ComplaintSense: Ruby Track Hackathon for Headstarter AI
This project is built for the Ruby Track of the Headstarter AI Hackathon. The project demonstrates skills in AI Engineering and Full Stack Development by building a multi-modal complaint management system for Ruby, a financial technology company. The system leverages AI to analyze and categorize customer complaints, integrates with a relational database, and employs a Retrieval-Augmented Generation (RAG) pipeline with a vector database for advanced complaint retrieval.
Task: Use an LLM API to determine if a call is a complaint and create a summary of the complaint.
Implementation:
Task: Assign a product category and a sub-product category similar to the sample data to the new complaint and save it to the database of complaints.
Implementation:
Task: Build a RAG pipeline using a vector database. Given a voice recording, find the most relevant complaints based on what is said in the voice recording.
Implementation:
Task: Make the inputs multi-modal. Handle voice, text, video, and text+picture inputs to identify and categorize complaints.
Implementation:
npm install
..env.local.example
file and rename it to .env.local
. Fill in the environment variables with your Firebase project configuration.npm run dev
.To generate a new migration, run the following command:
npm run db:generate
To run the migrations, use the following command:
npm run db:migrate
To check the consistency of the migrations, run the following command:
npm run db:check
For more information, see the Drizzle Kit documentation.