Platform - More agent related
- SlideDeck: W6.2-team2-agent24-full
- Version: current
- Lead team: team-2
- Notes: more on tools for LLM agent
In this session, our readings cover:
Required Readings:
Agent Tools/Libraries:
Introduces a cohesive AutoGen ecosystem that includes the framework, developer tools, and applications. The framework’s layered architecture clearly defines each layer’s functionality. It supports both first-party and third-party applications and extensions. Microsoft Research announces AutoGen v0.4, a major update to their multi-agent AI framework. The new version introduces a complete redesign with an asynchronous, event-driven architecture that improves code quality, robustness, and scalability. Key features include modular components, built-in debugging tools, cross-language support, and enhanced observability through OpenTelemetry integration.
The update brings a new three-layered framework architecture consisting of core building blocks, AgentChat API, and extensions. It also introduces improved developer tools including AutoGen Bench for performance testing, an upgraded AutoGen Studio with real-time agent updates and visual team building, and Magentic-One, a new generalist multi-agent application for handling web and file-based tasks. The release maintains backward compatibility through the AgentChat API, making the migration from v0.2 straightforward while adding new capabilities like streaming messages and improved task progress management.
- https://docs.ag2.ai/docs/blog/2025-02-13-DeepResearchAgent/index
one Survey Blogpost on agent2024 …
- https://open.substack.com/pub/victordibia/p/ai-agents-2024-rewind-a-year-of-building?r=ya7nu&utm_medium=ios
OpenAI Operator
- https://cdn.openai.com/operator_system_card.pdf
- “OpenAI introduces Operator, a research preview of a browser-controlling agent available to Pro users in the U.S. Powered by the Computer-Using Agent (CUA) model, Operator can perform web-based tasks like filling forms, ordering groceries, and creating memes by interacting with graphical interfaces through typing, clicking, and scrolling. The agent leverages GPT-4o’s vision capabilities and reinforcement learning to navigate websites without requiring API integrations.”
- “multiple safety features, including user takeover mode for sensitive information, task limitations, and defenses against malicious websites. OpenAI is partnering with companies like DoorDash, Instacart, and Uber to refine the technology, while also exploring public sector applications. “
