How to make chatbot stop the conversation and return a JSON with order

Short Course Q&A ChatGPT Prompt Engineering for Developers
1

The conversation is never ends, last message from chatbot is something like that:
You’re welcome! If you have any other questions or need further assistance, feel free to ask. Have a great day and enjoy your meal!

What is a good way to instruct ChatGPT to identify the end of a conversation and return a JSON with the order details?

2

To achieve this using methods within the prompt itself, follow these steps:

  1. Execute the first action.
  2. Perform the second action.
  3. Carry out the third action.
  4. Verify if the task is complete.
    • If it’s not complete, repeat steps 2 - 4.
    • If it’s complete, proceed to the next step.
  5. Respond in JSON format.

To instruct ChatGPT to identify the end of a conversation and return a JSON with order details, you must design a system to track the conversation’s context and extract the necessary details. Here’s a detailed step-by-step guide:

  1. Context Tracking: Implement a context tracking mechanism to keep track of the conversation’s state. This could be a variable or a flag that indicates whether the order process is complete. When the conversation starts, this flag could be set to False, and when the order is complete (for example, when the chatbot says, “Your order has been placed.”), the flag could be set to True.

  2. Order Detail Extraction: As the conversation progresses and the user provides details about their order, these details must be extracted and stored. This could be done using pattern matching, keyword spotting, or more advanced natural language understanding techniques. You must design your system to recognize and extract these details accurately.

  3. JSON Creation: Create a JSON object with the order details once the order process is complete. This could be as simple as a Python dictionary converted into a JSON object using json.dumps().

Here’s a simplified Python code example of how this might look:

import json

# Initialize conversation state and order details
conversation_ended = False
order_details = {}

# Simulate conversation
for message in a conversation:
    # Check if the conversation has ended
    if "enjoy your meal" in the message:
        conversation_ended = True

    # Extract order details (this is a simplification, in reality, you'd need more advanced techniques)
    if "order" in the message:
        items = message.split("order")[1].strip()
        order_details["items"] = items

# If the conversation has ended, create a JSON object with the order details
if conversation_ended:
    order_json = json.dumps(order_details)
    print(order_json)

Please note that this is a simplified example and doesn’t cover many real-world implementation complexities, such as handling multiple orders, dealing with changes to the order, or extracting more complex order details. However, it should give you a starting point for developing a system to handle these tasks.

1 Like
3

I tried adjusting a prompt, to summarize in a JSON format or to put a summary in a
but chatbot doesnt perform this instruction.
is there a better way to engineer a prompt? maybe there is an example of something similar?

context = [ {'role':'system', 'content':"""
You are development manager bot, your main task is to collect the current status of the project: "Building a new chatbot" \
You first greet the customer, then sync on project progress. \
Finally you summarise what customer said. \
Write your summary in <summary> tag
You respond in a short, very conversational friendly style. \
"""} ]