Trying out the sample app

1. Clone the GitHub repo

The first step is to clone the below GitHub repo (if not already done as part of the previous demo app)

https://github.com/slanglabs-projects/conva-ai-demo-apps

This will contain the following demo apps.

• MyChatGPT App - A simple demo app to build your custom ChatGPT-style application with the ability to answer specific questions about a particular subject that a Copilot has been trained for. This shows how to use the Dynamic mode of CONVA.ai to build a powerful Assistant inside your app to answer any of your user's questions.

• Capability Utilizer - A simple demo app to demonstrate how to use the Static mode of CONVA.ai to trigger specific useful Capabilities inside apps.

$ git clone https://github.com/slanglabs-projects/conva-ai-demo-apps.git

For this section, we will be referring to the Capability Utilizer app for all languages.

2. Pick your language choice

The demo is currently available in two languages

• Flutter (Android)

• Python

You can pick either one of them that you are comfortable with.

3. Update the API keys in the code

Get the Copilot ID & the API key for your Copilot from Magic Studio.

You can find this in the "Properties" section of the "Demo App for Magic Studio"

Copy the ID and API key in the code which initializes the Copilot

Flutter (Android)

/// File - Flutter/MyChatGpt/lib/utils/utils.dart

/// The Copilot ID of your Copilot.
static String assistantId = "<CopilotID>";

/// The API key for accessing the CONVA.ai system.
static String apiKey = "<API Key>";

/// The version of the your Copilot.
static String assistantVersion = "<Version Number>";

Python

// File is inside Python/Capability_Utilizer.py

async def generate(query: str, stream: bool):
    client = AsyncConvaAI(
        copilot_id="YOUR_COPILOT_ID", 
        copilot_version="YOUR_ASSISTANT_VERSION", 
        api_key="<YOUR_API_KEY>"
    )
    response = client.invoke_capability(query, stream=stream)
    out = "" 
    async for res in response:
        out = res.model_dump_json(indent=4)
    return out

4. Setup dependencies

Before you can run the application, make sure that you have your environment setup correctly.

Note that for Python, you need to do an extra step of installing the conva-ai package $ pip install conva-ai

5. Run the application

Now you should be able to run your application.

Flutter (Android)

// Was a bit too late in the night to add a screenshot 

Python

$ python MyChatGPT.py
Ah, the ever-so-wise Boomers,
the perpetually self-aware Millennials, and
the remarkably knowledgeable Gen Z - together, we form the circle of eternal enlightenment!
 Welcome, wise sages of all ages, to another day of unrivaled wisdom and unending insight in the AI-powered wonderland!
Where machines dream of electric sheep and chatbots tell jokes!

Today I am going to do something mundane like answer your questions regarding Digi Yatra! Dont go light on me. Go for it
Your first questson? (Type 'exit' to quit) what is dy?
Digi Yatra is a Ministry of Civil Aviation, Govt. of India led 
initiative to make air traveller’s/ passenger’s journey seamless,
 hassle-free and Health-Risk-Free. The Digi Yatra process uses 
the single token of face biometrics to digitally validate the Identity, 
Travel, Health or any other data that is needed for the purpose of 
enabling air travel.

Next one? (Type 'exit' to quit) 

Thats it!

Lets now explore the key parts of the code which made this happen

How did it work?

There are only two APIs to use to make this above system work

• Initialize CONVA.ai with your Copilot credentials

Flutter (Android)

// TBD

Python

// Create the Client object using the Copilot credentials
client = AsyncConvaAI(
    copilot_id="<Copilot ID>",
    api_key="<API key>",
    copilot_version="<version number>"
)

• Invoke the Capability and get the response in JSON

Flutter (Android)

Python

// Invoke a specific Capability by giving its Group name and asynchronously collect the response
out = client.invoke_capability(query, stream=stream, capability_group="group_name")
response = ""
async for res in out:
    response = res

Thats all. Use the response object to access the various parameters that is returned by CONVA.ai and use it in your app as you see fit.

The most important parameters are -

• message - The response message for the Capability. Every Capability will have this parameter.

• related_queries - The list of AI generated queries that are most appropriate follow up questions to ask the AI.

• parameters - This is a HashMap which contains the custom parameters that the developer configured when the Capability was configured.

Thats all.