Build AI Agents & RAG apps with NoCode

Today’s top AI Highlights:

1. Build smarter AI with knowledge graphs and logic, not just vectors

2. No-code platform to build full-stack AI apps with RAG and multiple agents

3. Lightweight library to build and run AI agents with a few lines of code

4. AI to auto-fix bad code and security vulnerabilities

& so much more!

Read time: 3 mins

AI Tutorials

Ever had your RAG system confidently give completely irrelevant information? Or watched it stubbornly stick to outdated data when better sources were just a web search away? You're not alone. Traditional RAG systems, while powerful, often act like that one friend who never admits when they need to double-check their facts.

In this tutorial, we'll fix that by building a Corrective RAG Agent that implements a multi-stage workflow with document retrieval, relevance assessment, and web search. Using LangGraph's workflow capabilities, we'll create a system that can evaluate its responses, adapt on the fly, and even reach out to the web when its local knowledge falls short. Think of it as RAG with a built-in fact-checker and research assistant.

We'll combine the analytical prowess of Claude 3.5 Sonnet with LangGraph's flexible workflow engine. By the end of this tutorial, you'll have a RAG system that's not just smarter but also more honest about what it knows (and doesn't know).

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 leveling 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

Logical Reasoning meets Semantic RAG 🧠

KAG is a new framework that elevates traditional RAG systems by combining logical reasoning with retrieval to handle complex queries for domain-specific knowledge bases. It takes a unique approach to processing both unstructured and structured data, creating a unified knowledge graph that can work with everything from documents to business rules.

What sets it apart is its ability to bring together multiple reasoning approaches for more accurate results - in testing, it showed significant improvements over existing methods.

Key Highlights:

1. Smart Knowledge Integration - KAG processes both unstructured content (like PDFs and Word docs) and structured data through a flexible pipeline that handles layout analysis, knowledge extraction, and semantic alignment. The mutual indexing between graph structure and text blocks makes retrieval more efficient.

2. Hybrid Reasoning Engine - The framework combines multiple problem-solving approaches - retrieval, knowledge graph reasoning, language reasoning, and numerical calculations - guided by logical forms. You can mix and match different operators like exact match retrieval, text search, or semantic reasoning based on your specific use case.

3. Ready for Production - KAG offers both a product mode for quick setup and a toolkit mode for developers. You can start with a Docker-based installation for the complete stack, or use the Python package to integrate specific components. The framework supports PostgreSQL for production deployments and SQLite for local development.

4. Model Agnostic - Built to work with various LLM providers through LiteLLM, including local models via llama-cpp-python. You can use OpenAI-compatible APIs like Qwen/DeepSeek or deploy models locally using vLLM/Ollama. The system also supports custom implementations through clear extension points for builders, solvers, and model integrations.

Build RAG and AI Agents Without Writing Backend Code 🧑‍💻

Momen is a no-code web app development platform that provides a robust suite of features to build AI agents and RAG into projects. The framework handles everything from frontend and backend to database operations, letting you focus on building your AI features. Along with its RAG engine that supports multiple data sources, Momen provides built-in tools for task planning and execution through multiple collaborating agents.

A particularly useful feature is its structured output system that binds AI responses directly to UI components and enables chaining with APIs and actions.

Key Highlights:

1. RAG Implementation - Integrate context from multiple sources including databases, APIs and files through the RAG engine. Create embeddings for any database text field to enable similarity-based retrieval, with built-in optimization for chunking and context handling. The system handles all the complexity of RAG while giving you full control over the pipeline.

2. Agent Architecture - Build AI applications where multiple agents can collaborate and be invoked through your backend workflow. The framework provides tools for task planning, execution monitoring, and agent coordination. Agents can access data sources, call external APIs, and interact with users through a human-in-the-loop interface.

3. Production-Ready Features - Take advantage of built-in user management with SSO and RBAC, Stripe integration for payments, and over 30 ready-to-use UI components. The BaaS solution exposes your AI agents through public GraphQL APIs, making it easy to integrate AI capabilities into existing applications.

4. Structured Outputs - Get more than just text strings from your AI - receive fully typed structured data that can be bound to different UI components or used to update application state. This enables seamless integration between AI outputs and your application logic, including chaining responses to other actions and API calls.

Quick Bites

Meta researchers have introduced COCONUT (Chain of Continuous Thought), a new way for LLMs to reason that replaces traditional step-by-step text reasoning with “continuous thoughts" in a latent space. Rather than generating explicit reasoning steps in natural language, COCONUT lets models explore multiple possible reasoning paths simultaneously and achieves better results on complex planning tasks while using fewer tokens.

Hugging Face just launched SmolAgents, a lightweight library to create and run powerful AI agents with just a few lines of code. This library supports code-writing agents and tool-calling agents, and integrates with various LLMs, including open-source options, and provides sandboxed environments for secure code execution.

OpenAI and Andrew Ng's DeepLearning.AI have launched a free 70-minute course called "Reasoning with o1," taught by OpenAI's Head of AI Solutions Colin Jarvis. The course focuses on effectively using OpenAI's o1 model for complex reasoning tasks like coding and image analysis, teaching techniques from basic prompting to advanced concepts like meta-prompting and model orchestration.

Tools of the Trade

1. CodeAnt AI: AI to detect and fix code quality issues, security vulnerabilities, and anti-patterns across 30+ programming languages. It integrates with popular development environments (VS Code, JetBrains) and version control platforms (GitHub, GitLab, etc.) to automatically analyze pull requests, enforce code standards, and give one-click fixes.

2. AI Agents Marketplace: Open-source platform to discover and access 100+ AI agents across different categories like coding, productivity, and data analysis. You can find AI agents for specific tasks, and developers can list their own AI agents for free to reach users worldwide.

3. Fabrice AI: A lightweight, functional TypeScript framework for creating multiple collaborative AI agents. You can define agents, workflows, and tools in a simple, composable way without needing classes or complex infrastructure.

4. Awesome LLM Apps: Build awesome LLM apps with RAG, AI agents, and more to interact with data sources like GitHub, Gmail, PDFs, and YouTube videos, and automate complex work.

Hot Takes

1. LLMs would advance so much faster if companies published their research instead of waiting for Deepseek to render it obsolete ~
   Tom Dörr

   
   

2. The world is going to look shockingly similar in 5 years, despite massive technological innovation enabled by AI. ~
   Logan Kilpatrick

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 😉