Making a demo

Over the past few weeks, we have worked on getting familiar with the Microsoft Graph API, the Webex API and the LangChain framework, and vibecoded a simple demo.

The process

To keep the code simple and get a demo out as soon as possible, we decided to keep the workflow simple: no agents, no LangChain. We decided to use LLama for the natural language processing (since we don't have any GPT tokens) and just have a Webex bot that can read the user's calendar but not able to book meetings.

We split the workload evenly between the four of us. Kavya focused on the Webex API, Bobby focused on creating a local server that runs Llama, and Fateh and Karl focused on the Microsoft Graph API.

The demo

After collaborating on the GitHub repo, we made a simple demo where the user can communicate with a bot on Webex and ask it a simple question like "What are my plans for next week?". The bot will then run a Llama instance on the user's computer (since we are keeping it simple and running the bot and model on localhost rather than on the cloud). The bot then understands what the user is asking for and sends an API request to Microsoft Graph, receives the response and outputs the response to the user in a nice format.

We had a meeting with Sunny and showed her the demo, which she found very impressive.

Future plans

Over the next couple of weeks, we will work on improving the code and adding more functionalities to it so that it can actually book meetings for users and save them time throughout the week.