app.py

import logging
import os
from fastapi import FastAPI, UploadFile, File, HTTPException
from fastapi.responses import HTMLResponse, JSONResponse
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
from langchain.vectorstores import Chroma
from langchain.llms import OpenAI
from langchain.vectorstores.cassandra import Cassandra
from langchain.indexes.vectorstore import VectorStoreIndexWrapper
from langchain.chains import RetrievalQA
from langchain.document_loaders import PyPDFLoader
from langchain.vectorstores.base import VectorStoreRetriever
from langchain.text_splitter import CharacterTextSplitter
from azure.core.credentials import AzureKeyCredential
from azure.ai.inference import EmbeddingsClient
import cassio
from pydantic import BaseModel
import shutil
from config import settings
app = FastAPI()
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

logging.basicConfig(level=logging.ERROR)
logger = logging.getLogger(__name__)


HUGGINGFACE_API_KEY = settings.huggingface_key
ASTRA_DB_APPLICATION_TOKEN = settings.astra_db_application_token
ASTRA_DB_ID = settings.astra_db_id
OPENAI_API_KEY = settings.openai_api_key
GITHUB_TOKEN = settings.github_token
AZURE_OPENAI_ENDPOINT = settings.azure_openai_endpoint
AZURE_OPENAI_MODELNAME = settings.azure_openai_modelname
AZURE_OPENAI_EMBEDMODELNAME = settings.azure_openai_embedmodelname




UPLOAD_FOLDER = '/uploads'
conversation_retrieval_chain = None
chat_history = []
llm = None
embedding = None
cassio.init(token=ASTRA_DB_APPLICATION_TOKEN, database_id=ASTRA_DB_ID)

class MessageRequest(BaseModel):
    userMessage: str
class AzureOpenAIEmbeddings:
    def __init__(self, client):
        self.client = client
        self.model_name = AZURE_OPENAI_EMBEDMODELNAME  # Store model name

    def embed_query(self, text: str):
        """Embed a query."""
        response = self.client.embed(
            input=[text],
            model=self.model_name
        )
        return response.data[0].embedding

    def embed_documents(self, texts: list):
        """Embed a list of documents."""
        response = self.client.embed(
            input=texts,
            model=self.model_name
        )
        return [item.embedding for item in response.data]

def init_llm():
    global llm, embedding
    llm = OpenAI(
        base_url=AZURE_OPENAI_ENDPOINT,
        api_key=GITHUB_TOKEN,
        model=AZURE_OPENAI_MODELNAME
    )
    embedding = EmbeddingsClient(
        endpoint=AZURE_OPENAI_ENDPOINT,
        credential=AzureKeyCredential(GITHUB_TOKEN),
        model=AZURE_OPENAI_EMBEDMODELNAME
    )

def process_document(document_path):
    init_llm()
    global conversation_retrieval_chain
    loader = PyPDFLoader(document_path)
    documents = loader.load()
    text_splitter = CharacterTextSplitter(
        chunk_size=800,
        chunk_overlap=200,
    )
    raw_text = "".join([doc.page_content for doc in documents])
    texts = text_splitter.split_text(raw_text)
    custom_embedding = AzureOpenAIEmbeddings(embedding)
    astra_vector_store = Cassandra(
        embedding=custom_embedding,
        table_name="qa_mini_demo",
        session=None,
        keyspace=None,
    )
    astra_vector_store.add_texts(texts[:500])
    retriever = VectorStoreRetriever(
        vectorstore=astra_vector_store, search_type="mmr", search_kwargs={'k': 1, 'lambda_mult': 0.25}
    )
    conversation_retrieval_chain = RetrievalQA.from_chain_type(
        llm=llm,
        chain_type="stuff",
        retriever=retriever,
        return_source_documents=False,
        input_key="question"
    )

def process_prompt(prompt):
    init_llm()
    global chat_history
    global conversation_retrieval_chain

    output = conversation_retrieval_chain({"question": prompt+"you should only give answer to the question ,do not give any other information", "chat_history": chat_history})
    answer = output["result"]

    chat_history.append((prompt, answer))
    return answer

# Define the route for the index page
@app.get("/", response_class=HTMLResponse)
async def index():
    return """
    <!DOCTYPE html>
    <html>
        <head>
            <title>File Upload</title>
        </head>
        <body>
            <h2>Upload a PDF Document</h2>
            <form action="/process-document" method="post" enctype="multipart/form-data">
                <input type="file" name="file" required>
                <button type="submit">Upload</button>
            </form>
            <h2>Chat with the Bot</h2>
            <form id="chat-form">
                <input type="text" id="userMessage" placeholder="Type your message here..." required>
                <button type="submit">Send
                </button>
            </form>
            <div id="chat-response"></div>
            <script>
                document.getElementById("chat-form").onsubmit = async (e) => {
                    e.preventDefault();
                    const userMessage = document.getElementById("userMessage").value;
                    const response = await fetch("/process-message", {
                        method: "POST",
                        headers: {
                            "Content-Type": "application/json",
                        },
                        body: JSON.stringify({ userMessage }),
                    });
                    const data = await response.json();
                    document.getElementById("chat-response").innerText = data.botResponse || data.error;
                    document.getElementById("userMessage").value = ""; // Clear input
                };
            </script>
        </body>
    </html>
    """

# Define the route for processing messages
@app.post("/process-message")
async def process_message_route(message: MessageRequest):
    try:
        user_message = message.userMessage  # Extract the user's message from the request
        if not user_message:
            raise HTTPException(status_code=400, detail="User  message is required.")

        bot_response = process_prompt(user_message)  # Process the user's message
        bot_response = bot_response.split("<|fim_suffix|>")[0].strip()
          # Remove everything after <|fim_suffix|> and trim
        bot_response = bot_response.split("\n")[0].strip()


        # Return the bot's response as JSON
        return JSONResponse(content={"botResponse": bot_response})
    except Exception as e:
        logger.error(f"Error processing message: {e}")
        raise HTTPException(status_code=500, detail="An error occurred while processing the message.")

# Define the route for processing documents
@app.post("/process-document")
async def process_document_route(file: UploadFile = File(...)):
    try:
        # Check if a file was uploaded
        if not file:
            raise HTTPException(status_code=400, detail="File not uploaded.")

        file_path = f"uploads/{file.filename}"  # Define the path where the file will be saved
        os.makedirs("uploads", exist_ok=True)  # Create the uploads directory if it doesn't exist
        with open(file_path, "wb") as buffer:
            shutil.copyfileobj(file.file, buffer)  # Save the file

        process_document(file_path)  # Process the document

        # Return a success message as JSON
        return JSONResponse(content={
            "botResponse": "Thank you for providing your PDF document. I have analyzed it, so now you can ask me any questions regarding it!"
        })
    except Exception as e:
        logger.error(f"Error processing document: {e}")
        raise HTTPException(status_code=500, detail="An error occurred while processing the document.")