perplexity-killer

Local, privacy-friendly search + grounded summary: SearXNG behind a small reverse proxy, plus a Tauri desktop UI that calls your self-hosted metasearch and (optionally) a local llama.cpp server (e.g. Nemotron-class GGUF).

Phase 0 (design)

• Design guide: docs/design/glass-ui.md
• Static HTML mockup: open mockup/search-glass.html in a browser

Run SearXNG (Docker)

From the repo root:

cd infra/searxng
docker compose up -d
# or: docker-compose up -d

JSON search (via Caddy on localhost only):

curl -sG 'http://127.0.0.1:8080/search' --data-urlencode 'q=searxng' --data-urlencode 'format=json' | head -c 400

If the container layout or settings path differs for your image tag, adjust docker-compose.yml and settings.yml per SearXNG docs.

Run the desktop app

cd apps/desktop
npm install
npm run tauri dev

1. Open Settings (gear).
2. Set SearXNG base URL (default http://127.0.0.1:8080).
3. Optionally set llama-server binary path, GGUF path, host, and port; click Start LLM (or run llama-server yourself).

The app uses Rust + reqwest for SearXNG (no browser CORS issues). Answers stream from the OpenAI-compatible endpoint on llama.cpp when the process is running.

llama.cpp

Use a build that provides llama-server with your model, for example:

llama-server -m ./model.Q4_K_M.gguf --host 127.0.0.1 --port 8081

Match host and port in app settings.

Project layout