Reading
OpenAI officially launched a JavaScript/TypeScript multi-agent framework that supports complete features such as agent collaboration, tool calling, security protection, human intervention, session management, and real-time voice, and is compatible with the OpenAI API and third-party providers.
Section 01
OpenAI officially launched a JavaScript/TypeScript multi-agent framework that supports complete features such as agent collaboration, tool calling, security protection, human intervention, session management, and real-time voice, and is compatible with the OpenAI API and third-party providers.
Section 02
With the continuous improvement of large language model capabilities, a single agent can no longer meet the needs of complex tasks. From simple Q&A to complex business process automation, from text processing to real-time voice interaction, modern AI applications increasingly require multiple agents to work collaboratively, each performing its own duties. Against this background, OpenAI officially launched the JavaScript/TypeScript version of the Agents SDK—a lightweight yet powerful multi-agent workflow framework.
Section 03
The design philosophy of OpenAI Agents JS can be summarized with three key words: lightweight, powerful, and flexible. Compared to the Python version of the Agents SDK, the JS version is deeply optimized for the JavaScript ecosystem while maintaining compatibility with the OpenAI API and third-party providers. The framework supports multiple runtime environments such as Node.js 22+, Deno, Bun, and Cloudflare Workers, providing developers with great deployment flexibility.
Section 04
Agents are the basic building blocks of the framework. Each agent is an entity configured with a large language model and includes the following elements:
This design allows developers to build complex multi-agent systems like building blocks, where each agent focuses on a specific domain and collaborates to complete the overall task.
Section 05
The framework provides two agent collaboration modes:
Agents as Tools: An agent can call other agents as tools, which is suitable for hierarchical task decomposition. For example, a main agent is responsible for understanding user intent, then calls a specialized sub-agent to handle specific tasks.
Handoffs: An agent can fully transfer control to another agent, which takes over the subsequent conversation. This mode is suitable for scenario switching, such as handoff from a customer service agent to a technical support agent.
Section 06
The framework supports multiple types of tools, greatly expanding the agent's capability boundaries:
This diverse tool support allows agents to interact seamlessly with databases, APIs, external services, etc.
Section 07
When deploying AI systems in production environments, security is an indispensable element. The framework provides configurable security check mechanisms that support:
These guardrails can take effect at various stages before, during, and after agent operation, forming a comprehensive security guarantee.
Section 08
Not all decisions are suitable for full automation. The framework has a built-in human intervention mechanism that allows pausing agent execution at key nodes to wait for human confirmation or input. This is particularly important in the following scenarios:
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
Cross-generational hardware testing based on NVIDIA A10G and H100 GPUs, comparing and analyzing the throughput, latency, and concurrent scalability of three mainstream LLM inference frameworks. SGLang achieves a 3.4x performance advantage over vLLM on H100, while Ollama faces architectural bottlenecks in high-concurrency scenarios.