app.py

# app.py
pip install -r requirements.txt
import gradio as gr
import pandas as pd
import numpy as np
import yfinance as yf
from ta import add_all_ta_features
from ta.utils import dropna
from transformers import pipeline, AutoModelForQuestionAnswering, AutoTokenizer
from sentence_transformers import SentenceTransformer
import faiss

# Load the dataset from Google Drive
def load_data():
    url = "https://drive.google.com/uc?export=download&id=1N1bCVRSacs7_nENJzleqqNRHA22-H9B5"
    df = pd.read_csv(url)
    return df

# Preprocess the data
def preprocess_data(df):
    df = dropna(df)
    df = add_all_ta_features(df, open="Open", high="High", low="Low", close="Close", volume="Volume", fillna=True)
    return df

# Train the FAISS index for RAG
def train_faiss_index(df):
    model = SentenceTransformer('all-MiniLM-L6-v2')
    embeddings = model.encode(df['Close'].astype(str).tolist())
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatL2(dimension)
    index.add(embeddings)
    return index, model

# Load the QA model
def load_qa_model():
    model_name = "deepset/roberta-base-squad2"
    model = AutoModelForQuestionAnswering.from_pretrained(model_name)
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    qa_pipeline = pipeline("question-answering", model=model, tokenizer=tokenizer)
    return qa_pipeline

# Get technical analysis
def get_technical_analysis(df):
    analysis = {
        "SMA_50": df['Close'].rolling(window=50).mean().iloc[-1],
        "SMA_200": df['Close'].rolling(window=200).mean().iloc[-1],
        "RSI": df['momentum_rsi'].iloc[-1],
        "MACD": df['trend_macd'].iloc[-1],
    }
    return analysis

# RAG-based QA function
def rag_qa(question, df, index, model, qa_pipeline):
    query_embedding = model.encode([question])
    distances, indices = index.search(query_embedding, k=1)
    context = df.iloc[indices[0][0]]['Close']
    result = qa_pipeline(question=question, context=str(context))
    return result['answer']

# Gradio Interface
def trading_analysis_app(question):
    df = load_data()
    df = preprocess_data(df)
    index, model = train_faiss_index(df)
    qa_pipeline = load_qa_model()
    
    analysis = get_technical_analysis(df)
    answer = rag_qa(question, df, index, model, qa_pipeline)
    
    return f"Technical Analysis: {analysis}\n\nAnswer: {answer}"

# Gradio Interface
iface = gr.Interface(
    fn=trading_analysis_app,
    inputs="text",
    outputs="text",
    title="RAG-based Trading Analysis",
    description="Enter your question about ICICIBANK's stock to get technical analysis and answers."
)

iface.launch()