Scrape, classify & chat over your LinkedIn Saved Posts using Supabase + pgvector, Vercel Edge Functions, Playwright MCP, Firecrawl MCP, and OpenRouter LLMs.
- One-time bootstrap scrape of every saved post (≈ 400 posts)
- Daily incremental sync via Vercel cron (stops on first duplicate)
- Rigid Topic → Category → Subcategory taxonomy for clean filtering
- Fast vector & full-text search (pgvector + Postgres)
- Chat assistant with switchable models:
o4-mini-high,o4-mini,gpt-4.1,gemini-2.5-flash - Firecrawl enrichment for external links (optional)
- Cursor AI rule-files for structured, SOLID, KISS code generation
| Layer | Tech |
|---|---|
| Front-end | Next.js (App Router), Tailwind, shadcn/ui |
| Back-end | Vercel Edge Functions + Playwright MCP |
| Database | Supabase Postgres + pgvector |
| AI / LLM | OpenRouter (OpenAI & Gemini models) |
| Agents | MCP Servers (playwright, firecrawl, tavily, git, github, sequential-thinking, memory) |
git clone https://github.com/BjornMelin/linkedin-saved-posts-ai.git
cd linkedin-saved-posts-ai
cp .env.example .env.local # add your keys + LinkedIn li_at cookie
pnpm install
pnpm dev # local Next.js + SupabaseThen open http://localhost:3000 and log in with Supabase Email-Link Auth.
Secrets stay only in Vercel’s encrypted env-store — never commit them.
GitHub secret-scanning is active for public repos. If a secret leaks, rotate immediately.