# NovaChat Platform

A production-grade AI chatbot platform with live web search, streaming responses, and a modern UI.

## Folder Structure
- backend
  - src
    - ai (routing, memory, context compression)
    - search (search, extraction, ranking, cache)
    - utils (LLM streaming client)
- frontend
  - src (React UI)

## Requirements
- Node.js 20+
- npm 10+

## Backend Setup
```bash
cd backend
npm install
npm run build
set LLM_API_KEY=YOUR_KEY_HERE
set LLM_MODEL=gpt-4o-mini
set LLM_API_BASE=https://api.openai.com
npm start
```

## Frontend Setup
```bash
cd frontend
npm install
npm run dev
```

The frontend is proxied to `http://localhost:8080` for API calls.

## Docker Setup (Compose)
```bash
set LLM_API_KEY=YOUR_KEY_HERE
set LLM_MODEL=gpt-4o-mini
set LLM_API_BASE=https://api.openai.com
docker compose up --build
```

Open `http://localhost:4173`.

## Performance Notes
- SSE streaming for token-by-token delivery.
- Parallel search fetch + content extraction.
- LRU caches for search results and conversation memory.
- Minimal sync blocking in the request path.

## Environment Variables
- `LLM_API_KEY`: API key for OpenAI-compatible endpoint
- `LLM_API_BASE`: Base URL for the OpenAI-compatible API (default: `https://api.openai.com`)
- `LLM_MODEL`: Model id (default: `gpt-4o-mini`)
- `PORT`: Backend port (default: `8080`)