Ubuntu, Open-webui, Ollama, Langchain 이용 자동 문서화 세팅

Prerequisite:
- Open-webui, Ollama 세팅

- 2025.03.25 - [컴퓨터/머신러닝 (Machine Learning)] - Ubuntu, AMD GPU, Ollama, open-webui setting

참고

https://github.com/open-webui/open-webui

GitHub - open-webui/open-webui: User-friendly AI Interface (Supports Ollama, OpenAI API, ...)

User-friendly AI Interface (Supports Ollama, OpenAI API, ...) - open-webui/open-webui

github.com

https://docs.openwebui.com/pipelines/

⚡ Pipelines | Open WebUI

DO NOT USE PIPELINES!

docs.openwebui.com

목표: 연구 개발 과정에서 사용되는 command들을 모아서 넣어주면 적절하게 자동으로 markdown 형식으로 documentation.

순서:

- Open WebUI 환경에서 문서화 정리를 요청

- LangChain과 Open WebUI의 pipeline 을 통해 ollama 에게 문서화 정리 prompt 전달

- Markdown 형식의 문서를 Open WebUI에 출력

Open WebUI (port:30000)
│
├── Ollama (port:11434)
│
└── Pipelines (port:9099)
     └── LangChain API (port:8000, host.docker.internal)

1. Langchain Setting

필요 패키지 설치

pip install fastapi uvicorn langchain ollama python-multipart

langchain을 구동할 app.py 작성

from fastapi import FastAPI, Form
from fastapi.middleware.cors import CORSMiddleware
from langchain.llms import Ollama
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain

app = FastAPI()

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_methods=["*"],
    allow_headers=["*"],
)

prompt_template = PromptTemplate(
    input_variables=["commands"],
    template="""
    You are a documentation assistant. Convert the given commands or scripts into clear, structured documentation in Markdown format.

    Commands or scripts:
    {commands}

    Documentation:
    """
)

@app.post("/generate-docs/")
async def generate_docs(commands: str = Form(...), model: str = Form(...)):
    llm = Ollama(base_url="http://localhost:11434", model="gemma3:12b")
    chain = LLMChain(llm=llm, prompt=prompt_template)
    result = chain.run(commands=commands)
    return {"documentation": result.strip()}

langchain 구동

uvicorn app:app --host 0.0.0.0 --port 8000

2. open-webui의 pipeline 기능과 langchain을 연결할 langchain.py 작성

*LANGCHAIN_API_URL의 주소를 위에서 구동한 langchain 이 구동된 ip주소와 포트 입력

"""
title: LangChain Documentation Generator
description: Sends commands to LangChain API for documentation generation.
version: 1.0
author: Your Name
"""

from typing import List, Union, Generator, Iterator
from schemas import OpenAIChatMessage
from pydantic import BaseModel, Field
import requests
import os

class Pipeline:

    class Valves(BaseModel):
        LANGCHAIN_API_URL: str = Field(
            default="http://192.168.0.41:8000/generate-docs/",
            description="LangChain API endpoint URL"
        )

    def __init__(self):
        self.valves = self.Valves(
            LANGCHAIN_API_URL=os.getenv("LANGCHAIN_API_URL", "http://192.168.0.41:8000/generate-docs/")
        )

    async def on_startup(self):
        # 서버 시작 시 호출되는 메서드 (필요시 초기화)
        pass

    async def on_shutdown(self):
        # 서버 종료 시 호출되는 메서드 (자원 정리 등)
        pass

    def pipe(
        self, 
        user_message: str, 
        model_id: str, 
        messages: List[dict], 
        body: dict
    ) -> Union[str, Generator, Iterator]:

        if not user_message.strip():
            return "No commands provided for documentation."

        payload = {
            "commands": user_message,
            "model": model_id
        }

        try:
            response = requests.post(self.valves.LANGCHAIN_API_URL, data=payload)
            response.raise_for_status()
            documentation = response.json().get("documentation", "No documentation generated.")
            return documentation
        except requests.RequestException as e:
            return f"Error communicating with LangChain API: {str(e)}"

3. Open-webui의 pipeline 컨테이너 부분 docker-compose.yml 파일 추가 및 구동

*위에서 준비한 langchain.py 파일이 담긴 폴더 마운트 필요 - "./data/pipelines:/app/pipelines" 부분

services:
  open-webui:
    image: 'ghcr.io/open-webui/open-webui:main'
    restart: always
    environment:
      OLLAMA_BASE_URL: "http://ollama:11434"  # <- 이 부분을 추가
    container_name: open-webui
    ports:
      - '30000:8080'
    volumes:
      - './data/open-webui:/app/backend/data'
      - './data/open-webui-plugins:/app/backend/data/plugins'
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
      - webui-network

  
  ollama:
    image: 'ollama/ollama:rocm'
    restart: always
    volumes:
      - './data/ollama:/root/.ollama'
    ports:
      - '11434:11434'
    devices:
      - "/dev/kfd"
      - "/dev/dri"
    group_add:
      - "${RENDER_GROUP_ID:-992}"
      - "video"
    networks:
      - webui-network

## pipeline 서비스를 위해 추가
  pipelines:
    image: 'ghcr.io/open-webui/pipelines:main'
    restart: always
    container_name: pipelines
    ports:
      - '9100:9099'
    volumes:
      - './data/pipelines:/app/pipelines'
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
      - webui-network

networks:
  webui-network:
    driver: bridge

이후 compose up