracy. Optimized for real-time, on-device use, it processes 10-second audio 5x faster and requires as little as 8MB RAM, making it ideal for resource-constrained devices
Runway has unveiled Act-One to generate expressive character performances with Gen-3 Alpha model using simple video inputs. It uses a video of a person’s performance, captures its essence, and transposes it to a completely new generated character. It works with a single driving video that can be shot on something as simple as a cell phone. Act-One will be gradually rolled out to users.
Apple has rolled out developer betas of iOS 18.2, iPadOS 18.2, and macOS Sequoia 15.2, introducing new Apple Intelligence features like Genmoji, Image Playground, and ChatGPT integration. Along with these updates come APIs for three key Apple Intelligence features: Writing Tools API, Genmoji API, and Image Playground API.
Tools of the Trade
Agent.exe: A free, opensource app that lets Claude 3.5 Sonnet control your computer with the new computer-use API. It works on Mac, Windows, and Linux, allowing the model to perform tasks like navigating browsers and executing commands.
Agentic LLM Vulnerability Scanner: Opensource tool for testing LLM security with customizable attacks and stress testing. It integrates with LLM APIs and supports fuzzing and dataset-driven vulnerability checks.
HuggingChat macOS: A native chat app for macOS that uses opensource language models for AI-powered conversations. It offers a seamless desktop experience with easy installation.
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
One of the things that excites me the most about Computer use is the effect it will have on UI/UX.
The web is designed for humans, but what will an AI-first website look like?
I can see a future where sites will have two versions, similar to the early days of mobile web. ~
Pietro SchiranoMira, Ilya, Elon, Sam, and Dario are now all competing with each other for AGI despite all having worked together at OpenAI just a few years ago ~
James Campbell
Meme of the Day
That’s all for today! See you tomorrow with more such AI-filled content.
🎁 Bonus worth $50 💵
Share this newsletter on your social channels and tag Unwind AI (X, LinkedIn, Threads, Facebook) to get AI resource pack worth $50 for FREE. Valid for a limited time only!
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 😉
Reply