PoliSync - server

A WebSocket & Proxy server for PoliSync's MVP.

Client Repo     Server Repo

PoliSync is a voice recognition-based the minutes of a meeting writing platform using National Assembly Public Data & API and Generative AI (Open AI).

Getting Started

Install

npm install

Start

npm run start

Env

Check the API on the OpenAI API Docs (EN) and Fill out .env

PORT={{PORT}}

OPENAI_URL={{OpenAI API URL}}
OPENAI_API_KEY={{OpenAI Organazation API key}}
OPENAI_ORGANIZATION_ID={{OpenAI Organazation ID}}

Features

• Creating the minutes of a meeting
  • Voice-based Live chat using WebSocket & Web Speech API
  • User classification using querystring (no-database)
• Searching References based on voice recognition & Member of the National Assembly (National Assembly data)
  • Infinite Scrolls with IntersectionObserver API
• Meeting Summary using OpenAI API (Generative AI) with proxy server in the middle

Tech Stack

Client

React, TypeScript, @tanstack/react-query, Recoil, Mui , Axios, webSocket

Server

JavaScript, NodeJS, Express, openai, webSocket, nodemon

Service Architecture

• The React App Client sends and receives requests to Node.js/Express Server and the National Assembly API.
• Node.js/Express Server acts as a proxy server for OpenAI, and at the same time acts as a WebSocket server.
• The external APIs OpenAI and National Assembly APIs send responses to servers and clients, respectively.

How it Works

The following is a sequence diagram of the key features.

Creating the minutes of a meeting (Voice-based Live chat)

1. When the user says, Browser (Web Speech API) converts it into text.
2. Client sends the converted text to the WebSocket Server.
3. WebSocket Server exchanges data with users through real-time communication.

Searching References

1. User refers to a specific keyword
2. The Client App recognizes this keyword, extracts it, and sends a GET request to the National Assembly API.
3. The National Assembly API sends a response, and users can check this information.
4. Infinite scrolling allows the Client App to send additional GET requests, while the National Assembly API sends additional responses.

Generating meeting summary

1. Client sends a POST request to the Proxy Server with the meeting notes data.
2. The Proxy Server forwards this request to the OpenAI API, using the GPT-3.5-turbo model.
3. The response from the OpenAI API is received by the Proxy Server, which then processes and sends the data back to the Client.

Reasons for using the proxy server: Use a proxy server for enhanced security. The proxy server acts as a middleman between the client and the OpenAI API.

Demo of this project

1. Enter a conference room by entering your name on the main screen:

2. Press and speak on the voice recognition button at the bottom of the chat section to convert to text:

3. When words of certain keywords are found, reference documents are found through the data of the National Assembly:

4. When words of certain keywords are found, reference documents are found through the data of the National Assembly: