OpenAI Opensources Agents SDK

Today’s top AI Highlights:

1. OpenAI’s new API with built-in RAG, web search, and Computer Use

2. Opensource Manus AI that can research, browse, code, and automate tasks

3. Apache 2.0-licensed 21B reasoning model beats OpenAI o1-mini

4. Global directory of tool-using AI agents you can directly use

5. Embed 600+ integrations from 400+ APIs in your SaaS

& so much more!

Read time: 3 mins

AI Tutorials

OpenAI just released its Agents SDK, a rebranded, production-ready, and advanced version of the OpenAI Swarm framework to build multi-agent applications. Keep reading for more details👇. We couldn't wait to get our hands on it and build something useful.

In this tutorial, we'll walk you through building a multi-agent research assistant using OpenAI's Agents SDK. You'll create a system where multiple specialized agents work together to research any topic, collect facts, and generate comprehensive reports — all within a user-friendly application that's easy to use and extend.

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

Responses API, Computer Use, and Agents SDK 🌐📚🖥️

OpenAI has launched a new set of tools for developers to build intelligent agents that can handle complex tasks on their own. The new Responses API consolidates the functionality of the Chat Completions and Assistants APIs, providing a unified API to build applications with built-in vector store for RAG, web search capability, and computer use.

The new Agents SDK, replacing the experimental Swarm, is an opensource robust framework for orchestrating multi-agent workflows, complete with features like intelligent handoffs, observability, and built-in safety mechanisms. The best part — it supports not just OpenAI’s models but also other providers supporting the OpenAI API format.

Key Highlights:

Agents SDK

1. Multi-Agent Workflows - The Python-based Agents SDK is a complete framework where you can define agents with instructions, models, tools, and output_type (including Pydantic models for structured outputs). It manages the entire agent loop, handling LLM calls, tool execution, and handoffs.

2. Sophisticated Handoffs - The handoff function in the Agents SDK is far more than a simple routing mechanism. You can define handoffs with custom logic, including overriding tool names, descriptions, and even filtering the conversation history passed to the next agent.

3. Integrated Guardrails - The SDK comes with safety checks that run alongside your agents to validate inputs and outputs. These guardrails can prevent unwanted behaviors by running lightweight validation before your main agent even starts processing.

4. Tracing - Get full visibility into how your agents work with built-in execution traces. The tracing system shows you exactly what happened at each step. It also connects with external tools like Logfire, AgentOps, and Braintrust for more advanced monitoring.

Responses API

5. The Unified Builder - Replacing both Chat Completions (for simple interactions) and Assistants API (for tool use), the Responses API is now the recommended starting point. It offers a single endpoint for creating agents that can use OpenAI's built-in tools, handle multiple model turns, and manage data storage. Usage is billed at standard token rates, plus tool-specific costs.

6. Powerful Built-in Tools - The Responses API unlocks immediate access to:

   • Web Search: Provides cited, up-to-date information from the web.

   • File Search: Creates vector stores for RAG on the uploaded documents through semantic and keyword search.

   • Computer Use (Research Preview): Allows agents to interact with a computer's UI, uses the Computer Use model powering OpenAI Operator. Available to select developers (tiers 3-5). You can try this free hosted version of the model here.

Opensource Alternative to Manus AI Agent 🦉🤖

While everyone’s desperately waiting for Manus AI agent’s invitation code, GitHub is exploding with opensource alternatives for the same. After OpenManus, check out the OWL framework by Camel AI which is gaining serious traction. This multi-agent framework handles complex tasks like research, web browsing, and coding through multi-agent collaboration.

What's impressive is that OWL currently ranks #1 among open-source projects on the GAIA benchmark with a 58.18 average score. It works with multiple LLMs including Claude 3.7 Sonnet, DeepSeek, GPT-4o, and even local models through Ollama.

Key Highlights:

1. Multi-Agent Collaboration - OWL uses multiple specialized AI agents working together on a single task. When you submit a question, the system creates a "society" of agents that divide the work, share findings, and collectively solve problems

2. Extensive Toolkit Library - The framework includes 20+ pre-built toolkits that agents can use for browser automation, web search, document processing (PDF, Word, Excel parsing), code execution, and specialized tools for GitHub, Reddit, Google Scholar, and more.

3. Multimodal Capabilities - OWL supports processing and understanding of images, videos, and audio data. This expands the range of tasks agents can handle, from content analysis to multimedia generation (requires compatible LLMs).

4. Multi-Model - OWL works with OpenAI models, Claude 3.7 Sonnet, DeepSeek, and local LLMs via Ollama. You can switch between models by changing a single configuration parameter.

Quick Bites

Reka AI has released Reka Flash 3, a powerful, open-source 21B parameter reasoning model that beats OpenAI's o1-mini, under an Apache 2.0 license. The model excels at general chat, coding, instruction following, and function calling, making it ideal for low-latency or on-device applications where it can be compressed to as small as 11GB with 4-bit quantization.

LangChain has launched Agent Chat UI, a new open-source web app for interacting with any LangGraph application. This interface supports human/AI messages, tool calls, and human-in-the-loop interactions, you can also run it locally or use the deployed site.

Perplexity AI has launched its Windows desktop app, bringing AI-powered search directly to the desktop with features like voice dictation, keyboard shortcuts, and multiple search modes. The app offers access to Perplexity's latest AI models and is available for free or with a paid subscription for all the features.

Tools of the Trade

1. Colossal: A global directory of tool-using AI agents that you can directly integrate into your applications. Instead of building and maintaining your own tools, you can connect to pre-built agents that handle various services like Zendesk, Twilio, and Slack with a single API call.

2. ToolJet: Open-source low-code framework to build and deploy internal business apps with a drag-and-drop UI. You can integrate various data sources, including databases, API endpoints with OpenAPI specs, SaaS tools as well as object storage services to fetch and write data.

3. Nango: A developer-first platform to integrate 400+ APIs into your SaaS products, handling the messy details like auth, sync, and rate limits, through pre-built or custom-coded integrations. It's a more flexible alternative to restrictive unified API or embedded iPaaS solutions, built for developers, no restriction.

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. What's crazy when you read about Taiwan Semiconductors (TSMC) is you realize Nvidia are essentially dropshippers ~
   Pieter Levels
   
   

2. doing an AI agent startup seems hopelessly unoriginal and needlessly competitive until you realize the TAM is the ~entire economy~ ~
   James Campbell

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 😉