AI ChatBot

Started on January 3, 2025

Status: In-progress

Democratize LLM's and access all through single application

Idea

Single Chatbot that help users to chat with best AI providers, such as OpenAI, DeepSeek, Claude, Mistral and Meta.

What we need

  • Xcode
  • AWS
  • API Keys
  • Auth setup
  • Dynamo DB
  • API Gateway
  • Lambda
AI Chatbot Logo

Website

chatwithai.us

Journal

02.27.2025

Defining high level requirements

  • The app should allow users to use LLM's from of AI providers, like
    • GPT 3.5
    • GPT 4.0
    • GPT O
    • DALLE
    • Claude
    • Grok
    • Deep Seek (In future)
  • Able to store all the history of chats.
  • Customized chat bots for each use case like content creation, email writing, solving homework all categorized for a quick selection.
  • Ability to bookmark such bots for quick access
  • We will be using Supabase for auth, storage and realtime sync

I have earlier implemented everything on AWS , but now switching to Supabase, due to the need to include web

Supabase is extermely efficient for quick prototype.

Products

  • Web - Next JS + Supabase + vercel
  • Mobile (iOS) - Xcode, iOS , Supabase
API Gateway Latency Stats
02.07.2025

Defining high level requirements

  • The app should allow users to use LLM's from of AI providers, like
    • GPT 3.5
    • GPT 4.0
    • GPT O
    • DALLE
    • Claude
    • Grok
    • Deep Seek (In future)
  • Able to store all the history of chats.
  • Customized chat bots for each use case like content creation, email writing, solving homework all categorized for a quick selection.
  • Ability to bookmark such bots for quick access

Game Plan

  • Use AWS Amplify for Authenticating users, through various methods

    1. Create an account directly with us
    2. Use google to authenticate

  • Once Authenticated, users can access free resources, like GPT 3.5 for few tasks with certain limits, Pro plans will have unlimited access to all other models.

  • Setup AWS Gateway with Post method that can take a payload and send it a lambda only for AWS IAM users, that are authenticated earlier.

  • Lambda will use different API keys from Secret Manager, to run the API calls with different providers and share the response back to the API Gateway.

  • To reduce Lambda spin up time (latency), we will trigger lambda every 5 mins using AWS Event Bridge.

API Gateway Latency Stats

  • Extract user information such as email/ unique ID from Cognito , along with the purchase plan information from Revenue Cat and store in AWS Dynamo DB depending on the type of user.

  • When the API call is made from the client, ensure Lambda checks the user requests against the limits and reject additional API calls.

Design

Image 1
Home Screen
Image 2
New Chat
Image 3
Chat History
Image 4
Request to Bot
Image 5
Response from Bot