---
# System prepended metadata

title: Langflow
tags: [llm]

---

## Experiement
"What is langflow"라는 질문에 대해 chatgpt와 langflow로 만든 서비스간 결과물 차이비교
- chatGPT
    ![](https://hackmd.io/_uploads/HJW2wJEK3.png)
- langflow
    ![](https://hackmd.io/_uploads/BycluyVt2.png)
    ![](https://hackmd.io/_uploads/rJPG_1VKh.png)


## Langflow tools
- langflow: https://github.com/logspace-ai/langflow
- flowise: https://github.com/FlowiseAI/Flowise

## Setup (langflow)
- conda setup
    ```python=
    conda create -n langflow python=3.10
    conda activate langflow
    ```
- langflow init
    ```python=
    pip install langflow
    python -m langflow
    ```
    ![](https://hackmd.io/_uploads/SyVW8R7t3.png)

## Basic usage
- langflow UI
    ![](https://hackmd.io/_uploads/ry3qPRQY2.png)
    - LLM > ChatOpenAI: OpenAI모델 기반 LLM 사용 (OpenAI API Key 필요)
    - Memory > ConversationBufferMemory: 나눴던 대화내용을 기억할 수 있도록

- test
    - 우측 하단 버튼 누르면 현재 flow기반 테스트 가능
        ![](https://hackmd.io/_uploads/r1ymuAmF2.png =100x)
    - 테스트에서 "what is its population"의 it이 무얼 가르키는지 memory를 통해 이해하고 있음
        ![](https://hackmd.io/_uploads/B1-DOAmth.png)


- export langflow architecture
    ![](https://hackmd.io/_uploads/rkvHt0mt2.png)