Sentional — AI-powered account intelligence for Twitter / X: sentiment analysis, activity patterns, and topic discovery in seconds.
Prerequisites: Node.js 20+, pnpm, Upstash Redis, TikHub API key, OpenRouter API key.
git clone https://github.com/febriadj/sentional.gitcd sentionalpnpm installmv .env.example .env.localEdit .env.local:
# Upstash Redis (from Vercel KV integration or upstash.com)
KV_REST_API_URL=https://your-redis.upstash.io
KV_REST_API_TOKEN=your_token
KV_REST_API_READ_ONLY_TOKEN=your_readonly_token
KV_URL=redis://your-redis.upstash.io
REDIS_URL=redis://your-redis.upstash.io
# TikHub — Twitter/X data provider
TIKHUB_BEARER_TOKEN=your_tikhub_token
# OpenRouter — AI model routing
OPENROUTER_API_KEY=sk-or-your_key
OPENROUTER_HTTP_REFERER=https://your-domain.com
OPENROUTER_APP_TITLE=Sentionalpnpm devOpen http://localhost:3000.
Add the same environment variables in your Vercel project settings. Enable Vercel KV from the Storage tab to provision Upstash Redis automatically.
Caution
Vercel Hobby Plan is not supported. The /api/analyses route runs on Edge Runtime and streams inference from a free-tier model with high latency (30–90s typical). Vercel caps Edge Function execution at 30 seconds on Hobby, the model response will exceed this limit before completion.
Options:
- Upgrade to Vercel Pro for a 30s Edge Function limit increase.
- Switch to a low-latency paid model (e.g.
google/gemini-2.5-flash-lite) — response time drops to under 10s, making Hobby viable. - Deploy on a platform with no duration limit (e.g. self-hosted, Cloudflare Workers).
| Layer | Technology |
|---|---|
| Framework | Next.js 16 (App Router, Edge Runtime) |
| UI | Tailwind CSS v4, shadcn/ui, Framer Motion |
| Charts | Recharts 3 |
| AI | OpenRouter → poolside/laguna-m.1:free (Best performing free-tier model). |
| Data | TikHub Twitter/X API |
| Storage | Upstash Redis (Vercel KV) |
| Language | TypeScript |
Not affiliated with X Corp. or Twitter, Inc.