__import__('pysqlite3') import sys sys.modules['sqlite3'] = sys.modules.pop('pysqlite3') import streamlit as st import os import io import asyncio import json import hashlib import shutil import time import requests import edge_tts from openai import OpenAI from langchain_community.document_loaders import PyPDFLoader, Docx2txtLoader, TextLoader from langchain_community.vectorstores import Chroma from langchain_huggingface import HuggingFaceEmbeddings from langchain_text_splitters import RecursiveCharacterTextSplitter from streamlit_lottie import st_lottie # --- 1. KONFIGURASI HALAMAN --- st.set_page_config( page_title="HarshBot - Sobat Digi Galak", page_icon="😤", layout="wide" ) # --- 2. CSS CUSTOM (TAMPILAN BERSIH & GALAK) --- st.markdown(""" """, unsafe_allow_html=True) # --- 3. FUNGSI UTILITAS (AUDIO & ANIMASI) --- def load_lottieurl(url: str): """Load animasi Lottie dari URL""" try: r = requests.get(url) if r.status_code != 200: return None return r.json() except: return None async def _generate_audio_async(text, voice="id-ID-ArdiNeural"): """Generate audio TTS (Async)""" communicate = edge_tts.Communicate(text, voice) mp3_fp = io.BytesIO() async for chunk in communicate.stream(): if chunk["type"] == "audio": mp3_fp.write(chunk["data"]) mp3_fp.seek(0) return mp3_fp def generate_audio_sync(text): """Wrapper Sync untuk TTS""" try: loop = asyncio.get_event_loop() except RuntimeError: loop = asyncio.new_event_loop() asyncio.set_event_loop(loop) if loop.is_running(): future = asyncio.run_coroutine_threadsafe(_generate_audio_async(text), loop) return future.result() else: return loop.run_until_complete(_generate_audio_async(text)) # --- 4. ENGINE: DATABASE & LLM --- api_key = os.environ.get("GROQ_API_KEY") or st.secrets.get("GROQ_API_KEY") if not api_key: st.error("⚠️ API Key Error: GROQ_API_KEY tidak ditemukan.") st.stop() client = OpenAI(api_key=api_key, base_url="https://api.groq.com/openai/v1") # Konfigurasi Path Database DB_PATH = "./chroma_db_store" METADATA_PATH = "db_metadata.json" def get_files_signature(folder_path="."): """Cek apakah ada perubahan file dokumen""" file_signatures = [] for root, dirs, files in os.walk(folder_path): for file in files: if "chroma_db" in root or file.startswith("."): continue if file.lower().endswith(('.pdf', '.docx', '.txt')) and "requirements" not in file: full_path = os.path.join(root, file) mtime = os.path.getmtime(full_path) file_signatures.append(f"{file}_{mtime}") return hashlib.md5("".join(sorted(file_signatures)).encode()).hexdigest() @st.cache_resource def get_vectorstore(): """Inisialisasi Database Vektor dengan HuggingFace Embeddings""" embedding_function = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2") current_signature = get_files_signature(".") last_signature = "" if os.path.exists(METADATA_PATH): with open(METADATA_PATH, "r") as f: try: last_signature = json.load(f).get("signature", "") except: pass db_exists = os.path.exists(DB_PATH) and len(os.listdir(DB_PATH)) > 0 is_changed = current_signature != last_signature if db_exists and not is_changed: print("✅ Menggunakan Database Lama.") return Chroma(persist_directory=DB_PATH, embedding_function=embedding_function) print("🔄 Membangun Ulang Database...") if os.path.exists(DB_PATH): shutil.rmtree(DB_PATH, ignore_errors=True) documents = [] for root, dirs, files in os.walk("."): if "chroma_db" in root: continue for file in files: full_path = os.path.join(root, file) try: if file.lower().endswith(".pdf"): documents.extend(PyPDFLoader(full_path).load()) elif file.lower().endswith(".docx"): documents.extend(Docx2txtLoader(full_path).load()) elif file.lower().endswith(".txt") and "requirements" not in file: documents.extend(TextLoader(full_path, encoding='utf-8').load()) except: pass if not documents: return None text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200) splits = text_splitter.split_documents(documents) db = Chroma.from_documents(documents=splits, embedding=embedding_function, persist_directory=DB_PATH) with open(METADATA_PATH, "w") as f: json.dump({"signature": current_signature}, f) return db # Load Database db = get_vectorstore() # --- 5. SIDEBAR: AVATAR & TOMBOL RESET --- with st.sidebar: st.markdown("### 🤖 Identity") # Animasi Robot (Lottie) # Ganti URL ini jika ingin animasi lain lottie_url = "https://lottie.host/5b064c5d-4f7f-4c12-9c16-2d3345701c40/3eJ5y2WpQY.json" avatar_anim = load_lottieurl(lottie_url) if avatar_anim: st_lottie(avatar_anim, height=180, key="avatar_main") else: st.write("🤖 [Avatar Error]") st.markdown("""

HarshBot

Jangan tanya hal bodoh.

""", unsafe_allow_html=True) st.markdown("---") # TOMBOL RESET DATABASE (Permintaan Anda) if st.button("🗑️ Hapus & Reset Database", type="primary"): with st.spinner("Menghapus ingatan..."): try: if os.path.exists(DB_PATH): shutil.rmtree(DB_PATH) if os.path.exists(METADATA_PATH): os.remove(METADATA_PATH) st.success("Database dihapus!") time.sleep(1) st.rerun() except Exception as e: st.error(f"Gagal: {e}") # --- 6. CHAT INTERFACE --- if "messages" not in st.session_state: st.session_state.messages = [] # Tampilkan History for message in st.session_state.messages: with st.chat_message(message["role"]): st.markdown(message["content"]) if message["role"] == "assistant" and "audio" in message: st.audio(message["audio"], format="audio/mp3") # Input User if user_input := st.chat_input("Tulis pertanyaan lo di sini..."): st.session_state.messages.append({"role": "user", "content": user_input}) with st.chat_message("user"): st.markdown(user_input) # Cari Konteks (RAG) konteks = "" if db: try: docs = db.similarity_search(user_input, k=3) konteks = "\n\n".join([d.page_content for d in docs]) except: pass # Prompt Galak system_prompt = f"""You are a very harsh, sarcastic, and nonchalant chatbot. CONTEXT DATA: {konteks} INSTRUCTION: 1. Act nonchalant and slightly annoyed. 2. Use non-formal Indonesian language (bahasa gaul/kasar sopan). 3. Provide the answer based on CONTEXT DATA if available, but do it grudgingly. 4. If no context matches, mock the user for asking something irrelevant. """ with st.chat_message("assistant"): try: # Generate Text history_payload = [{"role": m["role"], "content": m["content"]} for m in st.session_state.messages[-5:]] messages_final = [{"role": "system", "content": system_prompt}] + history_payload stream = client.chat.completions.create( model="llama-3.3-70b-versatile", messages=messages_final, temperature=0.7, stream=True ) response_placeholder = st.empty() full_response = "" for chunk in stream: content = chunk.choices[0].delta.content if content: full_response += content response_placeholder.markdown(full_response + "▌") response_placeholder.markdown(full_response) # Generate Audio audio_buffer = generate_audio_sync(full_response) st.audio(audio_buffer, format="audio/mp3") st.session_state.messages.append({ "role": "assistant", "content": full_response, "audio": audio_buffer }) except Exception as e: st.error(f"Error: {e}")