BAYES the Shopping Assistant

Inspiration

Speaking and seeing is so much by typing and clicking. I believe we will usher in a new era of sales interactions, where having an agent speaking about and showing a product will be the dominant way for personalized online shopping.

What it does

Our AI agent hears the user's request, identifies the best products with the latest information, and helps the user pick the best product for them.

How we built it

I took a multimodal-first approach, prioritizing voice-to-voice interactions, while adding additional features such as knowledge retrieval and website browsing.

Challenges we ran into

Hooking everything together, overcoming CORS issues while trying to build fast.

Accomplishments that we're proud of

A complete pipeline that ingests PDFs into our Pinecone vector DB, and a frontend that retrieves the knowledge and acts on it.

What we learned

More tools have emerged that make building agent-powered web apps easier than ever.

What's next for BAYES the Sales Agent

Fleshing out PMF and seeking feedback from potential funding partners.

Built With

• llama-index
• openai
• pinecone
• python
• toolhouse
• typescript