Update app.py

app.py

@@ -8,7 +8,6 @@ app = FastAPI()
 tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
 model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
 
-# Memory store per user
 chat_history = {}
 
 @app.get("/ai")
@@ -17,20 +16,17 @@ async def chat(request: Request):
     user_input = query_params.get("query", "")
     user_id = query_params.get("user_id", "default")
 
-    # Pull user history
-    user_history = chat_history.get(user_id, [])
-
-    # Tokenize with context
     new_input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors='pt')
-
-    # Concatenate history if available
+    user_history = chat_history.get(user_id, [])
     bot_input_ids = torch.cat(user_history + [new_input_ids], dim=-1) if user_history else new_input_ids
 
-    # Generate response
     output_ids = model.generate(bot_input_ids, max_new_tokens=100, pad_token_id=tokenizer.eos_token_id)
     response = tokenizer.decode(output_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
 
-    # Save history
     chat_history[user_id] = [bot_input_ids, output_ids]
+    return JSONResponse({"reply": response})
 
-    return JSONResponse({"reply": response})
+# ✅ Add this to launch properly on Hugging Face Spaces
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)