Update app.py

app.py

@@ -2,50 +2,49 @@ import gradio as gr
 from datasets import load_dataset
 from transformers import AutoTokenizer, TFAutoModelForSeq2SeqLM
 
-
-# Load model and tokenizer from your Hugging Face model repo
 model_name = "NinaMwangi/T5_finbot"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = TFAutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-# Load dataset ===
 dataset = load_dataset("virattt/financial-qa-10K")["train"]
 
-# Function to retrieve matching context
 def get_context_for_question(question):
     for item in dataset:
         if item["question"].strip().lower() == question.strip().lower():
             return item["context"]
     return "No relevant context found."
 
-# Define the prediction function (inference)
 def generate_answer(question, chat_history):
-    context = get_context_for_question(question)
-    prompt = f"Q: {question} Context: {context} A:"
-
-
-    inputs = tokenizer(
-        prompt,
-        return_tensors="tf",
-        padding="max_length",
-        truncation=True,
-        max_length=256
-    )
-
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=64,
-        num_beams=4,
-        early_stopping=True
-    )
-
-    #return tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-    answer = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
-    chat_history.append([question, answer])
-    return "", chat_history
+    try:
+        if chat_history is None:
+            chat_history = []
+        context = get_context_for_question(question)
+        prompt = f"Q: {question} Context: {context} A:"
+
+        inputs = tokenizer(
+            prompt,
+            return_tensors="tf",
+            padding="max_length",
+            truncation=True,
+            max_length=256
+        )
+
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=64,
+            num_beams=4,
+            early_stopping=True
+        )
+
+        answer = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
+        chat_history.append([question, answer])
+        return "", chat_history
+    except Exception as e:
+        if chat_history is None:
+            chat_history = []
+        chat_history.append([question, f"Error: {str(e)}"])
+        return "", chat_history
 
-# === Gradio UI ===
 with gr.Blocks(theme=gr.themes.Base()) as interface:
     gr.Markdown(
         """
@@ -65,25 +64,27 @@ with gr.Blocks(theme=gr.themes.Base()) as interface:
             submit_btn = gr.Button("Send")
 
     clear_btn = gr.Button("Clear Chat")
-
-    # Chat state
     state = gr.State([])
 
-    # Bind functionality
     submit_btn.click(
-    generate_answer,
-    inputs=[question_box, state],
-    outputs=[question_box, chatbot],
-)
+        generate_answer,
+        inputs=[question_box, state],
+        outputs=[question_box, chatbot],
+    )
+
+    question_box.submit(
+        generate_answer,
+        inputs=[question_box, state],
+        outputs=[question_box, chatbot]
+    )
 
     clear_btn.click(
         lambda: ("", [], []),
         inputs=[],
         outputs=[question_box, chatbot, state],
-)
-
+    )
 
-# === Launch the app ===
 interface.launch(share=True)
 
 
+