IT 기술 따라잡기

https://youtu.be/N4t-YQufGLg?si=ySA0Toof6Ng69APf

Today, I focused on refining the AI web application we previously developed by undertaking some critical refactoring and implementing enhancements.

The refactoring work addressed an issue where session_state in Streamlit was unintentionally sharing message data across different pages. By ensuring that shared API keys are used consistently while individual messages are managed separately, the application now functions more reliably across various pages.

On the enhancement side, I expanded the application's capabilities by integrating OpenAI's ChatGPT alongside Anthropic's Claude, offering more flexibility in model usage.

In addition to these improvements, I also covered some practical tips for developers. These include committing changes in Visual Studio Code, pushing updates to GitHub, and using GitLens to compare local and server-side files. These insights are grounded in real-world scenarios and are essential for efficient development workflows.

Moreover, you'll see how to methodically troubleshoot when results don't match expectations—a vital skill in any developer's toolkit.

As part of my journey into LangGraph, I also combined the Quick Start tutorial code with Streamlit to create an interactive web page. This integration serves as an excellent example of adding tools to enhance LangGraph functionality.

Lastly, I've left one error in the code as a challenge for you to solve before our next session. It’s an excellent opportunity for us to explore problem-solving together.

Check out the full video here: https://youtu.be/N4t-YQufGLg?si=uPlA2xqlUDUZFrgw

드디어 LangGraph를 시작합니다.

StateGraph, Nodes, Edges, Compile, Stream

LangGraph의 가장 기본 개념을 활용해서 AI 어플리케이션을 만듭니다.
게다가 Streamlit 의 Session State를 사용해서 챗봇 기능을 구현했습니다.

LLM 은 Anthropic 사의 Claude를 사용합니다.

소스코드 공개합니다.

가져가셔서 맘껏 활용 하세요.

처음 배울 때 확실하게 배워서 AI 어플리케이션 개발에선 제대로 된 개발자가 됩시다.

오늘 배우는 내용은 이곳으로 가시면 어플리케이션과 소스코드를 보실 수 있습니다.


https://youtu.be/Ri3Nrpac3nI?si=gagZU75-rNr9uz19

Catchup LangGraph Tutorial · Streamlit

Coursera와 deepliarning.ai를 만들고 Google Brain 도 만들고 활동했던 스탠포드 대학의 앤드류 응 교수의 AI Agentic Workflows 프로젠테이션 입니다.

2024년도 들어서면서 AI 어플리케이션 개발의 트렌드로 각광받고 있는 AI Agentic Workflows 그리고 Multi Agent를 먼저 체험하고 그 경험을 공유합니다.

AI 로 어플리케이션을 개발해야 한다면 꼭 보아야 하는 강의 입니다.

예전에 몇번 봤는데 한국어로 번역하면서 꼼꼼이 보고 싶어서 한번 더 봤습니다.

역시나 놓치고 지나친 곳도 있고 다시 한번 생각 해 보니까 또 다른 insight도 생기고 정말 좋은 시간 이었습니다.

한글로 번역한 김에 Youtube Clip에 한글 자막 달아서 올립니다.

혹시 한번도 못 보신 분이나... 예전에 보셨더라도 한국어 자막을 보면서 더 자세히 이해해 보고 싶으신 분들 추천 드립니다.

이번 강좌에서는 앤드류 응 교수가 AI Agentic Workflow는 일을 시키면 금방 답이 안나올 수도 있어서 좀 기다려야 한다. 초보 매니저가 팀원 일 시켜놓고 5분마다 다 됐냐고 물어보는 식으로 접근하면 안된다고 얘기했을 때 누군가가 크게 웃었거든요.
그 사람이 예전에 수시로 확인하면서 Micro Management를 했던 옛 상관이 생각나서 웃었는지 아니면 AI 는 기계인데 사람처럼 대하라는 말이 웃겨서 웃었는지는 모르겠는데. 하옇든간 크게 웃었습니다.
그 웃음소리를 듣고 앤드류 응 교수가 제가 보기에 동양인 특유의 그런 어떤 웃음을 지었는데 수줍은 웃음인 듯도 하고 (설명은 못하겠고 직접 보세요.) 그 웃는 표정이 가장 자연스럽고 활짝 웃었던 웃음 인것 같았습니다.

그래서 그 장면을 썸네일로 사용했습니다.

https://youtu.be/8b7CLBCS3pg