Infini Connect

Description

🤖💬 AzureOpenAI integrated Devoteam GPT App based on Nest TypeScript framework. Started as bot project turned into full fledge server side application.

Use case

• Get available desks only in Amsterdam office, based on from and to date and time (respond with full name like "Bar table #6", "Desk #1 - Dual Monitor,Desk #10")
• Get the employees professional work experience details based on a given name or certificate or skill (Work in progress)
• Make desk reservation for Amsterdam office, based on timeslot, morning, afternoon or full/all day (respond with full name like "Bar table #6", "Desk #1 - Dual Monitor,Desk #10")
• HR related questions, From Employee Handbook (Work in progress)

Example

Introduction

Employee Search

Desk Reservation

Architecture

Installation

$ yarn install

Running the app

# development
$ yarn run start

# watch mode
$ yarn run start:dev

# production mode
$ yarn run start:prod

Endpoint

http://localhost:8080/api http://localhost:8080/api/auth/google

Azure PG and OpenAI

az cognitiveservices account create \ 
--name DvtBrainOAI \
--location swedencentral \
--sku s0 \
--kind OpenAI \
--resource-group rg-dvt-parking-app

# Endpoint
az cognitiveservices account show --name DvtBrainOAI \
--resource-group rg-dvt-parking-app | jq -r .properties.endpoint

# Key
az cognitiveservices account keys list --name DvtBrainOAI \
--resource-group rg-dvt-parking-app | jq -r .key1

# Deploy embedding-ada-002 model and GPT4
az cognitiveservices account deployment create \
--name DvtBrainOAI \
--resource-group rg-dvt-parking-app \
--deployment-name text-embedding-ada-002 \
--model-name text-embedding-ada-002 \
--model-version "2" --model-format OpenAI \
--sku-capacity "100" --sku-name "Standard"

az cognitiveservices account deployment create \
--name DvtBrainOAI \
--resource-group rg-dvt-parking-app \
--deployment-name gpt-4-32k \
--model-name gpt-4-32k \
--model-version "0613" --model-format OpenAI \
--sku-capacity "80" --sku-name "Standard"



# Create Cosmos PostgreSQL Version 15 DB with lowest SKU possible
az postgres server create \
--resource-group rg-dvt-parking-app \
--name dvt-parking-pg-db \
--location europewest \
--admin-user dvtadmin \
--admin-password dvtadmin1234 \
--sku-name B_Gen5_1 \
--version 15



# azure cognitive search service create
az search service create \
--name dvt-parking-search \
--resource-group rg-dvt-parking-app \
--location westeurope \
--sku Basic \
--replica-count 1 \
--partition-count 1

# azure cognitive search service delete
az search service delete \
--name dvt-parking-search \
--resource-group rg-dvt-parking-app


# Retrieve AZURE_SEARCH_ADMIN_KEY
az search admin-key show \
--service-name dvt-parking-search \
--resource-group rg-dvt-parking-app
 
 # Retrieve  AZURE_SEARCH_ENDPOINT
az search show \
--service-name dvt-parking-search \
--resource-group rg-dvt-parking-app | jq -r .endpoint

# Create WebApp Plan for NodeJS App
az appservice plan create \
--resource-group rg-dvt-parking-app \
--name dvt-devo-gpt-app-plan \
--sku B1 \
--is-linux


# Create WebApp for nodejs application
az webapp create \
--resource-group rg-dvt-parking-app \
--plan dvt-devo-gpt-app-plan \
--name dvt-devo-gpt-app \
--runtime "NODE|18-lts" \
--deployment-local-git

Test

# unit tests
$ yarn run test

# e2e tests
$ yarn run test:e2e

# test coverage
$ yarn run test:cov

Features of app

Technical

• AzureOpenAI Embeddings and GPT4 integrated
• Demonstrated cutting edge function calling feature or GPT model
• Exposes endpoint for Google Chat bot
• Google Drive integration
• Azure cognitive search integration
• OAuth2 implementation, requires for communication with Google APIs
• JWT token (User session management)
• Modulerarized code
• DB repository configured
• TypeScript enabled
• Nest framework to support MVC patter
• Example on how auotmation script can talk to AWS or Google APIs.

Business

• Get or reserve(WIP) desk in Joan Desk via google chat.
• Search for employee with specific skill or certificate
• Summarize employees' professional experience
• WIP: Answer HR related questions from Employee Handbook

ToDO

Technical

• CICD, Automated infrastructure and deployment pipeline
• Development experience to Dev
• Introduce similar application in Python or GoLang, as DevOps usually prefers that over Node.
• Code enhancements
• Move infrastructure to private DB connection
• Change email sender from Hardik to no-reply
• store chat history for follow up and start new thread after few messages
• build correct inventory that Bot can understand today,tomorrow

Business

• Chat integration for users to manage their parking
• Extend functionality to seat occupation
• Allow user to reserve parking from chat
• timeslot for parking with two slots

Cons

• Chat messages could be expensive, we need to launch on trial basis, per individual
• Lack of JavaScript/TypeScript/NodeJS skill (same draw back as backstage.io)
• Lack of development experience within DevOps
• GPT model hallucinates a lot

License

Nest is MIT licensed.