Example Chat Agent

Example: ChatAgent

ChatAgent is an implementation of GrocAgent designed to handle chat-based queries. It demonstrates how to use GrocAgent to create a chatbot capable of processing user queries and generating responses using an external model (e.g., Modellake). Note: While this example uses Modellake to implement chatbot functionality, users can extend GrocAgent to incorporate other libraries or custom logic based on their project requirements.

Imports

Ensure the following imports are included when using GrocAgent:

from flask import Flask  # Flask is used as the web server framework
from groclake.modellake import Modellake  # (Optional) For handling chatbot logic
from groclake.utillake import GrocAgent  # Base class to define custom agents

Implementation Details

The ChatAgent example is initialized with:

1. Flask App Instance: Registers the agent with a Flask application.

2. Agent Name: Identifies the agent (e.g., "ChatAgent").

3. Initial Intent and Description: Defines the default intent ("chat") and its purpose.

4. Adaptor Configuration: Links external dependencies through configurations.

class ChatAgent(GrocAgent):
    def __init__(self, app, agent_name, initial_intent=None, intent_description=None, adaptor_config=None):
        super().__init__(app, agent_name, initial_intent, intent_description, self.default_handler, adaptor_config)

Default Handler

The default_handler method processes incoming payloads to:

• Extract key components: query_text, intent, entities, and metadata.

• Construct a chat payload for the Modellake API.

• Retrieve and return a response from the model, with error handling for fallback responses.

def default_handler(self, payload):
    try:
        query_text = payload.get('query_text', 'No query provided')
        intent = payload.get('intent', 'default_chat')
        entities = payload.get('entities', [])
        metadata = payload.get('metadata', {})

        chat_payload = {
            "messages": [
                {"role": "system", "content": "You are a helpful assistant. Only answer technical questions."},
                {"role": "user", "content": query_text}
            ],
            "token_size": 500
        }

        chat_response = Modellake().chat_complete(chat_payload)
        response_text = chat_response.get("answer", "I'm not sure how to help with that.")

        return {
            "response_text": response_text,
            "intent": intent,
            "entities": entities,
            "metadata": metadata,
            "query_text": query_text
        }
    except Exception as e:
        return {
            "response_text": f"Error: {str(e)}",
            "intent": payload.get('intent', 'default_chat'),
            "metadata": payload.get('metadata', {}),
            "query_text": payload.get('query_text', ''),
            "status": 500
        }

---

Request and Response Example

Request Example

{
  "groc_account_id": "",
  "header": {
    "version": "1.0",
    "message": "Request",
    "Content-Type": "application/json",
    "auth_token": "Authentication-Token",
    "apc_id": "c9d1c9b9-9f1b-4430-9bd4-6994dc1c89ee",
    "server_agent_uuid": "075fcb88-f25a-4390-a37e-e374a3c2b1df",
    "client_agent_uuid": "Client-Agent-UUID"
  },
  "body": {
    "query_text": "What is AI?",
    "intent": "chat",
    "entities": [{"ticket_id": "123314141"}],
    "metadata": {
      "context": "query generated from a support chatbot",
      "customer_id": "895985133"
    }
  }
}

Response Example

{
  "body": {
    "entities": [{"ticket_id": "123314141"}],
    "intent": "chat",
    "metadata": {
      "context": "query generated from a support chatbot",
      "customer_id": "895985133"
    },
    "query_text": "What is AI?",
    "response_text": "AI, or Artificial Intelligence, refers to the simulation of human intelligence processes by machines..."
  },
  "header": {
    "apc_id": "c9d1c9b9-9f1b-4430-9bd4-6994dc1c89ee",
    "auth_token": "Authentication-Token",
    "client_agent_uuid": "Client-Agent-UUID",
    "content-type": "application/json",
    "message": "response",
    "version": "1.0"
  }
}

---