Update app.py

app.py

@@ -9,18 +9,10 @@ from langchain.chains import RetrievalQA
 from langchain.chat_models import ChatOpenAI
 from typing import List
 from together import Together
-from transformers import AutoTokenizer, AutoModelForCausalLM
-from transformers import AutoTokenizer, AutoModel
-import torch
-from langchain_community.embeddings import HuggingFaceInstructEmbeddings
 
 
-import streamlit as st
-from PIL import Image
-
-st.set_page_config(page_title="چت‌ بات توانا", page_icon="🪖", layout="wide")
+st.set_page_config(page_title="چت‌ بات ارتش", page_icon="🪖", layout="wide")
 
-# استایل
 st.markdown("""
     <style>
     @import url('https://fonts.googleapis.com/css2?family=Vazirmatn:wght@400;700&display=swap');
@@ -30,7 +22,7 @@ st.markdown("""
         text-align: right;
     }
     .stApp {
-        background: url("military_bg.jpeg") no-repeat center center fixed;
+        background: url("./military_bg.jpeg") no-repeat center center fixed;
         background-size: cover;
         backdrop-filter: blur(2px);
     }
@@ -88,49 +80,47 @@ st.markdown("""
     </style>
 """, unsafe_allow_html=True)
 
-# لوگو در وسط با columns
 col1, col2, col3 = st.columns([1, 1, 1])
 with col2:
-    try:
-        image = Image.open("army.png")
-        st.image(image, width=240)
-    except FileNotFoundError:
-        st.error("📁 فایل 'army.png' پیدا نشد. مطمئن شو کنار فایل اصلی Streamlit هست.")
+    st.image("army.png", width=240)
 
-# تیتر
 st.markdown("""
     <div class="header-text">
-        <h1>چت‌ بات توانا</h1>
+        <h1>چت‌ بات ارتش</h1>
         <div class="subtitle">دستیار هوشمند برای تصمیم‌گیری در میدان نبرد</div>
     </div>
 """, unsafe_allow_html=True)
 
 
-from transformers import AutoTokenizer, AutoModel  
-
-class HuggingFaceEmbeddings(Embeddings):
-    def __init__(self, model_name: str):
-        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-        self.model = AutoModel.from_pretrained(model_name)
+class TogetherEmbeddings(Embeddings):
+    def __init__(self, model_name: str, api_key: str):
+        self.model_name = model_name
+        self.client = Together(api_key=api_key)
 
     def embed_documents(self, texts: List[str]) -> List[List[float]]:
-        embeddings = []
-        for text in texts:
-            inputs = self.tokenizer(text, return_tensors="pt", truncation=True, padding=True)
-            with torch.no_grad():
-                outputs = self.model(**inputs)
-            embeddings.append(outputs.last_hidden_state.mean(dim=1).squeeze().tolist())
-        return embeddings
+        response = self.client.embeddings.create(model=self.model_name, input=texts)
+        return [item.embedding for item in response.data]
 
     def embed_query(self, text: str) -> List[float]:
         return self.embed_documents([text])[0]
 
 @st.cache_resource
-def get_pdf_index(pdf_docs="test1.pdf"):
-  loader = PyPDFLoader('test1.pdf')
-  embeddings = HuggingFaceInstructEmbeddings(model_name="SajjadAyoubi/xlm-roberta-large-fa-qa")
-  index  = VectorstoreIndexCreator( embedding=embeddings, text_splitter=RecursiveCharacterTextSplitter(chunk_size=300, chunk_overlap=0)).from_loaders(pdf_reader)
-  return index
+def get_pdf_index():
+    with st.spinner('📄 در حال پردازش فایل PDF...'):
+        loader = [PyPDFLoader('test1.pdf')]
+        embeddings = TogetherEmbeddings(
+            model_name="togethercomputer/m2-bert-80M-8k-retrieval",
+            api_key="0291f33aee03412a47fa5d8e562e515182dcc5d9aac5a7fb5eefdd1759005979"
+
+
+
+
+        )
+        return VectorstoreIndexCreator(
+            embedding=embeddings,
+            text_splitter=RecursiveCharacterTextSplitter(chunk_size=300, chunk_overlap=0)
+        ).from_loaders(loader)
+
 index = get_pdf_index()
 
 llm = ChatOpenAI(
@@ -168,7 +158,7 @@ if st.session_state.pending_prompt:
         thinking = st.empty()
         thinking.markdown("🤖 در حال فکر کردن...")
 
-        response = chain.run(f'لطفاً فقط به زبان فارسی پاسخ بده: {st.session_state.pending_prompt}')
+        response = chain.run(f'question:پاسخ را فقط به زبان فارسی جواب بده {st.session_state.pending_prompt}')
         answer = response.split("Helpful Answer:")[-1].strip()
         if not answer:
             answer = "متأسفم، اطلاعات دقیقی در این مورد ندارم."