- unwind ai
- Posts
- AI Agent for APIs
AI Agent for APIs
PLUS: OpenAI drops Realtime API price, Build and scale agentic apps
Shubham Saboo & Gargi Gupta
October 31, 2024
Today’s top AI Highlights:
The first AI agent to build low-latency integrations with platforms lacking official APIs
Build powerful AI Agents at scale with IBM’s new opensource framework
Apple releases new Mac and M4 chips line-up built for on-device AI
OpenAI lowers Realtime API’s price to upto 80% using prompt caching
Build RAG, semantic search, and other AI apps directly in PostgreSQL
& so much more!
Read time: 3 mins
AI Tutorials
AI tools are changing how we handle financial data, and building a team of AI agents that can act as financial analysts makes it even better.
This guide shows you how to set up a multi-agent financial analyst system using GPT-4o in just 20 lines of Python code. The system integrates:
• A web agent for general internet research
• A finance agent for detailed financial analysis
• A team agent for coordinating between agents
working together to deliver meaningful financial insights quickly.
We are using Phidata, a framework designed for building agent-based systems to streamline the entire 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
Building integrations with platforms that lack official APIs can be a big headache. But what if you could bypass this limitation and access those functionalities with a simple AI solution?
Integuru, a newly released open-source AI agent, does just that by reverse-engineering internal APIs. It generates integration code, making it easier than ever to connect to services even without official API documentation.
Key Highlights:
Effortless Reverse Engineering - Integuru automatically analyzes browser network requests, identifies dependencies, and generates runnable Python code to interact with a platform's internal API.
Simplified Workflow - Setting up Integuru is straightforward: define your desired action using a prompt, capture network requests, and run the agent.
Beyond Code Generation - Integuru provides a visual dependency graph, giving you a clear understanding of the platform's internal API structure.
Real-World Applications - The project already includes unofficial APIs for platforms like Venmo and Robinhood.
IBM just opensourced the Bee Agent Framework, a powerful new framework for building and deploying agentic apps at scale. This framework simplifies creating complex, agentic workflows so you can focus on application logic rather than infrastructure. It boasts features like built-in caching and robust error handling, crucial for production environments. Bee Agent integrates with popular LLMs and offers an OpenAI-compatible API, ensuring easy model swapping and broad compatibility.
Key Highlights:
Simple Agentic Workflow Construction - The framework provides a streamlined approach to building agent-based applications. It handles the complexities of managing LLM interactions, tool integrations, and memory management.
Seamless LLM Integration - Bee Agent supports multiple LLMs through a consistent interface, including Llama 3.1, making it easy to switch between models. The OpenAI-compatible Assistants API and Python SDK allow for rapid prototyping and integration with existing AI projects.
Production-Ready Features - Built-in features such as caching, sophisticated error handling, and MLFlow integration for traceability ensure your applications are robust and reliable, ready for deployment in production environments.
Optimized Resource Management - The framework incorporates flexible memory management strategies specifically designed to minimize token consumption, reducing LLM costs of your applications.
Quick Start - Getting started is straightforward with installation instructions (
npm install bee-agent-frameworkoryarn add bee-agent-framework) and comprehensive example code in the repository.
Quick Bites
Wrapping three announcement days, Apple has refreshed its Mac lineup with new M4 chips, focusing on AI performance. The M4 Pro and M4 Max, built on second-gen 3nm tech, feature improved Neural Engines that double the speed of previous models, enhancing AI tasks like Apple Intelligence. All new Macs, including Mac Mini, MacBook Air and MacBook Pro, now come with 16GB of RAM by default to meet the demands of on-device AI processing.
Boston Dynamic’s humanoid robot Atlas is now being trained in an automotive factory where it is autonomously handling automotive parts and reacting to environmental changes, without any teleoperation. It has officially joined the fleet of robots (Figure 01, Apptronik, and Tesla) being trained for this environment.
But what makes Atlas stand out is its 360-degree joint rotation, making it faster and more efficient than other humanoid robots out there and bypassing the limitations of the human body.
Some quick updates and scoops from OpenAI:
The Realtime API now has 5 new voices. Also, OpenAI is lowering the price by using prompt caching. Cached text inputs are discounted 50% and cached audio inputs are discounted 80%.
Advanced Voice Mode is now available in the ChatGPT desktop apps. Download the latest version to get access.
OpenAI has opensourced a factuality benchmark to evaluate how accurately language models answer short, fact-based questions. This will help measure factuality, calibration, and minimize hallucinations in AI responses.
To meet the growing AI chip demand, OpenAI is working with Broadcom and TSMC to build its first in-house chip. They have also begun using AMD’s chips alongside Nvidia’s to further secure the supply.
Tools of the Trade
pgai: Build semantic search, RAG, and other AI applications directly in PostgreSQL. It complements popular extensions for vector search in PostgreSQL like pgvector and pgvectorscale, building on top of their capabilities.
Lighteval: All-in-one toolkit for evaluating LLMs across various backends like Hugging Face, vllm, and Nanotron. It lets you run performance tests, create custom tasks, and store results locally, on S3, or the Hugging Face Hub for analysis.
HumanLayer: Adds human oversight to AI agents by allowing humans to approve or deny key function calls. It works across platforms like Slack, email, and SMS, and integrates easily with any AI framework to ensure safe agent control.
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
Software, as know it, will slowly die… It will be replaced by autonomous AI agents and we will use a chat box or a voice agent to interact with them ~
Bindu ReddyTomorrow’s sun will have two extra layers: the panelsphere, capturing its energy, and the datasphere, using it to train AIs. ~
Pedro Domingos
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