Update app.py

app.py

@@ -1,42 +1,46 @@
 import gradio as gr
-from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
 
-# 1. Choose a bilingual or multilingual QA model
-MODEL_NAME = "deepset/xlm-roberta-large-squad2"
+# Use a modern, instruction-tuned model. FLAN-T5 Base is lightweight enough for free CPU usage.
+MODEL_NAME = "google/flan-t5-base"
 
-# 2. Load model + tokenizer
+# Load the tokenizer and model
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-model = AutoModelForQuestionAnswering.from_pretrained(MODEL_NAME)
+model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
 
-# 3. Initialize QA pipeline
-qa_pipeline = pipeline("question-answering", model=model, tokenizer=tokenizer)
+# Set up the text-to-text generation pipeline
+llm_pipeline = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
 
-# 4. Load or define custom knowledge base
-with open("knowledge.txt", "r", encoding="utf-8") as f:
-    knowledge_text = f.read()
-
-# 5. Define function to answer questions
-def answer_question(question):
-    if not question.strip():
-        return "Please ask a valid question."
-    try:
-        result = qa_pipeline(question=question, context=knowledge_text)
-        return result["answer"]
-    except Exception as e:
-        return f"Error: {str(e)}"
+def budtender_response(user_input: str) -> str:
+    """
+    This function crafts a prompt to instruct the model to act as BudtenderPro,
+    a bilingual, professional budtender for adult and medical cannabis customers in New Mexico.
+    """
+    # Construct a detailed system prompt for context.
+    prompt = (
+        "You are BudtenderPro, a friendly, knowledgeable, and culturally sensitive budtender for "
+        "both adult and medical cannabis customers in New Mexico. You are fluent in both English and Spanish. "
+        "You provide professional advice on strains, dosages, and product recommendations, keeping in mind legal regulations and community values. "
+        "Answer the user's question with clear, respectful, and useful guidance. \n\n"
+        f"User Question: {user_input}"
+    )
+    
+    # Generate the response. Adjust parameters (max_length, temperature) as needed.
+    outputs = llm_pipeline(prompt, max_length=256, do_sample=True, temperature=0.7)
+    answer = outputs[0]['generated_text']
+    return answer.strip()
 
-# 6. Build Gradio interface
+# Set up the Gradio interface
 iface = gr.Interface(
-    fn=answer_question,
-    inputs=gr.Textbox(lines=2, placeholder="Enter your question here..."),
-    outputs="text",
-    title="Budtender LLM (Bilingual QA)",
+    fn=budtender_response,
+    inputs=gr.Textbox(lines=3, placeholder="Ask your cannabis-related question here..."),
+    outputs=gr.Textbox(label="BudtenderPro Response"),
+    title="BudtenderPro LLM Demo",
     description=(
-        "A bilingual Q&A model trained on Spanish and English data. "
-        "Ask your cannabis-related questions here!"
+        "A modern, bilingual budtender assistant for adult and medical cannabis questions in New Mexico. "
+        "Powered by Google FLAN-T5 Base, this demo leverages an up-to-date Hugging Face workflow to provide professional guidance."
     )
 )
 
-# 7. Launch app
 if __name__ == "__main__":
     iface.launch()