Cafe order bot

restaurant chatbot agent
lambda function
connection to the endpoint
launch darkly model testing

Inspiration

Inspired by the struggles of busy San Francisco restaurant owners facing staffing shortages, we envisioned a solution that empowers them to do more with less. This project is our answer – an AI-powered ordering system using Agents that streamlines operations and frees up valuable time for restaurant teams.

What it does

Key functionalities:

Integrated Ordering with AI Assistant: Utilizes Agents built using Amazon Bedrock for a chatbot based ordering experience. This AI assistant leverages the power of Anthropic Claude LLM to understand customer requests. Customers can ask questions and receive real-time guidance throughout the ordering process, ensuring informed choices.

TODO: Complete LaunchDarkly Integration: Enables A/B testing of mutliple LLM models to determine which model works best.

How we built it

Bedrock Function Calling Agents: We leverage Bedrock Agents to create intelligent agents capable of reading and understanding the customer requests and calling the appropriate functions.

Challenges we ran into

We did not have enough time to building a custom web app llamaindex code integrating with Bedrock.

Accomplishments that we're proud of

Built a Bedrock agent within 60 minutes that is capable of taking cafe orders which was very difficult if we have to do everything using Python

What we learned

This project served as a valuable learning experience for our team. We gained significant exposure to AWS Bedrock and its capabilities. Additionally, we identified the potential benefits of leveraging Agents in Bedrock. This knowledge will be instrumental in our ongoing development efforts.

What's next ?

LaunchDarkly Integration: This feature flagging platform plays a vital role in testing our different LLM and determine what works best for our usecase.