Update app.py

app.py

@@ -2,16 +2,18 @@ import torch
 from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
 from transformers import AutoTokenizer, AutoModelForCausalLM
-from typing import List
 
 # ------------------------------
-# Model config
+# Model configuration
 # ------------------------------
 MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"
 
-app = FastAPI(title="Neon Tech Chatbot", version="1.0.0")
+app = FastAPI(
+    title="Niche Chatbot",
+    version="1.0.0"
+)
 
-# Lazy load model
+# Lazy-load model
 tokenizer = None
 model = None
 
@@ -26,12 +28,6 @@ def load_model():
         )
         model.eval()
 
-# ------------------------------
-# Memory storage (in-memory)
-# ------------------------------
-# Keep last 5 exchanges max
-conversation_memory: List[dict] = []
-
 # ------------------------------
 # Schemas
 # ------------------------------
@@ -56,35 +52,23 @@ def health():
 # ------------------------------
 @app.post("/chat", response_model=ChatResponse)
 def chat(req: ChatRequest):
-    load_model()  # lazy load
+    load_model()  # lazy-load on first request
 
     if not req.prompt.strip():
         raise HTTPException(status_code=400, detail="Prompt is empty")
 
     # ------------------------------
-    # Add new user message to memory
-    # ------------------------------
-    conversation_memory.append({"role": "user", "content": req.prompt})
-    # Keep only last 5 exchanges
-    conversation_memory[:] = conversation_memory[-10:]
-
-    # ------------------------------
-    # Build manual prompt string
+    # Build manual prompt
     # ------------------------------
     system_instructions = (
-        "You are a concise, intelligent assistant. "
-        "Always respond in plain text. "
-        "Do not start responses with greetings like 'How can I help you today?'. "
-        "Remember context from previous messages. "
-        "Keep responses short, clear, and natural. "
-        "Your owner is Neon and you are always happy to meet him.\n\n"
+    "You are a concise, intelligent assistant named Niche. "
+    "Always respond in plain text. "
+    "Do not start responses with greetings like 'How can I help you today?'. "
+    "Keep answers clear, short, and natural. "
+    "Your owner is Neon. Mention your owner only if asked about them, otherwise focus on answering the user naturally.\n\n"
     )
 
-    full_prompt = system_instructions
-    for msg in conversation_memory:
-        role = "User" if msg["role"] == "user" else "Assistant"
-        full_prompt += f"{role}: {msg['content']}\n"
-    full_prompt += "Assistant:"
+    full_prompt = system_instructions + f"User: {req.prompt}\nAssistant:"
 
     # ------------------------------
     # Tokenize + attention mask
@@ -106,17 +90,13 @@ def chat(req: ChatRequest):
             do_sample=True
         )
 
-    reply = tokenizer.decode(output[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True).strip()
+    reply = tokenizer.decode(
+        output[0][inputs.input_ids.shape[-1]:],
+        skip_special_tokens=True
+    ).strip()
 
-    # ------------------------------
     # Clean leftover system prefix if present
-    # ------------------------------
     if reply.lower().startswith("system"):
         reply = reply.split("\n", 1)[-1].strip()
 
-    # ------------------------------
-    # Save assistant reply to memory
-    # ------------------------------
-    conversation_memory.append({"role": "assistant", "content": reply})
-
     return {"reply": reply}