fast food bot order

telephony-bot-api/fastfood-order-bot-example/README.md

@@ -0,0 +1,117 @@
+# Project Overview
+
+The Project is to create a fast food ordering voicebot using Langchain and Voicegain Telephony Bot API. 
+The bot has the ability to order from a predefined menu that includes burgers, wraps, drinks, fries and potato wedges. 
+The bot can distinguish between all vegetarian items and non-veg items. 
+The user has the ability to customize their order by adding topics / add-ons to their burger or wrap 
+and have the ability to pick from 3 different sizes for the fries and potato wedges. 
+They can further create custom combo orders and have a choice between 3 - main item with a drink, 
+main item with a side of their fries or potato wedges, main with a side and a drink. 
+If an order qualifies for a combo order - the bot automatically creates a combo for the user. 
+The bot can summarize the users order at any point along with calculating the total cost of the order. 
+The user can further edit their order at any point - by either changing an item (replace item, change quantity etc) 
+or deleting an item entirely.
+
+# Project Steps
+
+There are mainly 3 steps to it - 
+- Implement the Bot Logic
+- Implement a Telephony Bot API service
+- Setting up the AIVR bot on Voicegain Customer Portal. 
+
+Each of these work hand in hand to create the bot.
+
+## 1. Bot Logic
+
+For the bot logic, we use LangChain which is a framework to build with LLMs by chaining interoperable components. 
+This helps in managing the workflow of the bot. 
+
+For Langchain Reference kindly refer the Documentation here - https://python.langchain.com/v0.2/docs/introduction/
+
+If the reader is interested in the detailed bot implementation, kindly read through --- 
+
+The bot logic is powered by LangChain and OpenAI's API. 
+In order to create the bot framework, chains are used to link the various functions that the bot can perform. 
+By using tools, the program is able to create a basic function framework that ChatGPT can use depending on the task. 
+It is important to note that the function description is key as this informs ChatGPT 
+what the function of each of these functions are, and can respectively pick the appropriate function for the task. 
+This logic for bot is present in [fast_food_bot.py](fast_food_bot.py).
+
+The core functions are therefore created as tool functions in LangChain. 
+Functions such as giving the order summary and the total order cost have their own function as well, 
+so as to prevent possible errors arising from ChatGPT.
+
+### 1.1 Tools
+
+- get_menu: Returns the menu and prices for burgers, wraps, fries, wedges, and drinks.
+- get_add_ons: Lists all possible add-ons and their costs.
+- get_combos(main_item, side=None, drink=None): Provides possible meal combos based on user selections.
+- update_order(new_order: OrderItem): Updates the current order, handles adding, updating, and deleting items.
+- get_final_price: Calculates the total price of the current order.
+- get_breakdown: Provides a detailed breakdown of the current order.
+
+### 1.2 Agents and Chains
+
+- ChatPromptTemplate: Defines the conversation prompt template.
+- create_tool_calling_agent: Creates an agent that uses the defined tools and prompt.
+- AgentExecutor: Manages the execution of the agent with tools.
+- RunnableWithMessageHistory: Manages conversation history and processes the inputs
+
+### 1.3 Steps to run the bot locally -
+
+- Install the requirements in requirements.txt
+- Set your openai api key in 'fast_food_bot.py' file.
+- Run the main.py file to start the local uvicorn server.
+- Run the http_client_example now to interact with the bot via text.
+
+
+## 2. Telephony Bot API Integration
+
+Reference Document: https://console.voicegain.ai/api-documentation#tag/aivr-callback
+
+Implement POST / PUT / DELETE
+
+- POST: We create a new call session with unique 'csid'. They can be used to uniquely identify the call.
+- PUT: If we get user output, call CHAIN to relay information to the bot. If not, reprompt the user for a response.
+- DELETE: Delete the session from Chain's memory
+
+This is the component that integrates Voicegain's Telephony Bot API with the function of the langChain bot. 
+This adds the ability to interact with the bot through speech, rather than text, 
+thereby enhancing the user interaction experience. This logic is present in [main.py](main.py).
+
+
+### 2.1 Functions
+
+- create_session(request: TelephonyBotPostReq): Adds a new call session with unique sid, csid.
+- update_session(request: TelephonyBotPutReq): Updates an existing order using the sid, csid based on the user input. Forwards input to the Fast Food Bot and returns the response.
+- delete_session(request: TelephonyBotDeleteReq): Removes the call session and associated chat history from the in-memory store by csid.
+
+
+## 3. Create AIVR App on Voicegain Portal.
+
+By making an AIVR App on the Voicegain Portal, we can enable the speech component to our food ordering bot. This has mainly 3 steps - 
+
+- Hosting Bot on a server
+- Providing the URL in the AIVR App created on the Voicegain Portal
+
+We explore these steps in detail below.
+
+### 3.1 Hosting Bot on Server
+
+When you've run the [main.py](main.py) file and the bot is running locally on your device, you need to host the bot on a public URL for using on AIVR app.
+One of the options for that is to use Ngrok - https://dashboard.ngrok.com/get-started/setup/windows
+
+### 3.2 Setting up AIVR App
+
+After you've configured the public URL for the food ordering bot, we need to setup the AIVR App using the following steps - 
+
+1. Login to the Voicegain console - https://console.voicegain.ai/specific/aivr-apps <img src="screenshots/Screenshot-1.png"/>
+2. Buy a phone number in the phone management section. Click on buy number, add your area code and buy. This will be used later in the Main DNS field for AIVR App. <img src="screenshots/Screenshot-2.png"/> <img src="screenshots/Screenshot-3.png"/> 
+3. Now move to 'Phone Apps' section from 'Phone Management' in the toolbar on the left. Add new AIVR App using the blue button on top-right and filling in the details. Remember to add the number you purchased in the 'Main DNIS' section. <img src="screenshots/Screenshot-4.png"/> <img src="screenshots/Screenshot-5.png"/>
+4. Click on Add Logic at the end and fill in details. The public URL where you host the bot goes in here <img src="screenshots/Screenshot-6.png"/>
+5. After that click on OK and Save to save your AIVR bot. You should get a screen like this with a number in Main DNIS field - <img src="screenshots/Screenshot-7.png"/>
+
+Now you can call on the phone number in the Main DNIS and you can use the food ordering bot on call.
+
+### Credits - 
+Voicegain Team

telephony-bot-api/fastfood-order-bot-example/fast_food_bot.py

@@ -0,0 +1,233 @@
+from langchain_openai import ChatOpenAI
+from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
+from langchain_core.pydantic_v1 import BaseModel, Field
+from langchain_core.chat_history import InMemoryChatMessageHistory
+from langchain_core.runnables.history import RunnableWithMessageHistory
+from langchain_core.tools import tool
+from langchain.agents import AgentExecutor, create_tool_calling_agent
+import os
+from typing import List
+
+os.environ["OPENAI_API_KEY"] = 'YOUR-OPENAI-KEY'
+
+llm = ChatOpenAI(model="gpt-4o-mini")
+
+
+# tool to contain the menu of the restaurant
+@tool
+def get_menu():
+    """Get fast food restaurant menu and price for each item. The price is in dollars
+    the menu has burgers and wraps. There are also fries and wedges that can be ordered, and there are 3 different sizes for them as well."""
+    menu = {
+        "burgers": [
+            {"name": "Cheeseburger", "type": "non-veg", "price": 10},
+            {"name": "Black Bean Burger", "type": "veg", "price": 12},
+            {"name": "Hamburger", "type": "non-veg", "price": 12},
+            {"name": "Crispy Chicken Burger", "type": "non-veg", "price": 12},
+            {"name": "Veggie Burger", "type": "veg", "price": 12},
+            {"name": "Fish Burger", "type": "non-veg", "price": 12},
+            {"name": "BBQ Chicken Burger", "type": "non-veg", "price": 12},
+            {"name": "Spicy Chicken Burger", "type": "non-veg", "price": 12},
+            {"name": "Turkey Burger", "type": "non-veg", "price": 12},
+            {"name": "Mushroom Swiss Burger", "type": "veg", "price": 12}
+        ],
+        "wraps": [
+            {"name": "Grilled Chicken Wrap", "type": "non-veg", "price": 10},
+            {"name": "Spicy Chicken Wrap", "type": "non-veg", "price": 10},
+            {"name": "BBQ Chicken Wrap", "type": "non-veg", "price": 10},
+            {"name": "Veggie Wrap", "type": "veg", "price": 10},
+            {"name": "Turkey Wrap", "type": "non-veg", "price": 10},
+            {"name": "Buffalo Chicken Wrap", "type": "non-veg", "price": 10},
+            {"name": "Chicken Caesar Wrap", "type": "non-veg", "price": 10},
+            {"name": "Steak Wrap", "type": "non-veg", "price": 10},
+            {"name": "Falafel Wrap", "type": "veg", "price": 10},
+            {"name": "Hummus Veggie Wrap", "type": "veg", "price": 10}
+        ],
+        "fries": [
+            {"name": "Regular Fries", "sizes": {"Small": 2, "Medium": 3, "Large": 4}},
+            {"name": "Curly Fries", "sizes": {"Small": 2.5, "Medium": 3.5, "Large": 4.5}},
+            {"name": "Sweet Potato Fries", "sizes": {"Small": 3, "Medium": 4, "Large": 5}},
+            {"name": "Waffle Fries", "sizes": {"Small": 3, "Medium": 4, "Large": 5}},
+            {"name": "Cajun Fries", "sizes": {"Small": 2.5, "Medium": 3.5, "Large": 4.5}},
+            {"name": "Chili Cheese Fries", "sizes": {"Small": 4, "Medium": 5, "Large": 6}},
+            {"name": "Garlic Fries", "sizes": {"Small": 3, "Medium": 4, "Large": 5}},
+            {"name": "Truffle Fries", "sizes": {"Small": 4.5, "Medium": 5.5, "Large": 6.5}},
+            {"name": "Loaded Fries", "sizes": {"Small": 5, "Medium": 6, "Large": 7}},
+            {"name": "Cheese Fries", "sizes": {"Small": 3.5, "Medium": 4.5, "Large": 5.5}}
+        ],
+        "wedges": [
+            {"name": "Potato Wedges", "sizes": {"Small": 3, "Medium": 4, "Large": 5}},
+            {"name": "Cajun Wedges", "sizes": {"Small": 3.5, "Medium": 4.5, "Large": 5.5}},
+            {"name": "Garlic Parmesan Wedges", "sizes": {"Small": 4, "Medium": 5, "Large": 6}},
+            {"name": "Buffalo Wedges", "sizes": {"Small": 4, "Medium": 5, "Large": 6}},
+            {"name": "Loaded Potato Wedges", "sizes": {"Small": 5, "Medium": 6, "Large": 7}},
+            {"name": "Cheesy Wedges", "sizes": {"Small": 3.5, "Medium": 4.5, "Large": 5.5}},
+            {"name": "BBQ Wedges", "sizes": {"Small": 4, "Medium": 5, "Large": 6}},
+            {"name": "Spicy Wedges", "sizes": {"Small": 3.5, "Medium": 4.5, "Large": 5.5}},
+            {"name": "Truffle Wedges", "sizes": {"Small": 4.5, "Medium": 5.5, "Large": 6.5}},
+            {"name": "Herb Wedges", "sizes": {"Small": 3.5, "Medium": 4.5, "Large": 5.5}}
+        ],
+        "drinks": [
+            {"name": "Coca-Cola", "size": "Medium", "price": 2},
+            {"name": "Pepsi", "size": "Medium", "price": 2},
+            {"name": "Sprite", "size": "Medium", "price": 2},
+            {"name": "Fanta Orange", "size": "Medium", "price": 2},
+            {"name": "Dr. Pepper", "size": "Medium", "price": 2},
+            {"name": "Mountain Dew", "size": "Medium", "price": 2},
+            {"name": "Root Beer", "size": "Medium", "price": 2},
+            {"name": "Iced Tea", "size": "Medium", "price": 2},
+            {"name": "Lemonade", "size": "Medium", "price": 2},
+            {"name": "Fruit Punch", "size": "Medium", "price": 2}
+        ]
+    }
+    return menu
+
+
+# list of all add ons for the items
+@tool
+def get_add_ons():
+    """all possible add ons """
+    add_ons = [
+        {"name": "Extra Cheese", "cost": "1"},
+        {"name": "Lettuce", "cost": "1"},
+        {"name": "Tomatoes", "cost": "1"},
+        {"name": "Bacon", "cost": "1.5"},
+        {"name": "Avocado", "cost": "2"},
+        {"name": "Jalapenos", "cost": "0.5"},
+        {"name": "Extra Patty", "cost": "3"},
+        {"name": "Guacamole", "cost": "2"},
+        {"name": "Pickles", "cost": "0.5"},
+        {"name": "Onions", "cost": "1"},
+        {"name": "Mushrooms", "cost": "1.5"},
+        {"name": "BBQ Sauce", "cost": "0.5"},
+        {"name": "Ranch Dressing", "cost": "0.5"},
+        {"name": "Honey Mustard", "cost": "0.5"},
+        {"name": "Chipotle Sauce", "cost": "0.5"},
+    ]
+    return add_ons
+
+
+# tool to help define what combos are possible
+@tool
+def get_combos(main_item: str, side: str = None, drink: str = None):
+    """Create a meal combo of the user's choice. The user can select a burger or wrap from the menu,
+    along with an optional drink and either fries or wedges with that as well, which is also optional."""
+    combos = {
+        "Main with a Side": {"main_item": main_item, "side": side, "price": 10},
+        "Main with a Drink": {"main_item": main_item, "drink": drink, "price": 11},
+        "Main with Side and Drink": {"main_item": main_item, "side": side, "drink": drink, "price": 12}
+    }
+    return combos
+
+
+class OrderItem(BaseModel):
+    item: str = Field(description="the item user ordered from the menu")
+    price: float = Field(description="price of the item")
+    delete: bool = Field(description="Whether the user want to delete this item")
+    quantity: int = Field(description="the amount of items ordered")
+    add_on: str = Field(description="the add on item ordered from the menu")
+    price_add_on: int = Field(description="the price of the add on items")
+    combined_cost: int = Field(description="the cost of the order including the items and the odd ons")
+
+
+# to store the current order
+my_order: List[OrderItem] = []
+
+
+@tool
+def update_order(new_order: OrderItem):
+    """This function needs to be called when the user order something, or update their order, or delete something"""
+    global my_order
+
+    if new_order.delete:
+        # Remove the item from the order
+        my_order = [item for item in my_order if not (item.item == new_order.item and item.add_on == new_order.add_on)]
+    else:
+        item_found = False
+        for item in my_order:
+            if item.item == new_order.item and item.add_on == new_order.add_on:
+                item_found = True
+                if new_order.quantity > 0:
+                    item.quantity = new_order.quantity
+                    item.price = new_order.price
+                    item.price_add_on = new_order.price_add_on
+                    item.combined_cost = item.price + item.price_add_on
+                else:
+                    my_order.remove(item)
+                break
+
+        if not item_found and new_order.quantity > 0:
+            my_order.append(new_order)
+
+    print(f"Updated order: {my_order}")
+
+
+@tool
+def get_final_price():
+    """ Get the final price of the order """
+    global my_order
+    total_price = sum(item.combined_cost * item.quantity for item in my_order)
+    return total_price
+
+
+# Function to get breakdown of the current order
+@tool
+def get_breakdown():
+    """ function to get the cost breakdown of the order """
+    global my_order
+    to_return = ""
+    total = 0
+    for item in my_order:
+        to_return += f'{item.item} with {item.add_on}: cost {item.combined_cost}\n'
+        total += item.combined_cost
+    to_return += f'total order cost is {total}'
+    return to_return
+
+
+tools = [get_menu, update_order, get_final_price, get_add_ons, get_combos, get_breakdown]
+
+
+prompt = ChatPromptTemplate.from_messages(
+    [
+        (
+            "system",
+            "You are a helpful assistant for a fast food ordering service. Answer all questions to the best of your ability."
+            "You are a voice bot, so only return sentences without formatting. ",
+        ),
+        MessagesPlaceholder("chat_history"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
+)
+
+chain = prompt | llm
+
+
+MEMORY_HISTORY_STORE = {}
+
+
+# storing chat history
+def get_session_history(session_id: str) -> InMemoryChatMessageHistory:
+    if session_id not in MEMORY_HISTORY_STORE:
+        MEMORY_HISTORY_STORE[session_id] = InMemoryChatMessageHistory()
+    return MEMORY_HISTORY_STORE[session_id]
+
+
+agent = create_tool_calling_agent(llm, tools, prompt)
+
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
+CHAIN = RunnableWithMessageHistory(
+    agent_executor,
+    get_session_history,
+    input_messages_key="input",
+    history_messages_key="chat_history",
+    output_messages_key="output",
+)
+
+
+if __name__ == "__main__":
+    while True:
+        input_text = input("input: ")
+        output = CHAIN.invoke({"input": input_text}, config={'configurable': {'session_id': 'foo'}})["output"]
+        print(output)