main.py

from fastapi import FastAPI,Request,File,UploadFile,Form
from fastapi.templating import Jinja2Templates
from fastapi.staticfiles import StaticFiles
from fastapi.responses import HTMLResponse,JSONResponse
from fastapi.middleware.cors import CORSMiddleware
import pandas as pd
import re 
import io
import base64
import matplotlib.pyplot as plt 
import torch
import tensorflow as tf

import fitz
from docx import Document
from pptx import Presentation
import seaborn as sns
import PIL.Image as Image
import fitz

from huggingface_hub import snapshot_download
from transformers import (
    AutoTokenizer, AutoModelForSeq2SeqLM,
    AutoModelForCausalLM,pipeline
)


try: 
    print("[Info] installing Salesforce/blip-image-captioning-base ....")
    blip_dir = "./models/blip-base-tf"
    snapshot_download("Salesforce/blip-image-captioning-base", local_dir=blip_dir, local_dir_use_symlinks=False)
    interpreter = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
    print("[Info] Salesforce/blip-image-captioning-base is installed.")
except Exception as exp:
    print("Can't load the model Salesforce/blip-image-captioning-base")
    print(f"[Error] {str(exp)}")


try:
    print("[Info] installing facebook/bart-large-cnn ....")
    bart_dir = "./models/bart-large-cnn"
    snapshot_download("facebook/bart-large-cnn", local_dir=bart_dir, local_dir_use_symlinks=False)
    bart_tokenizer = AutoTokenizer.from_pretrained(bart_dir)
    bart_model = AutoModelForSeq2SeqLM.from_pretrained(bart_dir)
    summarizer = pipeline("summarization", model=bart_model, tokenizer=bart_tokenizer)
    print("[Info] facebook/bart-large-cnn is installed")
except Exception as exp:
    print("Can't load the model facebook/bart-large-cnn")
    print(f"[Error] {str(exp)}")

try:
    print("[Info] installing deepseek-ai/deepseek-coder-1.3b-instruct ")
    deepseek_dir = "./models/deepseek-coder"
    snapshot_download("deepseek-ai/deepseek-coder-1.3b-instruct", local_dir=deepseek_dir, local_dir_use_symlinks=False)
    deepseek_tokenizer = AutoTokenizer.from_pretrained(deepseek_dir, trust_remote_code=True)
    deepseek_model = AutoModelForCausalLM.from_pretrained(deepseek_dir, trust_remote_code=True)
    generator = pipeline("text-generation", model=deepseek_model, tokenizer=deepseek_tokenizer)

    print("[Info] deepseek-ai/deepseek-coder-1.3b-instruct is installed")
except Exception as exp:
    print("Can't load the model deepseek-ai/deepseek-coder-1.3b-instruct")
    print(f"[Error] {str(exp)}")


app=FastAPI()
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)
MAX_SIZE= 1 * 1024 *1024

app.mount("/static",StaticFiles(directory='static'),'static')
templates = Jinja2Templates(directory='templates')

@app.get("/",response_class=HTMLResponse)
def index(req:Request):
    return  templates.TemplateResponse('index.html',{'request':req})
@app.get("/summarization",response_class=HTMLResponse)
def index(req:Request):
    return  templates.TemplateResponse('text-summarization.html',{'request':req})
@app.get("/datavisualisation",response_class=HTMLResponse)
def index(req:Request):
    return  templates.TemplateResponse('data-visualization.html',{'request':req})
@app.get("/imageinterpretation",response_class=HTMLResponse)
def index(req:Request):
    return  templates.TemplateResponse('image-interpretation.html',{'request':req})




    
@app.post("/interpret")
def interpret(file_img:UploadFile=File(...)):
    extension = file_img.filename.split(".")[-1]
    Supported_extensions = ["png","jpg","jpeg"]
    if extension not in Supported_extensions:
        return JSONResponse(content={"error": "Unsupported file type"},status_code=400)
    image = Image.open(file_img.file)
    global interpreter
    try:
        caption = interpreter(image)
    except Exception as exp:
        return JSONResponse(content={"error": "Can't interpret the image "},status_code=400)
    
    return  JSONResponse(content={"caption": caption[0]['generated_text']},status_code=200)

@app.post("/summerize")
def summerzation(file:UploadFile=File(...)):
    try:
        
        extension = file.filename.split(".")[-1]
        supported_ext=["pdf","xlsx","docx","ppt","xls"]
        if extension not in supported_ext :
            return JSONResponse(content={"error": "Unsupported file type"},status_code=400)
        if file.size > MAX_SIZE:
            return JSONResponse(content={"error": "file is too large "},status_code=400)
            
        
        file_bytes = file.file.read()
        
        
        if extension == "pdf":
            text = get_text_from_PDF(file_bytes)
        elif extension == "docx":
            text = get_text_from_DOC(file_bytes)
        elif extension == "pptx":
            text = get_text_from_PPT(file_bytes)
        elif extension == "xlsx":
            text = get_text_from_EXCEL(file_bytes)
        
        
        if not text.strip():
            return JSONResponse(content={'error':'File is emplty'},status_code=400)
        
        result=""
        global summarizer
        for i in range(0, len(text), 1024):
            try:
                summary = summarizer(text[i:i+1024], max_length=150, min_length=30, do_sample=False)
                result += summary[0]['summary_text']
            except Exception as e:
                return JSONResponse(content={"error": f"Summarization failed: {str(e)}"},status_code=403)
        return JSONResponse(content={"summary": result},status_code=200)
    except Exception as exp:
        return JSONResponse(content={"error":"Internel Server Error:"+str(exp)} ,status_code=500) 


@app.post("/plot")
async def plot(user_need:str=Form(...),file:UploadFile=File(...)):
    try:
        extension = file.filename.split(".")[-1]
        Supported_extensions = ["xlsx","xls"]
        if extension not in Supported_extensions:
            return JSONResponse(content={"error": "Unsupported file type"},status_code=400)
        file_bytes =  file.file.read()
        
            
        if len(file_bytes) >  MAX_SIZE :
            return JSONResponse(content={"error": "too large file "},status_code=400)
        
   
        
        df = pd.read_excel(io=io.BytesIO(file_bytes))
        

        prompt = [
    {"role": "system", "content": f'''You are a helpful assistant that helps users write Python code.
-you have to write the code to solve the task using the dataset df.
-you can use pandas to manipulate the dataframe.
-you can use matplotlib to plot the data.
-you can use seaborn to plot the data.
-don't use print or input statements in the code.
-don't use any other libraries except pandas, matplotlib, seaborn.
-don't use any other functions except the ones provided in the libraries.
-don't write the code for the dataframe creation.
-check if the columns has a nan values and raise exception if yes .
-exclude plt.show() from the code.
-you have to write the code in a markdown code block.
-make sure that the type of the chart is compatible with the dtypes of the columns 
-use only the column specified in the task.
-you have an information about the dataframe called df contains the following information:
df.columns:{df.columns.to_list()}
df.dtypes:{df.dtypes.to_dict()}
-you have to extract the column names and the plot type  from the prompt bellow  and use them in the code.
-if the user task is not clear or there is an error like the column names are not in the dataframe, raise an 
error.
    '''},
{"role": "user", "content": user_need},
]
        
        global generator
        output = generator(prompt, max_length=1000)
        
        match = re.search(r'```python(.*?)```', output[0]['generated_text'][2]['content'], re.DOTALL)
        
        code =''
        if not match:
          return JSONResponse(content={"error": "No Code was Generated"},status_code=403)
          
        code = match.group(1).replace("plt.show()\n","")
        
        
        safe_globals={
            "plt": plt,
            "sns": sns,
            "pd": pd,
            "df": df
        }
        try:
            exec(code,safe_globals)
            buf = io.BytesIO()
            plt.savefig(buf, format='png')
            buf.seek(0)
            base64_image = base64.b64encode(buf.getvalue()).decode('utf-8')
            
            return JSONResponse(content={"plot": f"data:image/png;base64,{base64_image}",'code':code},status_code=200)
        except Exception as e:
            print(e)
            return JSONResponse(content={"error":"Can't execute the code : "+ str(e) },status_code=400)
    except Exception as exp:
        print(exp)
        return JSONResponse(content={"error":"Internel Server Error: "+str(exp)} ,status_code=500)       






def get_text_from_PDF(file_content):
  doc = fitz.open(stream=file_content, filetype="pdf")
  text = ""
  for page in doc:
      text += page.get_text()
  return text

def get_text_from_PPT(file_content):
  prs = Presentation(io.BytesIO(file_content))
  text = ""
  for slide in prs.slides:
      for shape in slide.shapes:
          if hasattr(shape, "text"):
              text += shape.text
  return text


def get_text_from_DOC(file_content):
  doc = Document(io.BytesIO(file_content))
  text = ""
  for paragraph in doc.paragraphs:
      text += paragraph.text
  return text

def get_text_from_EXCEL(file_content):
  df = pd.read_excel(io=io.BytesIO(file_content))
  text = df.to_string()
  return text