DispatchBot

Inspiration

Incompetency in 911 calls which leads to delayed responses and uninformed agents on the scene

What it does

Gathers information about an emergency from a caller, redirects caller to a line for that department (triage), and asks questions specifically tailored to which department the caller has been redirected to and the given details of the incident

How we built it

Tech stack includes LangChain, GPT-3.5-Turbo, and OpenAI whisper, as well as various other libraries for utility functions

Challenges we ran into

Unfamiliarity with particular APIs and a lack of documentation for newer open-source projects

Accomplishments that we're proud of

Creating effective system prompt templates for the LLM

What we learned

Working with LLMs and various new APIs

What's next for DispatchBot

Building more capable voice features as well as giving the model the ability to recognize commonly-asked questions

Built With

• gpt
• langchain
• openai
• whisper