Update app.py

app.py

@@ -1,53 +1,51 @@
 import os
-import pinecone
-from transformers import T5Tokenizer, T5ForConditionalGeneration
-from sentence_transformers import SentenceTransformer
-import torch
 import gradio as gr
+from pinecone import Pinecone, ServerlessSpec
+from sentence_transformers import SentenceTransformer
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-# --- Load environment variables ---
+# بارگیری توکن‌ها از محیط
 HF_TOKEN = os.environ.get("HF_TOKEN")
 PINECONE_API_KEY = os.environ.get("PINECONE_API_KEY")
 PINECONE_INDEX_NAME = os.environ.get("PINECONE_INDEX_NAME")
 
-assert HF_TOKEN is not None, "❌ HF_TOKEN is missing!"
-assert PINECONE_API_KEY is not None, "❌ Pinecone API key is missing!"
+# بررسی متغیرها
+assert HF_TOKEN is not None, "❌ Hugging Face token missing!"
+assert PINECONE_API_KEY is not None, "❌ Pinecone API key missing!"
 assert PINECONE_INDEX_NAME is not None, "❌ Pinecone index name is missing!"
 
-# --- Initialize Pinecone ---
-pinecone.init(api_key=PINECONE_API_KEY, environment="gcp-starter")
-index = pinecone.Index(PINECONE_INDEX_NAME)
-
-# --- Load models ---
+# بارگذاری مدل‌های مورد نیاز
 embedding_model = SentenceTransformer("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2", token=HF_TOKEN)
-tokenizer = T5Tokenizer.from_pretrained("google/mt5-small", token=HF_TOKEN)
-model = T5ForConditionalGeneration.from_pretrained("google/mt5-small", token=HF_TOKEN)
-
-def generate_answer(question):
-    # Embed the question
-    question_embedding = embedding_model.encode(question).tolist()
-
-    # Query Pinecone for similar content
-    response = index.query(vector=question_embedding, top_k=3, include_metadata=True)
-    contexts = [match['metadata']['text'] for match in response['matches'] if 'text' in match['metadata']]
+tokenizer = AutoTokenizer.from_pretrained("google/mt5-small", token=HF_TOKEN)
+language_model = AutoModelForSeq2SeqLM.from_pretrained("google/mt5-small", token=HF_TOKEN)
 
-    # Concatenate context
-    context = "\n".join(contexts)
-    input_text = f"پرسش: {question}\nاطلاعات: {context}\nپاسخ:"
+# اتصال به Pinecone (نسخه جدید)
+pc = Pinecone(api_key=PINECONE_API_KEY)
+index = pc.Index(PINECONE_INDEX_NAME)
 
-    # Tokenize and generate
-    inputs = tokenizer(input_text, return_tensors="pt", truncation=True, max_length=512)
-    outputs = model.generate(**inputs, max_new_tokens=200)
-    answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return answer
+def get_similar_context(question, top_k=1):
+    question_embedding = embedding_model.encode(question).tolist()
+    results = index.query(vector=question_embedding, top_k=top_k, include_metadata=True)
+    if results and results['matches']:
+        return results['matches'][0]['metadata']['text']
+    return ""
 
-# --- Gradio UI ---
-iface = gr.Interface(
+def generate_answer(question):
+    context = get_similar_context(question)
+    if context:
+        prompt = f"پرسش: {question}\nاطلاعات مرتبط: {context}\nپاسخ:"
+    else:
+        prompt = f"پرسش: {question}\nپاسخ:"
+
+    inputs = tokenizer(prompt, return_tensors="pt", padding=True, truncation=True)
+    outputs = language_model.generate(**inputs, max_new_tokens=100)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# رابط گرافیکی
+gr.Interface(
     fn=generate_answer,
-    inputs=gr.Textbox(label="question"),
-    outputs=gr.Textbox(label="output"),
-    title="💬 چت‌بات هوشمند تیام",
-    description="سؤالات خود درباره خدمات دیجیتال مارکتینگ تیام را بپرسید."
-)
-
-iface.launch()
+    inputs=gr.Textbox(label="سؤال خود را وارد کنید", placeholder="مثلاً خدمات سئو تیام شامل چیست؟"),
+    outputs=gr.Textbox(label="پاسخ هوشمند"),
+    title="🤖 چت‌بات هوشمند تیام",
+    description="با استفاده از داده‌های اختصاصی، پاسخ هوشمندانه دریافت کنید."
+).launch()