app.py

import gradio as gr
import os
import re
import pandas as pd
import json
from ast import literal_eval
from langchain.llms import OpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
from langchain import OpenAI, ConversationChain
llm=OpenAI(openai_api_key=os.environ.get("OPENAI_API_KEY"))

def genCompetitorList(market):
    promptCompetitors = PromptTemplate(
        input_variables=["market"],
        template="You are a helpful Blueocean strategy advisor. Return top 3 {market}. Only response single-quote JSON array without reasoning.",
    )
    chain = LLMChain(llm=llm, prompt=promptCompetitors)
    responseCompetitorsArray = literal_eval(chain.run(market))
    print(responseCompetitorsArray)
    return gr.CheckboxGroup.update(choices=responseCompetitorsArray, interactive=True)


def genFeatureList(market,competitorList):
    promptFeatures = PromptTemplate(
        input_variables=["competitors"],
        template="return 3 benefits customers get or vendors provide, when customers use product and services of vendors {competitors}. Return only nouns, no adjective. Only response single-quote JSON array without reasoning.输出中文",
    )
    chain = LLMChain(llm=llm, prompt=promptFeatures)
    responseFeaturesArray=literal_eval(chain.run(' '.join(competitorList)))
    print(responseFeaturesArray)
    return gr.CheckboxGroup.update(choices=responseFeaturesArray, interactive=True)

def genRatingsPlot(market,competitorList,featureList):
    #responseRating='[{"competitor": "大理古城", "便利": 5, "体验": 4, "服务": 3}, {"competitor": "云龙", "便利": 2, "体验": 3, "服务": 4}, {"competitor": "双廊", "便利": 4, "体验": 5, "服务": 2}]'
    promptRatings = PromptTemplate(
        input_variables=["features","competitors"],
        template="return scoring 1 to 5 customers rate features:{features} for vendors:{competitors}.Only response single-quote JSON array without reasoning, features sequence and vendors sequence should be exactly the same as input, String \"competitor\" is the key cannot be omitted. Example: \"[\"competitor\": \"vendor1\", \"feature1\": 5, \"feature2\": 4 ; \"competitor\": \"vendor2\", \"feature1\": 2, \"feature2\": 3]\". 文字输出中文，标点符号用英文",
    )
    chain = LLMChain(llm=llm, prompt=promptRatings)
    responseRatings = chain.run({"features":' '.join(featureList),"competitors":' '.join(competitorList)})
    print(responseRatings.replace("'","\""))
    data = json.loads(responseRatings.replace("'","\""))
    df = pd.DataFrame(data)
    print(df)
    return gr.LinePlot.update(
        value=df,
        x="competitor",
        y=df.columns[1],
        x_title="竞品",
        y_title="特性评分",
        y_lim=[1,5],
        color_legend_position="right",
        width=500,
        height=300
    )

with gr.Blocks() as demo:

    market = gr.Textbox(label="细分市场：如大理附近小镇")
    btn1 = gr.Button(value="获取竞品")
    competitorList = gr.CheckboxGroup(label="竞品列表")
    btn1.click(genCompetitorList, inputs=[market], outputs=[competitorList])
    
    
    btn2 = gr.Button(value="获取特性")
    featureList = gr.CheckboxGroup(label="特性列表")
    btn2.click(genFeatureList, inputs=[market,competitorList], outputs=[featureList])
    
    
    btn3 = gr.Button(value="生成战略布局图")
    plot = gr.LinePlot(title="战略布局图")
    btn3.click(genRatingsPlot,inputs=[market,competitorList,featureList],outputs=[plot])

demo.launch()