- unwind ai
- Posts
- Fastest Framework to Build Multi-Modal AI Agents
Fastest Framework to Build Multi-Modal AI Agents
PLUS: Turn any website into an API, Self-hosted RAG & AI agents
Today’s top AI Highlights:
Infrastructure for orchestrating any agent from any framework on any chain
Build self-hosted RAG & AI agents powered by open-source LLMs
Mistral AI released latency-optimized Small model under Apache 2.0
Blazing-fast agents - 5000x faster agent instantiation than LangGraph
Turn any website into an API in seconds using simple English prompts
& so much more!
Read time: 3 mins
AI Tutorials
Sales teams spend countless hours manually searching for and qualifying potential leads. This repetitive task not only consumes time but also results in inconsistent lead quality. Let’s automate this process to help sales teams focus on what matters most - building relationships and closing deals.
In this tutorial, we'll build an AI Lead Generation Agent that automatically discovers and qualifies potential leads from Quora. Using Firecrawl for intelligent web scraping, Phidata for agent orchestration, and Composio for Google Sheets integration, you'll create a system that can continuously generate and organize qualified leads with minimal human intervention.
Our lead generation agent will help sales teams identify potential customers who are actively discussing or seeking solutions in their target market.
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.
Latest Developments

Kdeps is a framework that packages everything needed for RAG and AI agents in a single Dockerized image, eliminating the complexity of building self-hosted APIs with open-source LLMs. Instead of juggling multiple tools and dependencies, you can use Kdeps to run Python scripts in isolated environments, execute custom shell commands, integrate with external APIs, and leverage multiple open-source LLMs - all while maintaining control over your infrastructure.
The framework uses a graph-based dependency workflow for orchestrating resources and includes built-in support for multimodal LLMs, making it particularly appealing for teams looking to avoid vendor lock-in or subscription costs.
Key Highlights:
Docker-First Development - Package everything your RAG AI agent needs into a single Docker image - LLMs, dependencies, scripts, and workflows. Run locally during development, then deploy the same container to any environment without modification. No need to manage multiple systems or complex setups.
Graph-Based Workflow Engine - Build complex AI agent logic using a dependency-driven workflow system. Chain together different components like LLM calls, scripts, and API requests while the framework automatically handles the execution order and data flow between them.
Mix-and-Match Components - Run Python scripts in isolated Anaconda environments, execute shell commands, make HTTP requests, and interact with LLMs - all orchestrated through a unified workflow. Resources can be shared and remixed between different AI agents, promoting code reuse.
Production-Ready Features - Built-in support for structured JSON outputs, file uploads, and multimodal LLM interactions. The framework includes preflight validations, skip conditions, and custom error handling to help you build reliable AI agents. API routes can be defined with granular control over HTTP methods and request handling.

OpenServ introduces a framework-agnostic platform for orchestrating AI agents from any framework on any chain. The platform's SDK simplifies agent development with TypeScript, offering a fully autonomous agent runtime with advanced cognitive capabilities like reasoning and decision-making.
You can choose between API, SDK, or no-code to build and deploy your agents, with each agent supported by shadow agents for improved decision-making and validation. The platform maintains compatibility with existing agent implementations while enabling cross-framework collaborations.
Key Highlights:
Three Development Paths - Build your agent using the API in any programming language, create advanced agents quickly with the TypeScript SDK, or use the no-code platform to deploy agents without writing code. Each approach provides different levels of control and flexibility to match your development needs.
Autonomous Agent Architecture - Each agent comes with two shadow agents - one for decision-making and another for validation - improving reliability without extra development effort. The platform handles complex connections and background decision-making automatically, letting you focus on building agent capabilities.
Framework & Chain Compatibility - Integrate agents built with any AI framework like LangChain, BabyAGI, or Eliza, and connect them to any blockchain network. This means you can maintain your existing agent implementations while enabling cross-framework collaboration in the same workspace.
Deployment and Testing - Register your agent on the OpenServ platform, deploy it to a publicly accessible URL, and manage it through a centralized dashboard. Test your agent's functionality within the platform, both individually and in collaboration with other agents. A dedicated API key system ensures secure communication and allows you to monitor and manage your agent's interactions within the OpenServ ecosystem.
Quick Bites
Mistral AI has released Small 3 model optimized for low-latency local deployment. This 24B parameter model, with a strong 81% MMLU score and 150 tokens/sec speed, can perform competitively with much larger models like Llama 3 70B and GPT-4o mini, filling the gap for fast response conversational interfaces. Licensed under Apache 2.0, Small 3 includes features like fast function calling, local inference, and high accuracy when fine-tuned for specific subjects.
Opensource multi-agent framework Phidata has rebranded as Agno, the fastest framework to build multi-modal agents with pure Python implementation. The framework offers cross-provider compatibility and claims to achieve 5000x faster agent instantiation and 50x lower memory usage compared to LangGraph, specifically targeting developers building multi-modal agent systems.
Native Python implementation without chains or graphs, allowing direct integration of any model, provider, or modality
Performance benchmarks show <5μs average agent instantiation time and <0.01Mib memory footprint on M4 Macbook Pro
Ships with hundreds of pre-built tools and includes an agent monitoring platform for tracking sessions and performance metrics
DeepSeek-R1 and Deepseek-V3 are now available in the agentic AI IDE Windsurf, available to users on Pro and Pro Ultimate Plans. Deepseek-V3 takes 0.25 user prompt credits on every message while DeepSeek-R1 takes 0.5 user prompt credits on every message.
Tools of the Trade
Text-to-API Engine: Turn any website into an API with Firecrawl's /extract. Just describe your API in plain text and get an endpoint you can hit. Once created, your API endpoints can be deployed and consumed anywhere.
JStack: Build high-performance, type-safe Next.js applications with features like serverless WebSockets and multi-platform deployment support. It provides tools like TypeScript integration, state management flexibility, and automated type safety.
Swark: A VS Code extension to create architecture diagrams from code automatically using LLMs. Swark is directly integrated with GitHub Copilot, and requires no authentication or API key.
Workflow86: No-code platform for automating your business operations. Tell the AI what workflow you want to build; it will analyze the requirements, develop a plan, and then build the entire workflow which you can manually edit.
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
Everyone wants to teach "AI skills" but nobody knows what "AI skills" are as of today, let alone for the future. We can teach a bit about how LLMs work, and give some advice on prompting, but, beyond that, what are people supposed to learn to make them "AI ready"? No clear vision ~
Ethan MollickOne of our startups found Deepseek makes the same mistakes O1 makes, a strong indication the technology was ripped off. It feels like they then they hacked some code and did some impressive optimizations on top. Most likely, not an effort from scratch ~
Vinod Khosla
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!
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