This is a Next.js project bootstrapped with create-next-app. It is a function AI assistant with a barebones UI. Please read the TO-DO list below to review limitations!

Deployed with Vercel

Visit: https://ama-app-nu.vercel.app/

Running locally

To run locally the development server, you must provide your own OpenAI API key in a .env file. Add your key to .env as OPENAI_API_KEY="..."

npm run dev

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

Tools Used

• Frameworks

  • Next.js v16 + App Router
  • React v19

• AI

  • Vercel AI SDK - Extensive toolkit for supporting various AI SDK Providers.
  • OpenAI - Provider, language model support for OpenAI API's (completion, responses, chat, embeddings).
    • Model is set in the chat API route (api/chat/route.js).

• Styling

  • Tailwind CSS v4
  • next/font to automatically optimize and load Jost

• Database/object storage

  • Vercel Postgres - for storing chat metadata
  • Neon + pgvector - for Vector embedding storage/persistence.

Notes:

State of the app within 72 hours of creation

• General chatbot functionality implemented.
• General UX/styling implemented.
• Simple tool calls enabled - a simple, first-round implementation for simple tool calling is enabled, not too useful at the moment (fetches weather data, performs math conversions and formats output).
• Enabled multi-step tool calls.
• Set up Postgres database, Neon + pgvector extensions, set up a Vercel Blob store for raw object storage, installed necessary packages.

To-dos:

implement data pipelines for:

1: User file management - uploading:

• pipeline trigger: user uploads a file -> client-side data validation -> server-side data validation -> metadata generation -> db lookup -> db transaction -> blob storage -> response generation -> response processing -> app state update.

2: New file data ingestion:

• pipeline trigger: user selects file, taps button to trigger ingestion -> db/blob store validation (db metadata entry lookup + blob store lookup )
  -> blob store fetch -> file processing (server-side data validation -> nlp -> tool calls (not in 1st proto) -> chunking -> embedding generation -> metadata generation -> db transation -> response generation -> response processing -> app state update.

Postgres business logic

Blob Store interaction logic

file upload UI/UX