AI powered bike shop bot

Your Cosmic Concierge: AI-Powered Shop Assistance

Comment

Inspiration

As an Azure cloud consultant, I’ve witnessed the rapid evolution of technology. Over the past two years, AI has emerged as a transformative force, reshaping industries and enhancing user experiences. The allure of AI lies in its ability to learn, adapt, and make intelligent decisions. Whether it’s natural language processing, computer vision, or predictive analytics, AI has become an essential tool for innovation. As I embark on my AI journey, I’m excited to explore Azure’s AI services, experiment with generative models, and participate in hackathons. By understanding AI’s power, I hope to contribute to groundbreaking solutions and drive positive change.

What it does

This hackathon challage is to develop chat application (assistant agent) to get answers based on information stored in cosmos DB. It search information from cosmos DB and format it using Azure Open AI chat completion model.

How I built it

I used terraform cloud to deploy and configure azure infrastructure. I used GitHub for terraform code and VCS backed terraform workspace to initiate azure deployment. This helped me to quickly deploy resources and destroy (if required).

For web app , I used given public repository for website. Variables are used for backend configuration.

For Container app , I copied given repository and created new image in my container registry. I updated details from this repository for vector indexes, updated modules and small base image.

Post cosmos DB deployment , I used lab code (executed from laptop) to load and vectorise data.

Challenges we ran into

During this project , I ran into multiple challenges like

  1. Initially I uploaded data to cosmos DB with different vector index name. I got various errors as it was different from vector index name used in backend api.
  2. While uploading bulk data with vectors in cosmos db , I was getting exception due to timeout. I tried multiple options but was not able to use bulk write. So finally I changed logic to update and vectorise one record at a time. This is not best option but I was able to update all records.
  3. Web application code uses node js. I don't have any experience with node js, but I tried to change web application UI to add more options and suggestion. However it was failing so finally I gave up and used given repository for web application.

Accomplishments that we're proud of

From this hackathon , I learn

  1. To use azure free services effectively to reduce cost
  2. Container app deployment using image from container registry
  3. Docker image development with python
  4. The basic principles of generative AI and how can we develop intelligence app.

What's next for AI powered bike shop bot

This AI powered Bike shop bot uses fixed data (got from GitHub guide repo). I want to , update web app to have options to add/update/delete data. Also I might try to use Azure AI search to query data and then Open AI API for formating.

Built With

Share this project:

Updates