AI sentiment dashboard updated throughout the day for Claude, ChatGPT, Gemini, and Grok. Is your AI having a bad day?
Every day, thousands of developers share their real experiences with AI models across Reddit, Hacker News, Bluesky, Twitter/X, and the App Store. But there's no easy way to see the big picture: which models are loved, which are frustrating people, and what's trending.
LLM Vibes scrapes 5 social platforms, classifies sentiment with AI, and scores each model in daily 0-100 windows that refresh throughout the day. No surveys, no voting — just real conversations from real developers.
- 📊 Vibes dashboard updated throughout the day for 4 AI models (Claude, ChatGPT, Gemini, Grok)
- 📈 30-day historical charts with per-model detail pages — vendor-event overlays, official status cards, and anomaly correlation
- 📝 Research articles — long-form analyses with live embedded charts and CSV + Dataset (JSON-LD) export
- 🏷️ Complaint breakdowns — lazy responses, hallucinations, refusals, coding quality, speed, general drop, pricing/value, censorship, context window, API reliability, multimodal quality, reasoning
- 🌐 Source diversity tracking across 5 platforms
- ⚡ Automated pipeline with source-specific scrape windows throughout the day — no manual intervention needed
- 🔍 Keyword matching + AI relevance filtering to reduce noise
- 🌍 Auto-translation of non-English posts for global coverage
| Source | Method | Auth Required |
|---|---|---|
Apify (harshmaur/reddit-scraper, HTML-parsing) |
API token | |
| Hacker News | Algolia Search API | None |
| Twitter/X | Apify scraper | API token |
| Bluesky | AT Protocol search | App password |
| App Store reviews | Apple's iTunes RSS reviews feed | None |
| Mastodon | Public hashtag timelines (5 instances) — unscheduled since Aug 2026 | None |
| GitHub issues | GitHub Issues API — unscheduled since Sep 2026, historical rows only | None |
| Layer | Technology |
|---|---|
| Framework | React 18.3 + TypeScript 5.8 |
| Build | Vite 6 (SWC plugin) |
| Routing | React Router 7 (lazy-loaded pages) |
| UI | shadcn/ui (Radix + Tailwind CSS) |
| Charts | Recharts 2.15 |
| State | TanStack React Query 5 |
| Backend | Supabase (PostgreSQL + Edge Functions) |
| Sentiment AI | GPT-6 Sol via the OpenAI API (Gemini spillover fallback; provider-pluggable via CLASSIFIER_MODEL) |
- Clone the repo
git clone https://github.com/dkships/llm-moods.git cd llm-moods - Copy
.env.exampleto.envand fill in your Supabase credentials - Install dependencies
npm install
- Start the dev server
npm run dev
Note: The scraping edge functions run on Supabase and require their own setup with secrets for API keys (Bluesky, Apify, etc). See
supabase/functions/for details.
src/
├── components/ # React components (model cards, charts, chatter feed)
├── hooks/ # React Query hooks for data fetching
├── pages/ # Route pages (landing, dashboard, model detail, research)
├── lib/ # Utilities, constants, sentiment scale
└── integrations/ # Supabase client config (auto-generated types)
supabase/
├── functions/ # 17 deployable Deno edge functions (5 active scrapers + utilities)
└── migrations/ # Database schema migrations
Contributions welcome! See CONTRIBUTING.md for setup instructions, contribution ideas, and guidelines for adding new scrapers or models.
Built by David Kelly