Findatalab Frontend

Frontend for Findatalab projects, built with Next.js (App Router).

Stack

• Next.js 15
• React 19
• App Router API routes (app/api/**)
• Markdown rendering (marked) + sanitization (isomorphic-dompurify)

Features

• / — project landing page
• /fingpt — chat UI for admissions assistant
• /llmcity — news sentiment analysis + generated comments
• Server-side proxy routes for LLM calls:
  • /api/chat
  • /api/llmcity/chat

Prerequisites

• Node.js 20+
• npm 10+
• Running LLM backend (for local usage), e.g. Ollama or compatible API

Quick Start

1. Install dependencies:

npm install

2. Create local env file:

cp .env.example .env.local

If .env.example does not exist yet, create .env.local manually from the section below.

3. Start development server:

npm run dev

4. Open app:

• http://localhost:3000

If port 3000 is busy, Next.js may use 3001 automatically.

Environment Variables

Use .env.local for local development.

# Common chat endpoint (used by /api/chat as fallback)
CHAT_ENDPOINT=http://localhost:1416/chat/completions

# LLM model for /fingpt requests
NEXT_PUBLIC_CHAT_MODEL=

# Optional fixed chat history id in /fingpt
NEXT_PUBLIC_CHAT_HISTORY_ID=

# LLM City specific endpoint/model (optional)
LLMCITY_ENDPOINT=http://localhost:11434/api/chat
LLMCITY_MODEL=qwen3.5

Important

• Do not put secrets into NEXT_PUBLIC_* variables.
• Browser pages call internal routes (/api/...), and server routes call upstream LLM APIs.

Available Scripts

• npm run dev — start development server
• npm run build — create production build
• npm run start — start production server
• npm run lint — run lint checks

Production Run (PM2)

Build first:

npm run build

Start with PM2 on custom port:

pm2 start npm --name "findatalab-frontend" -- run start -- -p 33000

Check status/logs:

pm2 status
pm2 logs findatalab-frontend

API Flow

/fingpt

Client page sends request to /api/chat → server route forwards to CHAT_ENDPOINT.

/llmcity

Client page sends { news, commentCount } to /api/llmcity/chat → server route builds prompt, calls LLM endpoint, parses model response, returns structured JSON:

{
	"sentiment": {
		"label": "positive|negative|neutral|mixed",
		"score": 0.32,
		"explanation": "..."
	},
	"comments": ["...", "..."]
}

Troubleshooting

POST /api/llmcity/chat 404

• Restart dev server (npm run dev) after adding/changing route files.
• If needed, clear cache and restart:

rm -rf .next && npm run dev

Unexpected token '<' ... is not valid JSON

• Usually means HTML was returned instead of JSON (wrong URL, 404 page, proxy issue).
• Verify frontend is calling /api/llmcity/chat on the same host/port as the app.

502 from /api/llmcity/chat

• Upstream LLM responded in unexpected format or could not be parsed.
• Check server logs for [llmcity/chat] response_parse_error or upstream_error.

Development Notes

• Keep all external LLM calls on server-side routes.
• Keep client error messages user-friendly; keep technical details in server logs.
• Validate and sanitize all user-visible model output.