Update g1.py

g1.py

@@ -2,18 +2,9 @@ import groq
 import time
 import json
 import logging
-from typing import List, Dict, Generator, Optional, Any, Tuple
-from telegram.ext import (
-    Updater,
-    CommandHandler,
-    MessageHandler,
-    filters,  # Исправленный импорт
-    CallbackContext
-)
-from telegram import Update, Bot
+from typing import Generator
 from tqdm import tqdm
 
-# Конфигурация
 logging.basicConfig(
     format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
     level=logging.INFO
@@ -22,119 +13,44 @@ logger = logging.getLogger(__name__)
 
 class ResponseGenerator:
     def __init__(self):
-        self.models = {
-            'iterative': "llama3-70b-8192",
-            'final': "llama3-70b-8192",
-            'code': "llama3-8b-8192"
-        }
+        self.model = "llama3-70b-8192"
         self.max_tokens = {
             'intermediate': 2000,
             'final': 1000
         }
 
-    def make_api_call(self, messages: List[Dict], max_tokens: int, is_final: bool = False) -> Any:
-        """Метод для API вызовов с обработкой ошибок"""
-        model = self.models['final'] if is_final else self.models['iterative']
-        
-        for attempt in range(3):
-            try:
-                params = {
-                    "model": model,
-                    "messages": messages,
-                    "max_tokens": max_tokens,
-                    "temperature": 0.5
-                }
-                
-                if not is_final:
-                    params["response_format"] = {"type": "json_object"}
-                
-                response = groq.Groq().chat.completions.create(**params)
-                content = response.choices[0].message.content
-                
-                if is_final:
-                    return content
-                
-                try:
-                    parsed = json.loads(content)
-                    if not all(k in parsed for k in ("title", "content", "next_action")):
-                        raise ValueError("Missing required keys")
-                    return parsed
-                except json.JSONDecodeError as e:
-                    raise ValueError(f"Invalid JSON: {e}")
-                    
-            except Exception as e:
-                logger.error(f"Attempt {attempt + 1} failed: {e}")
-                if attempt == 2:
-                    return {
-                        "title": "Error",
-                        "content": str(e),
-                        "next_action": "final_answer"
-                    }
-                time.sleep(2 ** attempt)
+    def _make_api_call(self, messages, max_tokens, is_final=False):
+        try:
+            response = groq.Groq().chat.completions.create(
+                model=self.model,
+                messages=messages,
+                max_tokens=max_tokens,
+                temperature=0.5,
+                **({"response_format": {"type": "json_object"}} if not is_final else {})
+            return response.choices[0].message.content
+        except Exception as e:
+            logger.error(f"API Error: {str(e)}")
+            return str(e)
 
     def generate_response(self, prompt: str) -> Generator:
-        """Генерация ответа с прогресс-баром"""
-        system_msg = {
-            "role": "system",
-            "content": "Вы - эксперт, объясняющий мысли шагами. Формат ответа: JSON с title, content, next_action."
-        }
-        
-        messages = [system_msg, {"role": "user", "content": prompt}]
-        steps = []
+        messages = [
+            {"role": "system", "content": "You're an AI assistant. Explain your reasoning step-by-step in JSON format."},
+            {"role": "user", "content": prompt}
+        ]
         
-        with tqdm(total=10, desc="Генерация ответа", unit="step") as pbar:
-            for step_num in range(1, 11):
-                start = time.time()
-                step_data = self.make_api_call(messages, self.max_tokens['intermediate'])
-                elapsed = time.time() - start
-                
-                step = {
-                    "title": step_data.get("title", "Step"),
-                    "content": step_data.get("content", ""),
-                    "time": elapsed
-                }
-                steps.append(step)
-                
+        with tqdm(total=5, desc="Generating response") as pbar:
+            for _ in range(5):
+                result = self._make_api_call(messages, self.max_tokens['intermediate'])
+                messages.append({"role": "assistant", "content": result})
                 pbar.update(1)
-                pbar.set_postfix({"step": step_num, "time": f"{elapsed:.2f}s"})
-                messages.append({"role": "assistant", "content": json.dumps(step_data)})
+                yield result
 
-                if step_data.get("next_action") == "final_answer":
-                    break
-
-        # Финальный ответ
-        messages.append({"role": "user", "content": "Дай окончательный ответ в чистом тексте."})
-        final_answer = self.make_api_call(messages, self.max_tokens['final'], is_final=True)
-        
-        return {
-            "steps": steps,
-            "final_answer": final_answer,
-            "total_time": sum(s['time'] for s in steps)
-        }
-
-# Gradio интерфейс
-def format_response(response):
-    formatted = ""
-    for step in response['steps']:
-        formatted += f"### {step['title']}\n{step['content']}\n\n"
-    formatted += f"**Финальный ответ:**\n{response['final_answer']}\n\n"
-    formatted += f"**Общее время: {response['total_time']:.2f}с**"
-    return formatted
+        final_answer = self._make_api_call(messages, self.max_tokens['final'], is_final=True)
+        yield final_answer
 
 def gradio_interface(prompt: str) -> str:
-    generator = ResponseGenerator()
-    response = generator.generate_response(prompt)
-    return format_response(response)
-
-if __name__ == "__main__":
-    import gradio as gr
-    
-    demo = gr.Interface(
-        fn=gradio_interface,
-        inputs=gr.Textbox(lines=3, label="Ваш запрос"),
-        outputs=gr.Markdown(label="Результат"),
-        title="🧠 G1: Цепочки рассуждений с Llama-3",
-        description="Генерация пошаговых рассуждений с использованием Groq API"
-    )
-    
-    demo.launch()
+    gen = ResponseGenerator()
+    response = []
+    for chunk in gen.generate_response(prompt):
+        response.append(chunk)
+    return "\n\n".join(response)