Update app.py

app.py

@@ -12,78 +12,28 @@ from llama_index.core.llms import ChatMessage
 from llama_index.core.chat_engine.condense_plus_context import CondensePlusContextChatEngine
 from llama_index.core.schema import Document
 
-# ===================================
-# 1️⃣ Fungsi Membaca Data PKB.json
-# ===================================
-def read_pkb_json():
-    try:
-        with open("pkb.json", "r", encoding="utf-8") as file:
-            data = json.load(file)
-        
-        pkb_text = "=== Perjanjian Kerja Bersama ===\n"
-        for bab, content in data["perjanjian_kerja_bersama"].items():
-            pkb_text += f"\n## {content['judul']} ##\n"
-            for pasal, pasal_data in content.items():
-                if pasal != "judul":
-                    pkb_text += f"\n### {pasal_data['judul']} ###\n"
-                    for item in pasal_data["isi"]:
-                        if isinstance(item, dict):
-                            pkb_text += f"- {item['istilah']}: {item['definisi']}\n"
-                        else:
-                            pkb_text += f"- {item}\n"
-        return pkb_text
-    except Exception as e:
-        return f"❌ ERROR membaca PKB.json: {str(e)}"
-
-# ===================================
-# 2️⃣ Fungsi Membaca Data Google Spreadsheet
-# ===================================
 def read_google_sheets():
     try:
-        # Tentukan scope akses ke Google Sheets & Drive
         scope = ["https://www.googleapis.com/auth/spreadsheets", "https://www.googleapis.com/auth/drive"]
-        
-        # Load kredensial dari file credentials.json
         creds = ServiceAccountCredentials.from_json_keyfile_name("credentials.json", scope)
         client = gspread.authorize(creds)
-        
-        # ID Spreadsheet (tetap sama untuk semua sheet)
         SPREADSHEET_ID = "1e_cNMhwF-QYpyYUpqQh-XCw-OdhWS6EuYsoBUsVtdNg"
-
-        # 📌 Daftar nama worksheet yang akan dibaca
         sheet_names = ["datatarget", "datacuti", "dataabsen", "datalembur", "pkb"]
-
-        all_data = []  # 🔹 List untuk menyimpan semua data
-
-        # 📌 Loop untuk membaca setiap worksheet
+        all_data = []
         spreadsheet = client.open_by_key(SPREADSHEET_ID)
         for sheet_name in sheet_names:
             try:
                 sheet = spreadsheet.worksheet(sheet_name)
                 data = sheet.get_all_values()
-                
-                # Tambahkan header nama sheet sebelum data untuk membedakan
-                all_data.append(f"=== Data dari {sheet_name.upper()} ===")
                 all_data.extend([" | ".join(row) for row in data])
-                all_data.append("\n")  # Pisahkan tiap sheet dengan newline
-                
             except gspread.exceptions.WorksheetNotFound:
-                all_data.append(f"❌ ERROR: Worksheet {sheet_name} tidak ditemukan.")
-
-        # Gabungkan semua data menjadi satu string panjang
-        formatted_text = "\n".join(all_data)
-        
-        return formatted_text
-    
+                all_data.append(f"ERROR: Worksheet {sheet_name} tidak ditemukan.")
+        return "\n".join(all_data).strip()
     except gspread.exceptions.SpreadsheetNotFound:
-        return "❌ ERROR: Spreadsheet tidak ditemukan. Pastikan ID/nama benar!"
-    
+        return "ERROR: Spreadsheet tidak ditemukan."
     except Exception as e:
-        return f"❌ ERROR: {str(e)}"
+        return f"ERROR: {str(e)}"
 
-# ===================================
-# 2️⃣ Fungsi untuk Mengunduh Model Llama
-# ===================================
 def initialize_llama_model():
     model_path = hf_hub_download(
         repo_id="TheBLoke/zephyr-7b-beta-GGUF",
@@ -92,102 +42,65 @@ def initialize_llama_model():
     )
     return model_path
 
-# ===================================
-# 3️⃣ Inisialisasi Model dan Pengaturan
-# ===================================
 def initialize_settings(model_path):
     Settings.llm = LlamaCPP(
         model_path=model_path,
         temperature=0.7,
     )
 
-# ===================================
-# 4️⃣ Inisialisasi Index dari Data Spreadsheet
-# ===================================
 def initialize_index():
-    text_data = read_google_sheets() + "\n" + read_pkb_json()
+    text_data = read_google_sheets()
     document = Document(text=text_data)
-    documents = [document]
-    
     parser = SentenceSplitter(chunk_size=150, chunk_overlap=10)
-    nodes = parser.get_nodes_from_documents(documents)
-    
-    embedding = HuggingFaceEmbedding("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
-    Settings.embed_model = embedding
-    
-    index = VectorStoreIndex(nodes)
-    return index
+    nodes = parser.get_nodes_from_documents([document])
+    Settings.embed_model = HuggingFaceEmbedding("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
+    return VectorStoreIndex(nodes)
 
-# ===================================
-# 5️⃣ Inisialisasi Mesin Chatbot
-# ===================================
 def initialize_chat_engine(index):
     retriever = index.as_retriever(similarity_top_k=3)
-    chat_engine = CondensePlusContextChatEngine.from_defaults(
+    return CondensePlusContextChatEngine.from_defaults(
         retriever=retriever,
-        verbose=True,
+        verbose=False  # Matikan verbose agar output lebih bersih
     )
-    return chat_engine
 
-# ===================================
-# 6️⃣ Fungsi untuk Menghasilkan Respons Chatbot
-# ===================================
 def generate_response(message, history, chat_engine):
     if history is None:
         history = []
-    
-    # 🔹 Baca ulang data dari Google Sheets & PKB
-    text_data = read_google_sheets() + "\n" + read_pkb_json()
-    
-    document = Document(text=text_data)
-    documents = [document]
-    
-    parser = SentenceSplitter(chunk_size=150, chunk_overlap=10)
-    nodes = parser.get_nodes_from_documents(documents)
-    index = VectorStoreIndex(nodes)
-    retriever = index.as_retriever(similarity_top_k=3)
-    
-    chat_engine = CondensePlusContextChatEngine.from_defaults(
-        retriever=retriever,
-        verbose=True,
-    )
-    
+
+    # Prompt untuk memastikan jawaban tetap dalam Bahasa Indonesia
     chat_messages = [
         ChatMessage(
             role="system",
             content=(
                 "Anda adalah chatbot yang dirancang khusus untuk berbicara dalam Bahasa Indonesia. "
-                "Anda tidak diperbolehkan menjawab dalam bahasa lain, termasuk Inggris. "
+                "Anda tidak diperbolehkan menjawab dalam bahasa lain. "
                 "Gunakan gaya bahasa profesional tetapi tetap ramah. "
                 "Jika informasi tidak tersedia dalam dokumen, katakan dengan sopan bahwa Anda tidak tahu. "
                 "Pastikan setiap jawaban diberikan secara ringkas, jelas, dan sesuai konteks."
             ),
         ),
+        ChatMessage(role="user", content=message),
     ]
-    
-    response = chat_engine.stream_chat(message)
-    text = "".join(response.response_gen)
-    
-    history.append((message, text))
-    return history
 
-# ===================================
-# 7️⃣ Fungsi Utama untuk Menjalankan Aplikasi
-# ===================================
+    response = chat_engine.chat(chat_messages)  # Menggunakan prompt secara eksplisit
+    text = response.response  # Ambil teks respons dari model
+
+    history.append((message, text))  # Menyimpan riwayat percakapan dengan format yang benar
+    return text  # Mengembalikan teks respons langsung tanpa simbol aneh
+
 def main():
     model_path = initialize_llama_model()
-    initialize_settings(model_path)  
-
+    initialize_settings(model_path)
     index = initialize_index()
-    chat_engine = initialize_chat_engine(index)  
+    chat_engine = initialize_chat_engine(index)
 
     def chatbot_response(message, history):
-        return generate_response(message, history, chat_engine)  
+        return generate_response(message, chat_engine)
 
     gr.Interface(
         fn=chatbot_response,
-        inputs=["text"],
-        outputs=["text"],
+        inputs="text",
+        outputs="text",
     ).launch()
 
 if __name__ == "__main__":