①Azuzre OpenAI×Milvus×GradioによるRAGのWebアプリケーションの構築

データの準備

data.py

# %%
from datasets import load_dataset

ds = load_dataset("neural-bridge/rag-dataset-12000")
ds

# %%
import pandas as pd

df_train = pd.DataFrame(ds['train'])
df_train

# %%
df_test = pd.DataFrame(ds['test'])
df_test

# %%
df = pd.concat([df_train, df_test])
df

# %%
df.rename(columns={'answer': 'reference'}, inplace=True)
df

# %%
import csv

df.iloc[:100].to_csv('data.csv', index=False, quoting=csv.QUOTE_ALL)

# %%

データをmilvusへinsert

milvus.py

# %%
import pandas as pd

filepath = "data.csv"
df = pd.read_csv(filepath)
df

# %%
from langchain_core.documents import Document

documents = []
for index, row in df.iterrows():
    question = row.question
    context = row.context
    reference = row.reference
    metadata = {'question': question, 'reference': reference}
    page_content = context
    documents.append(Document(page_content=page_content, metadata=metadata))
documents

# %%
len(documents)

# %%
from langchain_huggingface import HuggingFaceEmbeddings

embedding = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")

# %%
from langchain_milvus import Milvus

URI = "milvus.db"
connection_args = {"uri": URI}
vector_store = Milvus.from_documents(documents=documents, embedding=embedding, connection_args=connection_args, drop_old=True)

# %%

RAGの実装

rag.py

# %%
from langchain_huggingface import HuggingFaceEmbeddings

embedding = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")

# %%
from langchain_milvus import Milvus

URI = "milvus.db"
connection_args = {'uri': URI}
vector_store = Milvus(embedding_function=embedding, connection_args=connection_args)

# %%
from dotenv import load_dotenv

load_dotenv()

# %%
import os
from azure.identity import ClientSecretCredential, get_bearer_token_provider

tenant_id = os.environ.get('AZURE_TENANT_ID')
client_id = os.environ.get('AZURE_CLIENT_ID')
client_secret = os.environ.get('AZURE_CLIENT_SECRET')
credential = ClientSecretCredential(
    tenant_id=tenant_id,
    client_id=client_id,
    client_secret=client_secret
)

# %%
scopes = "https://cognitiveservices.azure.com/.default"
azure_ad_token_provider = get_bearer_token_provider(credential, scopes)

# %%
from langchain_openai import AzureChatOpenAI

azure_endpoint=os.environ.get('AZURE_OPENAI_ENDPOINT')
azure_deployment = os.environ.get("AZURE_OPENAI_DEPLOYMENT_NAME")
api_version = os.environ.get("API_VERSION")
temperature = 0
max_tokens = 4096
llm = AzureChatOpenAI(
    azure_endpoint=azure_endpoint,
    api_version=api_version,
    azure_deployment=azure_deployment,
    azure_ad_token_provider=azure_ad_token_provider,
    temperature=temperature,
    max_tokens=max_tokens,
)

# %%
from langchain_core.prompts import PromptTemplate

template = """# System:
Answer the question based on the given context.

# Context:
{context}

# Question:
{question}

# Answer:
"""
prompt_template = PromptTemplate.from_template(template=template)

# %%
#import pandas as pd

#filepath = "data.csv"
#df = pd.read_csv(filepath_or_buffer=filepath)
#df

# %%
#for index, row in df.iterrows():
#    context = row.context
#    question = row.question
#    try:
#        response = llm.invoke(input=prompt_template.format(context=context, question=question))
#        generated_text = response.content
#    except Exception as e:
#        generated_text = ""
#    print(generated_text)

# %%

Webアプリケーションの実装

app.py

# %%
import gradio as gr
from rag import vector_store, prompt_template, llm

# %%
def rag(question):
    try:
        results = vector_store.similarity_search(query=question)
    except Exception as e:
        context = ""
        generated_text = ""
        return generated_text, context

    context = ""
    for i in range(len(results)):
        context += f"# {i+1}"
        context += "\n"
        context += results[i].page_content
        context += "\n"
    prompt = prompt_template.format(context=context, question=question)

    response = llm.invoke(input=prompt)
    generated_text = response.content 

    return generated_text, context

# %%
with gr.Blocks() as app:
    question = gr.Textbox(label="question")
    generated_text = gr.Textbox(label="generated_text")
    context = gr.Textbox(label="context")
    button = gr.Button()
    button.click(fn=rag, inputs=[question], outputs=[generated_text, context])
    question.submit(fn=rag, inputs=[question], outputs=[generated_text, context])

# %%
app.launch(server_name="0.0.0.0")

# %%