Update app.py

app.py

@@ -16,18 +16,16 @@ tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(model_name)
 model.eval()
 
-# 🔹 настройки
 MAX_HISTORY = 40
 NUM_WORKERS = 3
 
-db = OrderedDict()   # message -> {status, reply}
+db = OrderedDict()
 queue = []
 
 class Message(BaseModel):
     message: str
 
 
-# 🔥 STREAMING GENERATION
 def generate_ai_stream(message: str):
     prompt = f"User: {message}\nAssistant: Answer clearly and fully:\n"
 
@@ -51,24 +49,22 @@ def generate_ai_stream(message: str):
 
     partial = ""
 
+    # 🔥 STREAMING
     for text in streamer:
         partial += text
 
-        # 🔥 обновляем ответ в реальном времени
         if message in db:
             db[message]["reply"] = partial
 
-   
-
- final_text = partial.strip() + " full generated"
+    # 🔥 КОНЕЦ ГЕНЕРАЦИИ (ВОТ ТУТ ДЕТЕКТ)
+    final_text = partial.strip() + " full generated"
 
     if message in db:
         db[message]["reply"] = final_text
-        db[message]["status"] = "done"
 
     return final_text
 
-# 🔥 WORKER
+
 def worker():
     while True:
         if queue:
@@ -86,7 +82,6 @@ def worker():
             time.sleep(0.01)
 
 
-# 🔥 запускаем 3 воркера (ускорение x2-x3)
 for _ in range(NUM_WORKERS):
     threading.Thread(target=worker, daemon=True).start()
 
@@ -96,19 +91,14 @@ async def root():
     return PlainTextResponse("AI server работает")
 
 
-# 🔹 ASK
 @app.get("/ask")
 async def ask(message: str):
 
-    # кеш
     if message in db and db[message]["status"] == "done":
         return PlainTextResponse("cached")
 
     if message not in db:
-        db[message] = {
-            "status": "pending",
-            "reply": ""
-        }
+        db[message] = {"status": "pending", "reply": ""}
         queue.append(message)
 
         if len(db) > MAX_HISTORY:
@@ -117,7 +107,6 @@ async def ask(message: str):
     return PlainTextResponse("accepted")
 
 
-# 🔹 GET (визуальный стриминг)
 @app.get("/get")
 async def get(message: str):