customer chatbot

examples/panel/customer-chatbot/app.py

@@ -0,0 +1,189 @@
+"""
+Customer chatbot.
+
+Input:
+Users can submit queries that describe the type of books they are looking for, e.g., suggest fiction novels.
+Users can also ask the chat assistant for books by specific author, e.g., recommend books by Dan Brown.
+Answers to user's queries will be based on the Goodreads dataset:
+https://www.kaggle.com/datasets/cristaliss/ultimate-book-collection-top-100-books-up-to-2023
+
+Application logic:
+The app determines the closest matches by comparing user query's embedding to the available
+book embeddings. Embeddings of books are pre-computed on the description column of every book
+and stored in the assets/ folder.
+
+Response:
+The chat assistant then determines the top relevant answers shortlisted by comparing embeddings and
+provides the top 5 recommendations.
+"""
+
+import json
+
+import panel as pn
+from openai import OpenAI
+from scipy.spatial import KDTree
+import numpy as np
+import pandas as pd
+from pathlib import Path
+from _wandb import WeightsBiasesTracking
+import datetime
+
+from util import get_embedding_from_text
+
+ORDER_CANCEL = False
+
+WEIGHTS_AND_BIASES_TRACKING = False
+
+if WEIGHTS_AND_BIASES_TRACKING:
+    wandb_client = WeightsBiasesTracking()
+
+
+client = OpenAI()
+
+df = pd.read_csv("orders.csv")
+
+
+def detect_order_number(user_query):
+    system_prompt = f"""
+    You're a system that determines the invoice number or the order number in the user query. 
+
+    You need to return only the order number.If no invoice number or order number is found 
+    return the string None.
+"""
+    response = client.chat.completions.create(
+        model="gpt-3.5-turbo",
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": "What were the items in order 536365?"},
+            {"role": "system", "content": "536365"},
+            {"role": "user", "content": user_query},
+        ],
+        seed=42,
+        n=1,
+    )
+    invoice_number = response.choices[0].message.content
+    print(f"invoice_number: {invoice_number}")
+    return invoice_number if invoice_number != "None" else ""
+
+
+def detect_intent(user_query):
+    system_prompt = f"""
+    You're a system that determines the intent of the user query. 
+
+    You need to return only the intent and no additional sentences.
+    If relevant intent is not found then return the string None.
+
+    Valid intents = ["TOTAL_ORDER_COST", "CANCEL", "NUMBER_OF_ITEMS", "ORDER_ITEM_DETAILS", "CUSTOMER_ORDERS"]
+"""
+    response = client.chat.completions.create(
+        model="gpt-3.5-turbo",
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": "What is the total cost of the order 536365"},
+            {"role": "system", "content": "TOTAL_ORDER_COST"},
+            {"role": "user", "content": "total cost of all items in the order 536365"},
+            {"role": "system", "content": "TOTAL_ORDER_COST"},
+            {"role": "user", "content": "Which items were ordered in 536364"},
+            {"role": "system", "content": "ORDER_ITEM_DETAILS"},
+            {"role": "user", "content": "Can i cancel the order 458891"},
+            {"role": "system", "content": "CANCEL"},
+            {"role": "user", "content": "How many items were ordered in Invoice 558420"},
+            {"role": "system", "content": "NUMBER_OF_ITEMS"},
+            {"role": "user", "content": "Purchases made by Customer ID 17850"},
+            {"role": "system", "content": "CUSTOMER_ORDERS"},
+            {"role": "user", "content": user_query},
+        ],
+        seed=42,
+        n=1,
+    )
+    intent = response.choices[0].message.content
+    return intent if intent != None else ""
+
+
+def cancel_order(invoice_number):
+    cancel = {}
+    print(df.head())
+    invoice_details = df.loc[df['InvoiceNo'] == invoice_number]
+    print(len(invoice_details))
+    if len(invoice_details) == 0:
+        cancel['eligible'] = False
+        cancel['reason'] = f"Order not found: {invoice_number}"
+        ORDER_CANCEL = False
+        return cancel
+
+    invoice_date = invoice_details.iloc[0]['InvoiceDate']
+    print(f"invoice_date: {invoice_date}")
+    date_object = datetime.datetime.strptime(invoice_date.split()[0], "%d/%m/%y")
+    print(date_object)
+    start_date = datetime.datetime(2010, 11, 1)
+    end_date = datetime.datetime(2010, 12, 2)
+    if start_date < date_object < end_date:
+        cancel['eligible'] = True
+        ORDER_CANCEL = True
+        return cancel
+    else:
+        cancel['eligible'] = False
+        cancel['reason'] = f"Order {invoice_number} not eligible for cancellation."
+        ORDER_CANCEL = False
+        return cancel
+
+
+def customer_chatbot_agent(user_query, verbose=False, tracking=False):
+    """An agent that can respond to customer's queries regarding orders"""
+
+    if ORDER_CANCEL and "yes" in user_query:
+        ORDER_CANCEL = False
+        return f"Order successfully cancelled"
+
+    invoice_number = detect_order_number(user_query)
+    if not invoice_number:
+        return "Please provide an Order ID"
+    intent = detect_intent(user_query)
+    print(intent)
+    if not intent:
+        return "Please provide more details regarding the action you want to take on the order."
+
+    if intent != "CANCEL":
+        return "We only support order cancellation requests. "
+
+    return cancel_order(invoice_number)
+
+    return "Sorry, we couldn't understand your query!"
+
+    # start_time_ms = datetime.datetime.now().timestamp() * 1000
+    #
+    # response = client.chat.completions.create(
+    #     model="gpt-3.5-turbo",
+    #     messages=[
+    #         {"role": "system", "content": system_prompt},
+    #         {"role": "user", "content": user_query},
+    #     ],
+    #     seed=42,
+    #     n=1,
+    # )
+    #
+    # end_time_ms = round(datetime.datetime.now().timestamp() * 1000)
+    #
+    # if tracking:
+    #     wandb_client.create_trace(system_prompt, response, user_query, start_time_ms, end_time_ms)
+
+    # return response.choices[0].message.content
+
+
+def callback(contents: str, user: str, instance: pn.chat.ChatInterface):
+    return customer_chatbot_agent(contents, tracking=WEIGHTS_AND_BIASES_TRACKING)
+
+
+chat_interface = pn.chat.ChatInterface(callback=callback, callback_exception='verbose')
+chat_interface.send(
+    "I am a customer chat assistant! "
+    "Currently we support only order cancellations. Please mention Order ID in your request."
+    "You can deploy your own by signing up at https://ploomber.io",
+    user="System",
+    respond=False,
+)
+
+pn.template.MaterialTemplate(
+    title="Customer Chatbot",
+    main=[chat_interface],
+).servable()

examples/panel/customer-chatbot/requirements.txt

@@ -0,0 +1,5 @@
+panel
+openai
+scipy
+numpy
+wandb