pip install lida

# 設定對應的api keyexport OPENAI_API_KEY=<API_KEY>

from dotenv import load_env 
import os load_dotenv() 

# read the .env file 
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

from lida import Manager, TextGenerationConfig , llm 
from lida.utils import plot_raster 
import warnings
from dotenv import load_dotenv
import os

load_dotenv()
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

warnings.filterwarnings("ignore")

# 初始化 LIDA
lida = Manager(text_gen = llm("openai", api_key=str(OPENAI_API_KEY))) # !! input your openai or other LLM api key
textgen_config = TextGenerationConfig(n=1, temperature=0.5, model="gpt-3.5-turbo-0301", use_cache=True)

import pandas as pd  
# 資料目前是使用官方推薦的資料集 

cars data = pd.read_csv("<https://raw.githubusercontent.com/uwdata/draco/master/data/cars.csv>") data.head()

# 數據摘要：從資料集生成簡短摘要
summary = lida.summarize( "https://raw.githubusercontent.com/uwdata/draco/master/data/cars.csv" , summary_method= "default" , textgen_cnotallow=textgen_config)   

print (summary)

# 目標生成：根據資料摘要生成視覺化圖表的目標, n=3 表示生成3 個目標
goals = lida.goals(summary, n= 3 , textgen_cnotallow=textgen_config) 

# 查看目前要生成的目標
for goal in goals: 
    print ( "=" * 20 ) 
    print ( f"Question: {goal.index} " ) 
    # print the question, visualization and rationale with each goal 
    print (goal.question) 
    print (goal.visualization) 
    print (goal.rationale) 

```輸出結果
==================== 
Question: 0
 What is the distribution of Retail_Price? 
histogram of Retail_Price 
This tells about the spread of prices of cars in the dataset . 
==================== 
Question: 1
 What is the distribution of Engine_Size__l_ among different car types? 
box plot of Engine_Size__l_ for each car type
 This will help  in identifying if there is  any difference in engine size among different car types. 
==================== 
Question: 2
 What is the relationship between Horsepower_HP_ and City_Miles_Per_Gallon? 
scatter plot of Horsepower_HP_ vs City_Miles_Per_Gallon 
This will help  in identifying if there is  any correlation between horsepower and fuel efficiency of cars.

library = "matplotlib"  # 可選"altair", "seaborn", "plotly", "matplotlib"

 textgen_config = TextGenerationConfig(n= 1 , temperature= 0.2 , use_cache= True ) 
for i in  range ( len (goals)): 
    # print the question, visualization and rationale with each goal 
    print ( "Question: " , goals[i].question) 
    print ( "Visualization: " , goals[i].visualization) 
    print ( "Rationale: " , goals[i] .rationale) 
    charts = lida.visualize(summary=summary, goal=goals[i], textgen_cnotallow=textgen_config, library=library) 
    plot_raster(charts[ 0 ].raster)

# 改變圖表顏色和字體大小
instructions = [ "change the color to red " , "scale the word size to 50%" ] 

edited_charts = lida.edit(code=charts[ 0 ].code, summary=summary, instructinotallow=instructions ) 
plot_raster(edited_charts[ 0 ].raster)

code = charts[ 0 ].code 
explanations = lida.explain(code=code, library=library, textgen_cnotallow=textgen_config) 

for row in explanations[ 0 ]: 
    print (row[ "section" ], " ** " , row[ "explanation" ]) 
    
# 輸出結果
accessibility ** The code creates a scatter plot using the matplotlib.pyplot library to visualize the relationship between two variables - Horsepower_HP_ and City_Miles_Per_Gallon. The plot is colored blue with an alpha value of 0.5 to show the density of the data points. The x-axis is labeled 'Horsepower_HP_'  and the y-axis is labeled 'City_Miles_Per_Gallon' . The title of the plot is  'What is the relationship between Horsepower_HP_ and City_Miles_Per_Gallon?' . 
transformation ** There is no data transformation happening in this code. The plot is made using the original data as it is . 
visualization ** The code first imports the required libraries - matplotlib.pyplot and pandas. The function plot() takes a pandas DataFrame as  input  and creates a scatter plot using the plt.scatter() method. The x-axis of the plot is the 'Horsepower_HP_' column of the input DataFrame and the y-axis is the 'City_Miles_Per_Gallon' column of the input DataFrame. The alpha parameter controls the transparency of the data points and the color parameter sets the color of the data points. The plt.xlabel() and plt.ylabel() methods add labels to the x-axis and y-axis respectively. The plt.title() method adds a title to the plot. The wrap parameter in plt.title() is  set to True to wrap the title text if it exceeds the width of the plot. Finally, the function returns the plot object .

evaluations = lida.evaluate(code=code, goal=goals[i], library=library)[ 0 ] 
for  eval  in evaluations: 
    print ( eval [ "dimension" ], "Score" , eval [ "score" ], " / 10" ) 
    print ( "\t" , eval [ "rationale" ][: 200 ]) 
    print ( "\t*********************** ***********" ) 

# 輸出結果
bugs Score 10 / 10
   No bugs, syntax errors, or typos found. 
***************** ***************** 
transformation Score 10 / 10
   No data transformation needed for a scatter plot. 
******************* *************** 
compliance Score 8 / 10
   The code meets the specified visualization goal, but the title could be improved by removing the question mark and rephrasing it as a statement. 
**** ****************************** 
type Score 9 / 10
   A scatter plot is an appropriate visualization type  for exploring the relationship between two continuous variables. 
********************************** 
encoding Score 9 / 10
   The data is encoded appropriately with Horsepower_HP_ on the x-axis and City_Miles_Per_Gallon on the y-axis. 
********************************** 
aesthetics Score 9 / 10
   The aesthetics of the visualization are appropriate with a blue color and an alpha of 0.5 to show overlapping points. ***************************** *****

textgen_config = TextGenerationConfig(n= 1 , temperature= 0 , use_cache= True ) 
recommended_charts = lida.recommend(code=code, summary=summary, n= 3 , textgen_cnotallow=textgen_config) 

print ( f"Recommended { len (recommended_charts)} charts " ) 
for chart in recommended_charts: 
    plot_raster(chart.raster) 
    pass

# 先繼承class 'lida.datamodel.Goal' 
from lida.datamodel import Goal 

# datamodel 總共有4 個object，分別是index, question, visualization and rationale
 custom_goal = Goal( 
    index= 0 , 
    questinotallow= "What is the distribution of the Type?" , 
    visualizatinotallow= "Bar Chart" , 
    ratinotallow= "The type of the car is an important feature of the dataset."
 ) 
# 生成圖表
custom_chart = lida.visualize(summary=summary, goal=custom_goal, textgen_cnotallow=textgen_config , library=library) 
plot_raster(custom_chart[ 0 ].raster) 
# 編輯客制化生成圖表
custom_instructions = [ "change the color to blue tone on tone color" ] # 改變Bar Chart 的顏色
edited_custom_charts = lida.edit(code= custom_chart[ 0 ].code, summary=summary, instructinotallow=custom_instructions) 
plot_raster(edited_custom_charts[ 0 ].raster)

pip install lida 

export OPENAI_API_KEY=<your key> 

lida ui --port=8080 --docs