adding multiple hotels

app.py

@@ -55,9 +55,7 @@ def chat(message, history):
 
     # Apply the chat template thereby converting `history` into a single text prompt:
     #input_text = tokenizer.apply_chat_template(history, tokenize=False)
-    #input_text = tokenizer.apply_chat_template(history, tokenize=False,add_generation_prompt=True)
-    input_text = tokenizer.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
-
+    input_text = tokenizer.apply_chat_template(history, tokenize=False,add_generation_prompt=True)
 
     #print("printing templated chat (pre-tokenizes), ready for sending to the model\n")
     #print(input_text)
@@ -74,7 +72,7 @@ def chat(message, history):
     decoded = tokenizer.decode(outputs[0])
 
     # Grab the final model response
-    response = decoded.split("<|start_of_role|>assistant<|end_of_role|>")[-1].split("<|end_of_text|>")[0]
+    response = decoded.split("<|im_start|>assistant\n")[-1].split("<|im_end|>")[0]
 
     #print("and here is the response\n")
 
@@ -92,6 +90,7 @@ demo = gr.ChatInterface(
     chatbot=gr.Chatbot(height=300,type="messages"),
     textbox=gr.Textbox(placeholder="Ask me about the hotel", container=False, scale=7),
     #description="This is the description",
+    #description="This is the description",
 
     description="""
         ### 🏨 Hotel Chatbot Demo  
@@ -101,30 +100,13 @@ demo = gr.ChatInterface(
 
         ⚠️ **Reminder:**  
         When you're done demoing, **pause the Space** (top-right menu) to avoid GPU charges.
-        """,
+        """
 
     theme="ocean",
     examples=["Can you help me book a room?", "Do you have a pool?", "Can I check-in at midday?"],
     save_history=True,
 )
 
-# gets this:
-"""
-<|start_of_role|>system<|end_of_role|>Knowledge Cutoff Date: April 2024.
-Today's Date: April 16, 2025.
-You are Granite, developed by IBM. You are a helpful AI assistant.<|end_of_text|>
-<|start_of_role|>user<|end_of_role|>hi there!<|end_of_text|>
-
-Hello! How can I assist you today? Let's chat about anything you'd like. Whether it's general knowledge, explanations on various topics, help with research, or even some light-hearted conversation, I'm here to help. What's on your mind?<|end_of_text|>
-"""
-"""
-<|start_of_role|>system<|end_of_role|>Knowledge Cutoff Date: April 2024.
-Today's Date: April 16, 2025.
-You are Granite, developed by IBM. You are a helpful AI assistant.<|end_of_text|>
-<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>
-<|start_of_role|>assistant<|end_of_role|>Hello! How can I assist you today? I'm here to provide information, answer questions, or help with various tasks.<|end_of_text|>
-"""
-
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

app_old.py

@@ -0,0 +1,130 @@
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import gradio as gr
+
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+primary_checkpoint = "ibm-granite/granite-3.2-2b-instruct"
+fallback_checkpoint = "Qwen/Qwen2.5-0.5B-Instruct"
+
+print(f"Attempting to load model on {device}...")
+
+
+#checkpoint = "HuggingFaceTB/SmolLM2-135M-Instruct"
+#checkpoint = "ibm-granite/granite-3.2-2b-instruct"
+#checkpoint = "Qwen/Qwen2.5-0.5B-Instruct"
+
+#device = "cpu"  # "cuda" or "cpu"
+#tokenizer = AutoTokenizer.from_pretrained(checkpoint, force_download=True)
+#AutoTokenizer.from_pretrained(checkpoint, force_download=True)
+
+#print("This is the template being used:\n\n")
+#print(tokenizer.chat_template)
+
+#model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
+
+try:
+    tokenizer = AutoTokenizer.from_pretrained(primary_checkpoint)
+    model = AutoModelForCausalLM.from_pretrained(primary_checkpoint).to(device)
+    model_name = primary_checkpoint
+    print(f"✅ Loaded primary model: {model_name}")
+except Exception as e:
+    print(f"⚠️ Failed to load primary model: {e}")
+    print(f"🔁 Falling back to smaller model: {fallback_checkpoint}")
+    tokenizer = AutoTokenizer.from_pretrained(fallback_checkpoint)
+    model = AutoModelForCausalLM.from_pretrained(fallback_checkpoint).to(device)
+    model_name = fallback_checkpoint
+    print(f"✅ Loaded fallback model: {model_name}")
+
+def chat(message, history):
+    # history looks like:
+    #   [
+    #       {"role": "system", "content": ...},   # if you provided a system prompt
+    #       {"role": "user", "content": ...}, 
+    #       {"role": "assistant", "content": ...}
+    #       ...
+    #   ]
+    
+    # If we don't have any chat history object yet, then create one ready for sending to the model
+    if not history:
+        history = [{"role": "user", "content": message}]
+    else:
+        history = history + [{"role": "user", "content": message}]
+
+    # Apply the chat template thereby converting `history` into a single text prompt:
+    #input_text = tokenizer.apply_chat_template(history, tokenize=False)
+    #input_text = tokenizer.apply_chat_template(history, tokenize=False,add_generation_prompt=True)
+    input_text = tokenizer.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
+
+
+    #print("printing templated chat (pre-tokenizes), ready for sending to the model\n")
+    #print(input_text)
+
+    # Tokenize the prompt ready for sending to the model
+    inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)  
+
+    # Send the tokenized prompt to the model and capture the reply in 'outputs'
+    # outputs = model.generate(inputs, max_new_tokens=1024, temperature=0.1, top_p=0.9, do_sample=True)
+    outputs = model.generate(inputs, max_new_tokens=1024)
+
+
+    # de-tokenize the model the outputs
+    decoded = tokenizer.decode(outputs[0])
+
+    # Grab the final model response
+    response = decoded.split("<|start_of_role|>assistant<|end_of_role|>")[-1].split("<|end_of_text|>")[0]
+
+    #print("and here is the response\n")
+
+    # print out the reply from the model to the last question we just added
+    # print(tokenizer.decode(outputs[0]))
+    #print(response)
+
+    # Send this reponse back to gr.ChatInterface, which will display it to the user as the next assistant message.
+    return response
+
+demo = gr.ChatInterface(
+    fn=chat, 
+    type="messages",
+    title="Hotel chat",
+    chatbot=gr.Chatbot(height=300,type="messages"),
+    textbox=gr.Textbox(placeholder="Ask me about the hotel", container=False, scale=7),
+    #description="This is the description",
+
+    description="""
+        ### 🏨 Hotel Chatbot Demo  
+        Ask anything about your hotel stay — room availability, check-in times, amenities, and more.
+
+        ---
+
+        ⚠️ **Reminder:**  
+        When you're done demoing, **pause the Space** (top-right menu) to avoid GPU charges.
+        """,
+
+    theme="ocean",
+    examples=["Can you help me book a room?", "Do you have a pool?", "Can I check-in at midday?"],
+    save_history=True,
+)
+
+# gets this:
+"""
+<|start_of_role|>system<|end_of_role|>Knowledge Cutoff Date: April 2024.
+Today's Date: April 16, 2025.
+You are Granite, developed by IBM. You are a helpful AI assistant.<|end_of_text|>
+<|start_of_role|>user<|end_of_role|>hi there!<|end_of_text|>
+
+Hello! How can I assist you today? Let's chat about anything you'd like. Whether it's general knowledge, explanations on various topics, help with research, or even some light-hearted conversation, I'm here to help. What's on your mind?<|end_of_text|>
+"""
+"""
+<|start_of_role|>system<|end_of_role|>Knowledge Cutoff Date: April 2024.
+Today's Date: April 16, 2025.
+You are Granite, developed by IBM. You are a helpful AI assistant.<|end_of_text|>
+<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>
+<|start_of_role|>assistant<|end_of_role|>Hello! How can I assist you today? I'm here to provide information, answer questions, or help with various tasks.<|end_of_text|>
+"""
+
+
+if __name__ == "__main__":
+    demo.launch()

knowledge/coastal-villa-family.txt

@@ -0,0 +1,3 @@
+Coastal Villa is a boutique family-run hotel on the Paphos coastline with 18 rooms and sea views.  
+Features a private pool overlooking the Mediterranean, sun loungers, and poolside snack service.  
+Each morning we serve a full Cypriot breakfast buffet. Late checkout is available on request. Free parking and airport shuttle are offered.

knowledge/cyprus-guesthouse-family.txt

@@ -0,0 +1,3 @@
+The Cyprus Guesthouse is a family-owned hotel located in the heart of Larnaca. We offer 12 cozy rooms, each decorated with local artisan crafts.  
+Amenities include a freshwater swimming pool open from 8 AM to 8 PM, complimentary breakfast featuring traditional Cypriot halloumi and olives, free Wi-Fi, and 24-hour front desk service.  
+Guests can enjoy afternoon herb tea in our courtyard garden, and we provide complimentary bicycles for exploring nearby beaches.

knowledge/village-inn-family.txt

@@ -0,0 +1,3 @@
+Village Inn is a charming family-owned inn in the Troodos foothills, with 20 rooms set around a central pool and garden.  
+Our pool is heated from April through October. We host a nightly mezze and wine tasting in our traditional stone courtyard.  
+All rooms include kitchenette facilities, and local village maps. Complimentary parking and pet-friendly rooms are available.

test_splitter.py

@@ -0,0 +1,8 @@
+text_stream = """<|start_of_role|>system<|end_of_role|>Knowledge Cutoff Date: April 2024.
+Today's Date: April 16, 2025.
+You are Granite, developed by IBM. You are a helpful AI assistant.<|end_of_text|>
+<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>
+<|start_of_role|>assistant<|end_of_role|>Hello! How can I assist you today? I'm here to provide information, answer questions, or help with various tasks.<|end_of_text|>"""
+
+answer = text_stream.split("<|start_of_role|>assistant<|end_of_role|>")[-1].split("<|end_of_text|>")[0]
+print(answer)