Build a Custom ChatGPT-like Chatbot with Chainlit

This tutorial will guide you through building a custom chatbot using Chainlit and AskYoda, providing a powerful and customizable conversational AI tool for various applications 🙌

‍ ‍

Advantages of Eden AI and Chainlit for AI Chatbot Development

AskYoda and Chainlit offer significant advantages for those looking to create and customize AI chatbots.

Chainlit, an open-source Python framework, provides the capability to develop Conversation AI interfaces with ease, allowing for customization through various providers.

On the other hand, AskYoda, developed by Eden AI, empowers users to craft and train AI chatbots using their specific data.

AskYoda’s integration with providers such as OpenAI, Cohere, and Google Cloud expands its functionality, addressing limitations by enabling seamless data integration and training in multiple programming languages.‍

How to build a custom chatbot using ChainLit?‍

Step 1. Prerequisites

💡 Make sure you have the following libraries installed: chainlit, requests, json, dotenv, os

import chainlit as cl 
import requests 
import json 
from dotenv 
import load_dotenv 
import os 
from typing import Optional 
from chainlit.types import ThreadDict 
from chainlit.input_widget import Select, Slider‍

Step 2. Decorator and Chat Settings

Chainlit works with decorators for handling interactions, so you will need to use it for each of the interaction you want to handle. Here’s an example decorator, followed by the function that will be triggered when a new message is posted.

@cl.on_message #The decorator 
async def main(message: cl.Message): 
    await cl.Avatar( 
        name="AskYoda", 
        path = "public/logo_light.png" 
        ).send()‍

You can also customize chat settings using cl.ChatSettings in the triggered function:

cl.ChatSettings( 
        [ 
            Select( 
               id="llm_provider", 
               label="LLM Provider", 
               values=["cohere", "amazon", "anthropic", "ai21labs", "openai", "google"], 
               initial_index = 4 
               ), 
             Select( 
               id="llm_model", 
               label="LLM Model", 
               values= [ 
                  "command", 
                  "command-light" 
)])‍

Step 3. Retrieve Chat Settings

You can retrieve the selected chat settings like this:

llm_model = cl.user_session.get("llm_model") 
llm_provider = cl.user_session.get("llm_provider")‍

Step 4. Create a project in AskYoda

‍

1. Create an Account

To begin utilizing AskYoda, the first step is to create a free account on Eden AI.

Once registered, you can obtain your API key 🔑 directly from the homepage. This key can then be used with the complimentary credits provided by Eden AI.

G‍et your API key for FREE.

2. Create a new project

Once logged in, navigate to the AskYoda feature. Start by creating a new project within AskYoda.

3. Upload Your Data‍

Upload the data you want to train your chatbot on.

4. Chat with Your Data

Once your project is set up, you can start chatting on the interface using your own data:‍

If you want to go further check our article on our website

‍

Step 5. Make API Calls to AskYoda

Before making API calls, ensure you have the bearer token and project ID ready:

headers = {"Authorization": f"Bearer {Api_Key}"} 
url = f"https://api.edenai.run/v2/aiproducts/askyoda/{project_id}/ask_llm"‍

Then, gather all the necessary information and make the API call to Eden AI:

json_payload = {"query" : f"{message.content}", 

            "llm_provider" : llm_provider, 

            "history" : history, 

            "llm_model" : llm_model, 

            #Customize the personnality of your assistant 

            "personnality" : {}, 

            "chunks" : chunks 

            } 

    response = requests.post(url, json=json_payload, headers=headers) 

    result = json.loads(response.text) 

if 'result' in result: 
        a = str(result['result']) 
        history.append({"user": f"{message.content}", "assistant": f"{a}"}) 
        # Send a response back to the user 
        await cl.Message( 
              content=f"{result['result']}", author = 'AskYoda' 

        ).send() 
else: 
        await cl.Message( 
            content=f"Key 'result' not found in the response. : {result}", 
        ).send()‍



    

    

Step 6. Run the Chat Interface

To run your custom ChatGPT interface with AskYoda using Chainlit, type:
chainlit run app.py

Congratulations! 😎 You’ve created your own AskYoda chatbot interface using Chainlit!

You can now use it with the LLM of your choice from a diverse selection of providers on Eden AI, including OpenAI, Google, Cohere, Anthropic, AI21 Labs, Amazon, and many others!

There are still opportunities for further improvements such as authentication, chat history management, and adding more features as needed.

Explore more possibilities with Chainlit and AskYoda for creating powerful AI conversational agents tailored to your needs.‍

7. Embed your chat on your app/product

The chat is deployable in Copilot. Learn more about Copilot deployment here.

The Copilot is tailored to assist users in maximizing their app experience by offering contextual guidance and performing actions on their behalf:

Check it out!

Create your Account on Eden AI.