Update app.py

app.py

@@ -13,13 +13,37 @@ from llama_index.core.chat_engine.condense_plus_context import CondensePlusConte
 from llama_index.core.schema import Document
 
 # ===================================
-# 1️⃣ Fungsi untuk Membaca Google Spreadsheet dari Beberapa Worksheet
+# 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:
         scope = ["https://www.googleapis.com/auth/spreadsheets", "https://www.googleapis.com/auth/drive"]
         creds = ServiceAccountCredentials.from_json_keyfile_name("credentials.json", scope)
         client = gspread.authorize(creds)
+
         SPREADSHEET_ID = "1e_cNMhwF-QYpyYUpqQh-XCw-OdhWS6EuYsoBUsVtdNg"
         sheet_names = ["datatarget", "datacuti", "dataabsen", "datalembur"]
 
@@ -34,23 +58,13 @@ def read_google_sheets():
                 all_data.append("\n")
             except gspread.exceptions.WorksheetNotFound:
                 all_data.append(f"❌ ERROR: Worksheet {sheet_name} tidak ditemukan.")
-        return "\n".join(all_data)
-    except Exception as e:
-        return f"❌ ERROR: {str(e)}"
 
-# ===================================
-# 2️⃣ Fungsi untuk Membaca File PKB JSON
-# ===================================
-def read_pkb_json():
-    try:
-        with open("pkb.json", "r", encoding="utf-8") as file:
-            pkb_data = json.load(file)
-        return json.dumps(pkb_data, indent=2, ensure_ascii=False)
+        return "\n".join(all_data)
     except Exception as e:
         return f"❌ ERROR: {str(e)}"
 
 # ===================================
-# 3️⃣ Fungsi untuk Mengunduh Model Llama
+# 3️⃣ Inisialisasi Model Llama
 # ===================================
 def initialize_llama_model():
     model_path = hf_hub_download(
@@ -61,33 +75,20 @@ def initialize_llama_model():
     return model_path
 
 # ===================================
-# 4️⃣ Inisialisasi Model dan Pengaturan
-# ===================================
-def initialize_settings(model_path):
-    Settings.llm = LlamaCPP(
-        model_path=model_path,
-        temperature=0.7,
-    )
-
-# ===================================
-# 5️⃣ Inisialisasi Index dari Data Spreadsheet & PKB JSON
+# 4️⃣ Inisialisasi Index Data PKB & Spreadsheet
 # ===================================
 def initialize_index():
-    text_data = read_google_sheets()
-    pkb_data = read_pkb_json()
-    combined_text = text_data + "\n\n" + pkb_data
-
-    document = Document(text=combined_text)
-    documents = [document]
+    text_data = read_google_sheets() + "\n" + read_pkb_json()
+    document = Document(text=text_data)
     parser = SentenceSplitter(chunk_size=150, chunk_overlap=10)
-    nodes = parser.get_nodes_from_documents(documents)
+    nodes = parser.get_nodes_from_documents([document])
     embedding = HuggingFaceEmbedding("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
     Settings.embed_model = embedding
     index = VectorStoreIndex(nodes)
     return index
 
 # ===================================
-# 6️⃣ Inisialisasi Mesin Chatbot
+# 5️⃣ Inisialisasi Chatbot
 # ===================================
 def initialize_chat_engine(index):
     retriever = index.as_retriever(similarity_top_k=3)
@@ -98,22 +99,43 @@ def initialize_chat_engine(index):
     return chat_engine
 
 # ===================================
-# 7️⃣ Fungsi untuk Menghasilkan Respons Chatbot
+# 6️⃣ Fungsi untuk Respons Chatbot
 # ===================================
 def generate_response(message, history, chat_engine):
     if history is None:
         history = []
+    
+    chat_messages = [
+        ChatMessage(
+            role="system",
+            content=(
+                "Anda adalah chatbot yang dirancang khusus untuk berbicara dalam Bahasa Indonesia. "
+                "Anda adalah chatbot HRD yang membantu karyawan dalam memahami administrasi dan data perusahaan. "
+                "Anda tidak diperbolehkan menjawab dalam bahasa lain, termasuk Inggris. "
+                "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."
+                "Jawaban harus singkat, jelas, dan dalam Bahasa Indonesia."
+            ),
+        ),
+    ]
+    
     response = chat_engine.stream_chat(message)
     text = "".join(response.response_gen)
+    
     history.append((message, text))
     return history
 
 # ===================================
-# 8️⃣ Fungsi Utama untuk Menjalankan Aplikasi
+# 7️⃣ Fungsi Utama untuk Menjalankan Aplikasi
 # ===================================
 def main():
     model_path = initialize_llama_model()
-    initialize_settings(model_path)
+    Settings.llm = LlamaCPP(
+        model_path=model_path,
+        temperature=0.7,
+    )
+    
     index = initialize_index()
     chat_engine = initialize_chat_engine(index)