RedNote AI Chat Agent

Chat with any RedNote (小红书) post or creator — paste a link, and the AI extracts the full content (text, images, video) and lets you have a real conversation about it.

Total cost: $0 — built entirely on free-tier APIs and open-source tools.

---

What it does

• Single post: paste a note link → AI reads the full text, OCR-scans images, and transcribes video audio
• Creator profile: paste a profile link → AI ingests the creator's recent posts and lets you chat as if you're talking to the creator
• Smart paste: automatically extracts the URL from the messy share text RedNote copies to your clipboard
• Multi-turn chat: remembers the conversation history within each session

---

Tech stack

Layer	Tools
Frontend	Next.js 14, TypeScript, Tailwind CSS
Backend	Python, FastAPI, Playwright (headless Chrome)
AI	Groq API — Llama 3.1 8B (free)
Video	faster-whisper (local Whisper transcription) + yt-dlp + ffmpeg
OCR	PaddleOCR (Chinese + English)

---

Project structure

RedNote-AI-Chat-Agent/
├── frontend/
│   └── app/
│       ├── page.tsx        # Main chat UI
│       └── layout.tsx
├── backend/
│   ├── main.py             # FastAPI app + TTL session store
│   ├── crawler.py          # Playwright-based RedNote scraper
│   ├── video_processor.py  # yt-dlp + ffmpeg + Whisper
│   ├── ocr_processor.py    # PaddleOCR image text extraction
│   └── requirements.txt
└── README.md

---

Setup

Prerequisites

# macOS
brew install python@3.11 node ffmpeg

# Ubuntu
sudo apt install python3.11 nodejs npm ffmpeg

1. Backend

cd backend
python3 -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt
playwright install chromium

Create your .env file:

cp .env.example .env

Open .env and set your Groq API key (free at console.groq.com):

GROQ_API_KEY=your_key_here

2. Frontend

cd frontend
npm install

---

Running

Open two terminals:

Terminal 1 — backend

cd backend
source venv/bin/activate
uvicorn main:app --host 0.0.0.0 --port 8000

Terminal 2 — frontend

cd frontend
npm run dev

Then open http://localhost:3000

---

How to use

1. Copy any RedNote share link (the whole text — URL extractor handles the extra copy text automatically)
2. Paste into the input box and click Start Analysis
3. Wait for processing (image posts ~10–30s, video posts ~1–5 min)
4. Chat freely about the content

---

Optional: stay logged in to RedNote

If posts are blocked behind login, point the crawler at your real Chrome profile:

# In backend/.env
CHROME_USER_DATA_DIR="/Users/yourname/Library/Application Support/Google/Chrome"

The crawler uses a separate _crawler copy of the profile so it never conflicts with your running Chrome.

---

Troubleshooting

Problem	Fix
Crawl returns login page	Set CHROME_USER_DATA_DIR (see above)
Video transcription missing	Install faster-whisper: pip install faster-whisper
OCR not working	PaddleOCR requires Python ≤ 3.12; skip or use a compatible env
Groq rate limit	Lower request frequency or upgrade to a paid Groq plan

---

Deployment

Service	Notes
Frontend	Vercel (free)
Backend	Railway or Render (free tier available)

Note: the Playwright crawler requires a full Chrome install — serverless platforms won't work for the backend. Use a container-based host.