Delete llm_model.py

llm_model.py

@@ -1,91 +0,0 @@
-import os
-import traceback
-from datetime import datetime
-from pydantic import BaseModel
-from unsloth import FastLanguageModel
-from transformers import pipeline
-import torch
-from log import log
-
-class Message(BaseModel):
-    user_input: str
-
-class LLMModel:
-    def __init__(self):
-        self.model = None
-        self.tokenizer = None
-        self.pipe = None
-
-    def setup(self, s_config, project_config, project_path):
-        try:
-            log("🧠 LLMModel setup() başladı")
-            device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-            log(f"📡 Kullanılan cihaz: {device}")
-
-            model_base = project_config["model_base"]
-            hf_token = s_config.get_auth_token()
-
-            # Ortam değişkenleri
-            os.environ.setdefault("HF_HOME", "/app/.cache")
-            os.environ.setdefault("HF_DATASETS_CACHE", "/app/.cache")
-            os.environ.setdefault("HF_HUB_CACHE", "/app/.cache")
-            os.environ.setdefault("TRITON_CACHE_DIR", "/tmp/.triton")
-            os.environ.setdefault("TORCHINDUCTOR_CACHE_DIR", "/tmp/torchinductor_cache")
-
-            log(f"📦 UnsLoTH modeli yükleniyor: {model_base}")
-            self.model, self.tokenizer = FastLanguageModel.from_pretrained(
-                model_name=model_base,
-                load_in_4bit=True,
-                token=hf_token,
-                cache_dir="/app/.cache"
-            )
-            FastLanguageModel.for_inference(self.model)
-
-            self.pipe = pipeline(
-                "text-generation",
-                model=self.model,
-                tokenizer=self.tokenizer,
-                device_map="auto"
-            )
-
-            log("✅ LLMModel setup() başarıyla tamamlandı.")
-
-        except Exception as e:
-            log(f"❌ LLMModel setup() hatası: {e}")
-            traceback.print_exc()
-
-    async def generate_response_with_messages(self, messages, project_config, system_prompt):
-        try:
-            # Son 10 mesajla sınırla
-            max_history = 10
-            recent_messages = messages[-max_history:]
-
-            # System prompt + mesaj geçmişini birleştir
-            all_messages = [{"role": "system", "content": system_prompt}] + recent_messages
-            history_text = ""
-            for msg in all_messages:
-                if msg["role"] == "user":
-                    history_text += f"Kullanıcı: {msg['content']}\n"
-                elif msg["role"] == "assistant":
-                    history_text += f"Asistan: {msg['content']}\n"
-                elif msg["role"] == "system":
-                    history_text += f"{msg['content']}\n"
-
-            full_prompt = f"{history_text}Asistan:"
-            log("🧠 LLM çağrısı başlatılıyor...")
-
-            outputs = self.pipe(
-                full_prompt,
-                max_new_tokens=512,
-                temperature=0.2,
-                top_p=0.95,
-                repetition_penalty=1.1
-            )
-            answer = outputs[0]["generated_text"].replace(full_prompt, "").strip()
-            log("✅ LLM cevabı başarıyla alındı.")
-            return answer
-
-        except Exception as e:
-            log(f"❌ generate_response_with_messages sırasında hata oluştu: {e}")
-            traceback.print_exc()
-            return "Hata: Model cevabı alınamadı."