app.py

import time
import streamlit as st
from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings.base import Embeddings
from langchain.vectorstores import FAISS
from langchain.indexes import VectorstoreIndexCreator
from langchain.chains import RetrievalQA
from langchain.chat_models import ChatOpenAI
from typing import List
from together import Together
import pandas as pd
import streamlit as st
from sklearn.metrics.pairwise import cosine_similarity


import numpy as np
from sentence_transformers import SentenceTransformer
import faiss


# ----------------- تنظیمات صفحه -----------------
st.set_page_config(page_title="رزم یار ارتش", page_icon="🪖", layout="wide")

# ----------------- استایل سفارشی -----------------
st.markdown("""
    <style>
    @import url('https://fonts.googleapis.com/css2?family=Vazirmatn:wght@400;700&display=swap');
    html, body, [class*="css"] {
        font-family: 'Vazirmatn', Tahoma, sans-serif;
        direction: rtl;
        text-align: right;
    }
    .stApp {
        background: linear-gradient(to left, #f0f4f7, #d9e2ec);
    }
    .sidebar .sidebar-content {
        background-color: #ffffff;
        border-left: 2px solid #4e8a3e;
        padding-top: 10px;
    }
    .sidebar .sidebar-content div {
        margin-bottom: 10px;
        font-weight: bold;
        color: #2c3e50;
        font-size: 15px;
    }
    .stButton>button {
        background-color: #4e8a3e !important;
        color: white !important;
        font-weight: bold;
        border-radius: 8px;
        padding: 5px 16px;
        transition: 0.3s;
        font-size: 14px;
    }
    .stButton>button:hover {
        background-color: #3c6d30 !important;
    }
    .header-text {
        text-align: center;
        margin-top: 15px;
        margin-bottom: 25px;
        background-color: rgba(255, 255, 255, 0.85);
        padding: 16px;
        border-radius: 16px;
        box-shadow: 0 4px 10px rgba(0,0,0,0.1);
    }
    .header-text h1 {
        font-size: 36px;
        color: #2c3e50;
        margin: 0;
        font-weight: bold;
    }
    .subtitle {
        font-size: 16px;
        color: #34495e;
        margin-top: 5px;
    }
    .chat-message {
        background-color: rgba(255, 255, 255, 0.95);
        border: 1px solid #4e8a3e;
        border-radius: 12px;
        padding: 14px;
        margin-bottom: 10px;
        box-shadow: 0 4px 8px rgba(0,0,0,0.08);
        animation: fadeIn 0.5s ease;
    }
    .stTextInput>div>input, .stTextArea textarea {
        background-color: rgba(255,255,255,0.9) !important;
        border-radius: 8px !important;
        direction: rtl;
        text-align: right;
        font-family: 'Vazirmatn', Tahoma;
    }
    img.small-logo {
        width: 90px;
        margin-bottom: 15px;
        display: block;
        margin-right: auto;
        margin-left: auto;
    }
    .menu-item {
        display: flex;
        align-items: center;
        gap: 8px;
        padding: 6px 0;
        font-size: 15px;
        cursor: pointer;
    }
    .menu-item img {
        width: 20px;
        height: 20px;
    }
    </style>
""", unsafe_allow_html=True)

# ----------------- بدنه اصلی -----------------
with st.sidebar:
    st.image("log.png", width=90)

    
    st.markdown("""
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/3596/3596165.png" />
        گفتگوی جدید
    </div>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/709/709496.png" />
        تاریخچه
    </div>
    <hr/>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/1828/1828932.png" />
        مدل‌های هوش مصنوعی
    </div>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/681/681494.png" />
        تولید محتوا
    </div>
    <hr/>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/3601/3601646.png" />
        دستیار ویژه
    </div>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/709/709612.png" />
        ابزار مالی
    </div>
    <hr/>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/2099/2099058.png" />
        تنظیمات
    </div>
    <div class="menu-item">
        <img src="https://cdn-icons-png.flaticon.com/512/597/597177.png" />
        پشتیبانی
    </div>
    """, unsafe_allow_html=True)
st.markdown("""
    <style>
    /* تنظیم سایز سایدبار */
    [data-testid="stSidebar"] {
        width: 220px !important;
        flex-shrink: 0;
    }
    [data-testid="stSidebar"] > div {
        width: 220px !important;
    }
    </style>
""", unsafe_allow_html=True)

# محتوای اصلی
st.markdown("""
    <div class="header-text">
        <h1>رزم یار ارتش</h1>
        <div class="subtitle">دستیار هوشمند ارتشی برای پشتیبانی و راهبری</div>
    </div>
""", unsafe_allow_html=True)

st.markdown('<div class="chat-message">👋 سلام! چطور میتونم کمکتون کنم؟</div>', unsafe_allow_html=True)

# چت اینپوت کاربر
#user_input = st.text_input("پیام خود را وارد کنید...")

#if user_input:
#   st.markdown(f'<div class="chat-message">📩 شما: {user_input}</div>', unsafe_allow_html=True)




# ⚙️ مدل Embedding ساده و سریع
@st.cache_resource
def get_embedding_model():
    return SentenceTransformer("HooshvareLab/bert-fa-zwnj-base")

@st.cache_resource
def process_csv(csv_file):
    df = pd.read_csv(csv_file)
    texts = df.iloc[:, 0].astype(str).tolist()
    texts = [text for text in texts if text.strip()]

    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=200,
        chunk_overlap=50,
        length_function=len,
        separators=["\n\n", "\n", " ", ""]
    )

    split_texts = []
    for text in texts:
        split_texts.extend(text_splitter.split_text(text))

    # مدل امبدینگ
    model = get_embedding_model()
    embeddings = model.encode(split_texts, show_progress_bar=True)

    dim = embeddings.shape[1]
    index = faiss.IndexHNSWFlat(dim, 32)  
    index.hnsw.efSearch = 50 
    index.add(np.array(embeddings))

    return split_texts, embeddings, index

# مسیر فایل CSV
csv_file_path = 'output (1).csv'

texts, vectors, index = process_csv(csv_file_path)

# رابط چت
if 'messages' not in st.session_state:
    st.session_state.messages = []

if 'pending_prompt' not in st.session_state:
    st.session_state.pending_prompt = None

for msg in st.session_state.messages:
    with st.chat_message(msg['role']):
        st.markdown(msg['content'], unsafe_allow_html=True)

query = st.chat_input("سؤالت را بپرس...")

if query:
    st.session_state.messages.append({'role': 'user', 'content': query})
    st.session_state.pending_prompt = query
    st.rerun()

if st.session_state.pending_prompt:
    with st.chat_message("ai"):
        thinking = st.empty()
        thinking.markdown("🤖 در حال جستجو...")

        model = get_embedding_model()
        query_vector = model.encode([st.session_state.pending_prompt])
        
        D, I = index.search(np.array(query_vector), k=10)
        
        top_indices = I[0]
        top_texts = [texts[i] for i in top_indices]
        top_vectors = np.array([vectors[i] for i in top_indices])

        
        similarities = cosine_similarity(query_vector, top_vectors)[0]
        
        # پیدا کردن دقیق‌ترین متن
        best_match_relative_index = np.argmax(similarities)
        best_match_index = top_indices[best_match_relative_index]
        best_match_text = texts[best_match_index]
        response = "🧠 پاسخ سوال :\n\n" .join(best_match_text)

        thinking.empty()
        full_response = ""
        placeholder = st.empty()
        for word in response.split():
            full_response += word + " "
            placeholder.markdown(full_response + "▌")
            time.sleep(0.02)

        placeholder.markdown(full_response)
        st.session_state.messages.append({'role': 'ai', 'content': full_response})
        st.session_state.pending_prompt = None