Update app.py

app.py

@@ -27,28 +27,24 @@ def format_prompt(message, history):
     prompt += f"[INST] {message} [/INST]"
     return prompt
 
-def generate_response(prompt, database):
-    if prompt in database:
-        return database[prompt]
-    else:
-        response = next(client.text_generation(prompt, details=True, return_full_text=False)).token.text
-        database[prompt] = response
-        save_database(database)
-        return response
-
 def generate(
-    prompt, history, database, temperature=0.9, max_new_tokens=2000, top_p=0.9, repetition_penalty=1.2,
+    prompt, history, temperature=0.9, max_new_tokens=2000, top_p=0.9, repetition_penalty=1.2,
 ):
+    database = load_database()  # Load the database
     temperature = float(temperature)
     if temperature < 1e-2:
         temperature = 1e-2
     top_p = float(top_p)
 
     formatted_prompt = format_prompt(prompt, history)
-    response = generate_response(formatted_prompt, database)
-    yield response
+    if formatted_prompt in database:
+        response = database[formatted_prompt]
+    else:
+        response = next(client.text_generation(formatted_prompt, details=True, return_full_text=False)).token.text
+        database[formatted_prompt] = response
+        save_database(database)  # Save the updated database
 
-database = load_database()
+    yield response
 
 css = """
   #mkd {
@@ -61,8 +57,7 @@ css = """
 with gr.Blocks(css=css) as demo:
     gr.ChatInterface(
         generate,
-        examples=[["What is the secret to life?"], ["Write me a recipe for pancakes."], ["Write a short story about Paris."]],
-        database=database
+        examples=[["What is the secret to life?"], ["Write me a recipe for pancakes."], ["Write a short story about Paris."]]
     )
 
 demo.launch(debug=True)