main.py

import io
from pydantic import BaseModel
from fastapi import FastAPI, HTTPException, File, UploadFile
from pdfminer.high_level import extract_text
from datetime import datetime
from fastapi.middleware.cors import CORSMiddleware
from sqlalchemy import create_engine, MetaData, Table, Column, Integer, String
from databases import Database

from langchain.chains.question_answering import load_qa_chain
from langchain_community.llms import HuggingFaceEndpoint
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import CharacterTextSplitter
from langchain.docstore.document import Document as LangchainDocument


import os

app = FastAPI()

# Set up CORS (Cross-Origin Resource Sharing) for allowing requests from all origins
origins=["*"]
app.add_middleware(
    CORSMiddleware,
    allow_origins=origins,
    allow_credentials=True,
    allow_methods=["GET", "POST", "PUT", "DELETE"],
    allow_headers=["*"],
)

# Define SQLAlchemy engine and metadata
DATABASE_URL = "sqlite:///./test.db"
engine = create_engine(DATABASE_URL)
metadata = MetaData()

# Define the document table schema
documents = Table(
    "documents",
    metadata,
    Column("id", Integer, primary_key=True),
    Column("filename", String),
    Column("upload_date", String),
    Column("content", String),
)

# Create the document table in the database
metadata.create_all(engine)

# Define Pydantic model for the document
class Document(BaseModel):
    filename: str
    upload_date: str
    content: str

# Function to save uploaded files
# async def save_uploaded_file(file: UploadFile, destination: str):
#     with open(destination, "wb") as buffer:
#         while chunk := await file.read(1024):
#             buffer.write(chunk)

# Endpoint for uploading PDF files
@app.post("/upload/")
async def upload_pdf(file: UploadFile = File(...)):

    # Check if the uploaded file is a PDF
    if not file.filename.lower().endswith('.pdf'):
        raise HTTPException(status_code=400, detail="Only PDF files are allowed.")
    
    # Read content of the uploaded PDF file
    content = await file.read()
    
    # Extract text from the PDF
    with io.BytesIO(content) as pdf_file:
        text_content = extract_text(pdf_file)
    
    # Create a document object
    doc = Document(filename=file.filename, upload_date=str(datetime.now()), content=text_content)
    
    # Insert the document data into the database
    async with Database(DATABASE_URL) as database:
        query = documents.insert().values(
            filename=doc.filename,
            upload_date=doc.upload_date,
            content=doc.content
        )
        last_record_id = await database.execute(query)
    
    # Save the uploaded PDF file
    # destination = f"files/{file.filename}"
    # await save_uploaded_file(file, destination)
    
    # Return the document object
    return doc

# Pydantic model for input data
class DataInput(BaseModel):
    responseData: str
    userInput: str

# Endpoint for processing user data
@app.post("/doc/")
async def process_data(data: DataInput):
    # Access responseData and userInput
    response_data = data.responseData
    user_input = data.userInput
    
    # Load required models and components from Langchain librar
    # os.environ['HUGGINGFACEHUB_API_TOKEN'] =HUGGINGFACEHUB_API_TOKEN
    # HUGGINGFACEHUB_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")
    os.environ.get("EXAMPLE")

    dom = [LangchainDocument(page_content=response_data, metadata={"source": "local"})]

    text_spliter = CharacterTextSplitter(chunk_size=3000, chunk_overlap=0)
    docs = text_spliter.split_documents(dom)

    embeddings = HuggingFaceEmbeddings()
    db = FAISS.from_documents(docs, embeddings)
    llm = HuggingFaceEndpoint(
        repo_id="google/flan-t5-large",
        temperature=0.8,
    )

    chain = load_qa_chain(llm, chain_type="stuff")

    # Perform similarity search and question answering
    dm = db.similarity_search(user_input)
    result = chain.run(input_documents=dm, question=user_input)
    
    return result