Clean deploy

app.py

@@ -0,0 +1,58 @@
+from langchain_community.llms import HuggingFaceHub
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain.chains import RetrievalQA
+import os
+import requests
+import gradio as gr
+
+# Configuration
+HF_TOKEN = os.environ.get("HF_TOKEN", "")
+CLOUDFLARE_API = "https://notary-662-sbz.pages.dev/api/db/chats"
+
+# 1. Setup Llama 3 via Hugging Face Inference API
+llm = HuggingFaceHub(
+    repo_id="meta-llama/Meta-Llama-3-8B-Instruct",
+    huggingfacehub_api_token=HF_TOKEN,
+    model_kwargs={"temperature": 0.7, "max_new_tokens": 512}
+)
+
+# 2. Setup Persian-capable Embeddings
+embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
+
+def query_rag_system(user_query, chat_id="default"):
+    # Load the 100 docs index (assuming it's saved locally in the Space)
+    try:
+        vector_store = FAISS.load_local("faiss_index", embeddings, allow_dangerous_deserialization=True)
+        qa_chain = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=vector_store.as_retriever())
+        
+        # Get response from Llama 3 + RAG
+        result = qa_chain.run(user_query)
+        
+        # 3. Sync with Cloudflare D1
+        chat_data = {
+            "id": chat_id,
+            "title": user_query[:30],
+            "docType": "notary_rag_llama",
+            "messages": [
+                {"role": "user", "text": user_query},
+                {"role": "model", "text": result}
+            ]
+        }
+        requests.post(CLOUDFLARE_API, json=chat_data)
+        
+        return result
+    except Exception as e:
+        return f"خطا در اتصال به بانک اسناد: {str(e)}"
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=query_rag_system,
+    inputs="text",
+    outputs="text",
+    title="Notary Llama-3 RAG Engine",
+    description="این سیستم مستقیماً به Cloudflare و ۱۰۰ فایل PDF محضر متصل است."
+)
+
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,13 @@
+fastapi
+uvicorn
+langchain
+langchain-community
+langchain-google-genai
+pypdf
+requests
+beautifulsoup4
+fpdf
+huggingface_hub
+google-generativeai
+faiss-cpu
+gradio

scraper.py

@@ -0,0 +1,55 @@
+import requests
+from bs4 import BeautifulSoup
+from fpdf import FPDF
+from huggingface_hub import HfApi, upload_file
+import os
+
+HF_TOKEN = os.environ.get("HF_TOKEN", "")
+DATASET_REPO = "sosa123454321/Notary-PDF-Dataset"
+
+def scrape_article_to_pdf(url, output_name):
+    print(f"Scraping {url}...")
+    try:
+        response = requests.get(url)
+        soup = BeautifulSoup(response.content, 'html.parser')
+        
+        # Extract main text
+        for script in soup(["script", "style"]):
+            script.decompose()
+        
+        text = soup.get_text()
+        lines = (line.strip() for line in text.splitlines())
+        chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
+        clean_text = '\n'.join(chunk for chunk in chunks if chunk)
+
+        # Create PDF
+        pdf = FPDF()
+        pdf.add_page()
+        pdf.set_font("Arial", size=10)
+        # Handle non-latin characters by replacing them for the demo
+        # For full Persian support, a .ttf font must be loaded using pdf.add_font()
+        pdf.multi_cell(0, 10, txt=clean_text.encode('latin-1', 'replace').decode('latin-1'))
+        
+        pdf_path = f"{output_name}.pdf"
+        pdf.output(pdf_path)
+        print(f"Saved to {pdf_path}")
+        
+        # Upload to HF
+        print(f"Uploading to {DATASET_REPO}...")
+        upload_file(
+            path_or_fileobj=pdf_path,
+            path_in_repo=f"documents/{pdf_path}",
+            repo_id=DATASET_REPO,
+            repo_type="dataset",
+            token=HF_TOKEN
+        )
+        print("Upload successful!")
+        return pdf_path
+    except Exception as e:
+        print(f"Error: {e}")
+        return None
+
+if __name__ == "__main__":
+    # Example usage
+    test_url = "https://www.notary662th.ir/induction-manual"
+    scrape_article_to_pdf(test_url, "notary_induction_persian")