aranlucas / resume-chat

Chat about my resume

Resume Chat

This application was created during my recent hiring experience. I have a separate repository where I use LangChain to load my resume and a question bank of interview questions which then get fed to a vector database (PineCone). For each request, we do a "similarity" search on the DB, then feed that as context to ChatGPT and answers the question.

This should be pretty close to what I would answer if I had a good vector database inside my head.

Getting Started

First, run the development server:

npm run dev
# or
yarn dev
# or
pnpm dev

Open http://localhost:3000 with your browser to see the result.

You can start editing the page by modifying app/page.tsx. The page auto-updates as you edit the file.

This project uses next/font to automatically optimize and load Inter, a custom Google Font.

FAQ

How does it work?

There is 2 parts to this project

1. The loader
2. Chat API

The loader (configured in a separate repo) works the following way

sequenceDiagram
Script->>LangChain: Load all documents

LangChain->>LangChain: Read, Split, transform to embeddings

LangChain->>PineCone: Insert new embededdings into DB

PineCone->>Script: Return files in Index

Loading

Chat API

sequenceDiagram
API->>PineCone: Find the closest documents related to this question

PineCone->>API: returns K similar documents

API->>ChatGPT: With added context, answer the question

ChatGPT->>API: Return streaming response to client

Loading

resume-chat's Issues

Integrating Vercel KV

Hey Lucas!

This is a great repo! Have you tried integrating Vercel KV yet for chat history? I have been stuck on this exact thing after cloning down verel's template and its driving me mad!

Curious if you gave it a go, or if you have any pointers here, much appreciated