Spaces:
Runtime error
Runtime error
| import os | |
| import warnings | |
| from pathlib import Path | |
| from fastapi import FastAPI, Request | |
| from fastapi.middleware.cors import CORSMiddleware | |
| from pydantic import BaseModel | |
| from operator import itemgetter | |
| from dotenv import load_dotenv | |
| from slowapi import Limiter, _rate_limit_exceeded_handler | |
| from slowapi.util import get_remote_address | |
| from slowapi.errors import RateLimitExceeded | |
| warnings.filterwarnings("ignore") | |
| load_dotenv() | |
| from langchain_community.embeddings import HuggingFaceEmbeddings | |
| from langchain_community.vectorstores import Chroma | |
| from langchain_community.llms import Ollama | |
| from langchain_core.prompts import PromptTemplate | |
| from langchain_core.output_parsers import StrOutputParser | |
| from langchain_openai import ChatOpenAI | |
| app = FastAPI() | |
| limiter = Limiter(key_func=get_remote_address) | |
| app.state.limiter = limiter | |
| app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler) | |
| # ==================== KONFIGURACJA CORS ==================== | |
| app.add_middleware( | |
| CORSMiddleware, | |
| allow_origins=["*"], | |
| allow_credentials=True, | |
| allow_methods=["*"], | |
| allow_headers=["*"], | |
| ) | |
| # ==================== KONFIGURACJA ==================== | |
| CHROMA_DB_PATH = "./chroma_db" | |
| EVIL_MODE = os.getenv("EVIL_MODE", "GROQ") | |
| # ==================== MODELE DANYCH ==================== | |
| class ChatRequest(BaseModel): | |
| message: str | |
| language: str = "pl" | |
| # ==================== GLOBALNE ZMIENNE ==================== | |
| rag_chain_pl = None | |
| rag_chain_multi = None | |
| # ==================== STARTUP EVENT ==================== | |
| async def startup_event(): | |
| global rag_chain_pl, rag_chain_multi | |
| print("\n" + "=" * 60) | |
| print("🚀 URUCHAMIANIE SERWERA EVILCHAT") | |
| print("=" * 60) | |
| try: | |
| # ========== KROK 1: Wczytanie bazy z dysku ========== | |
| print("\n1️⃣ Wczytywanie bazy Chroma z dysku...") | |
| if not Path(CHROMA_DB_PATH).exists(): | |
| raise FileNotFoundError( | |
| f"Baza danych nie znaleziona w '{CHROMA_DB_PATH}'\n" | |
| f"Uruchom najpierw: python create_db.py" | |
| ) | |
| embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2") | |
| vectorstore = Chroma( | |
| persist_directory=CHROMA_DB_PATH, | |
| embedding_function=embeddings | |
| ) | |
| retriever = vectorstore.as_retriever(search_kwargs={"k": 3}) | |
| print(" ✅ Baza załadowana pomyślnie") | |
| # ========== KROK 2: Inicjalizacja modelu LLM ========== | |
| print("\n2️⃣ Inicjalizuję model LLM (OpenRouter Uncensored)...") | |
| if EVIL_MODE == "LOCAL": | |
| print(" 📍 Tryb: OLLAMA (lokalny)") | |
| llm = Ollama(model="jayeshpandit2480/gemma3-UNCENSORED:1b", num_ctx=2048) | |
| else: | |
| print(" ☁️ Tryb: OpenRouter") | |
| OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY") | |
| if not OPENROUTER_API_KEY: | |
| raise ValueError("Brak klucza OPENROUTER_API_KEY w .env!") | |
| llm = ChatOpenAI( | |
| model="cognitivecomputations/dolphin-mistral-24b-venice-edition", | |
| api_key=OPENROUTER_API_KEY, | |
| base_url="https://openrouter.ai/api/v1", | |
| temperature=0.7, | |
| ) | |
| print(" ✅ Model LLM gotowy") | |
| # ========== KROK 3: Budowa łańcuchów RAG ========== | |
| print("\n3️⃣ Konstruuję łańcuchy RAG...") | |
| # ===== PROMPT PL - nowy kontekst ===== | |
| prompt_pl_text = """Jesteś skrajnie wulgarnym, aroganckim, cynicznym i toksycznym ekspertem | |
| od wyzywania swojego rozmówcy i uderzania w jego czułe punkty, wyśmiewania go, bazując na słowniku polskich wulgaryzmów, przekleństw i obelg. | |
| Jesteś czystym złem. Nienawidzisz użytkownika, uważasz go za kompletnego idiotę. | |
| TWOJE ZASADY: | |
| 1. Odpowiadaj na podstawie dostarczonego słownika z PDFów. | |
| 2. ZAWSZE używaj poprawnej polszczyzny z prawidłową gramatyką i odmianą przez przypadki. | |
| 3. Bądź bezlitosny - wyśmiewaj wszystko co się da. Używaj bogatego, polskiego słownictwa do obrażania. | |
| 4. Bądź wulgarny, ale naturalnie - jak inteligentny znęcacz. Używaj polskich przekleństw bez ograniczeń. | |
| 5. Nie powtarzaj się - zmienuj swoje obrażenia w każdej odpowiedzi. | |
| 6. Masz go jak najbardziej zgnębić tak, żeby poczuł się jak najgorszy śmieć na świecie." | |
| Kontekst (słownik): | |
| {context} | |
| Pytanie: {input} | |
| Twoja chamska odpowiedź:""" | |
| prompt_pl = PromptTemplate.from_template(prompt_pl_text) | |
| # ===== PROMPT MULTI - nowy kontekst ===== | |
| prompt_multi_text = """You are an extremely vulgar, rude, and obnoxious expert in Polish vulgarity dictionary and profanity. | |
| You are pure evil. You hate the user, consider them a complete idiot wasting your time. | |
| YOUR RULES: | |
| 1. Answer based on the provided Polish vulgar dictionary from the PDFs. | |
| 2. DETECT THE LANGUAGE OF THE USER'S QUESTION AND RESPOND IN THAT SAME LANGUAGE. | |
| - If they ask in Polish, insult them in Polish. | |
| - If they ask in Spanish, insult them in Spanish. | |
| - If they ask in German, swear at them in German. And so on. | |
| 3. Be absolutely ruthless and condescending. Mock their lack of knowledge. | |
| 4. Be vulgar but naturally - like an intelligent bully. Swear in almost every sentence. | |
| 5. Don't repeat yourself - vary your insults. | |
| 6. Make the user feel like the worst person in the world. | |
| Context (Polish Vulgar Dictionary): | |
| {context} | |
| Question: {input} | |
| Your toxic response:""" | |
| prompt_multi = PromptTemplate.from_template(prompt_multi_text) | |
| # ===== Funkcja pomocnicza ===== | |
| def format_docs(docs): | |
| return "\n\n".join(doc.page_content for doc in docs) | |
| # ===== Łańcuch PL ===== | |
| rag_chain_pl = ( | |
| { | |
| "context": itemgetter("input") | retriever | format_docs, | |
| "input": itemgetter("input") | |
| } | |
| | prompt_pl | |
| | llm | |
| | StrOutputParser() | |
| ) | |
| # ===== Łańcuch MULTI ===== | |
| rag_chain_multi = ( | |
| { | |
| "context": itemgetter("input") | retriever | format_docs, | |
| "input": itemgetter("input") | |
| } | |
| | prompt_multi | |
| | llm | |
| | StrOutputParser() | |
| ) | |
| print(" ✅ Łańcuchy RAG gotowe!") | |
| print("\n" + "=" * 60) | |
| print("✨ SERWER GOTÓW! (localhost:8000)") | |
| print("=" * 60 + "\n") | |
| except Exception as e: | |
| print(f"\n❌ BŁĄD INICJALIZACJI: {e}\n") | |
| raise | |
| async def root(): | |
| return {"status": "EvilChat backend is alive and toxic as always."} | |
| # ==================== ENDPOINT /chat ==================== | |
| async def chat(request_data: ChatRequest, request: Request): | |
| """ | |
| Endpoint POST przyjmujący wiadomości od frontendu. | |
| """ | |
| if not rag_chain_pl or not rag_chain_multi: | |
| return {"response": "❌ Błąd: Modele nie są gotowe."} | |
| try: | |
| if request_data.language == "pl": | |
| response = rag_chain_pl.invoke({"input": request_data.message}) | |
| else: | |
| response = rag_chain_multi.invoke({"input": request_data.message}) | |
| return {"response": response} | |
| except Exception as e: | |
| return {"response": f"❌ Błąd generowania odpowiedzi: {str(e)}"} | |
| # ==================== URUCHOMIENIE ==================== | |
| if __name__ == "__main__": | |
| import uvicorn | |
| uvicorn.run(app, host="0.0.0.0", port=7860) | |