演講 / Speaking

OpenAI Agent SDK 與 LangGraph 原始碼深度解析

Source Code Deep Dive into OpenAI Agent SDK vs LangGraph

日期 / Date
2025/08/09
類型 / Type
演講 / Speaking
活動 / Event
COSCUP x RubyConf Taiwan 2025
地點 / Location
國立臺灣科技大學 TR409-2

Context

內容摘要

中文

這是我在 COSCUP x RubyConf Taiwan 2025 的演講,COSCUP session 頁與 Pretalx 頁面均公開標示講者 Ko Ko、議程時間、場地與協作筆記。

本次演講從原始碼角度比較 OpenAI Agent SDK 與 LangGraph,拆解 Agent 執行迴圈、工具呼叫、記憶體管理、狀態轉移與多 Agent 協作。這不是只介紹使用方式,而是讓聽眾理解兩個框架在設計哲學、控制流程與可擴充性上的差異。

English

This talk was delivered at COSCUP x RubyConf Taiwan 2025. The COSCUP session page and Pretalx page publicly list Ko Ko as the speaker, along with the session time, room, and collaborative notes.

The talk compared OpenAI Agent SDK and LangGraph from a source-code perspective, covering agent execution loops, tool invocation, memory handling, state transitions, and multi-agent collaboration. It went beyond usage examples to explain the design philosophy, control flow, and extensibility tradeoffs between the two frameworks.

Details

技術分類與活動連結

主題

  • AI Agent
  • OpenAI Agent SDK
  • LangGraph
  • Source Code

Microsoft 技術

  • Azure OpenAI

Event Records

活動紀錄

COSCUP session 頁截圖,包含 Ko Ko 講者資訊
Screenshot / 截圖