AI Support Copilot

Upload any file and chat with AI about it — like ChatGPT, but every answer is grounded in your upload with source citations.

Features

Capability	Description
File upload	PDF, text, JSON, Markdown, CSV, HTML at /chat
AI chat	ChatGPT-style streaming chat grounded in your files
Hybrid RAG	Vector + BM25 keyword search
Source citations	Every answer links back to your uploaded file
Usage dashboard	Estimated OpenAI cost at /usage — tracked locally

Routes

Route	Description
/	Landing page
/chat	Upload a file and chat with AI Support Copilot
/usage	Local usage & cost dashboard
/docs	Product guide

Getting started

npm install
cp .env.example .env.local
# Set OPENAI_API_KEY in .env.local
npm run dev

Open http://localhost:3010

Architecture

• Next.js — landing, chat UI, usage dashboard, and API routes
• POST /api/chat — SSE streaming; embeddings + chat via OpenAI
• POST /api/demo/parse — stateless PDF text extraction
• Browser storage — documents (sessionStorage), usage dashboard (IndexedDB)
• Server token budget — 100K trial enforced per device/network (data/guest-usage/)

Deploy

Works on Vercel (or any Node host that runs Next.js). Set OPENAI_API_KEY in project environment variables. No database or separate API server required.

Environment

Variable	Required	Description
OPENAI_API_KEY	Yes	OpenAI API key
OPENAI_CHAT_MODEL	No	Default gpt-4o-mini
OPENAI_EMBEDDING_MODEL	No	Default text-embedding-3-small
USD_TO_INR	No	INR display in usage dashboard

License

MIT — see LICENSE.