Open-source visual research engine. No account, no tracking, no search backend. Runs in your browser.
Query 2,400+ museums, archives, libraries, and open image sources in parallel. Results stream in as each source responds — no waiting for the slowest link.
Exact mode matches your terms precisely for known subjects. Explore mode casts wider across sources for visual discovery. Switch with `m`.
3-lane load-more dispatcher (fresh pages, query variation, pool widening) keeps distinct results coming past page 5. RRF + MMR reduce near-duplicates.
IIIF-compatible works open in OpenSeadragon. Examine high-resolution pieces down to brushstroke level — pan and zoom with mouse or keyboard.
Dominant colours are extracted from every result. Filter and sort the entire grid by palette — find works by colour relationship, not just subject.
Visualise results as a navigable point cloud arranged by visual similarity. Surfaces unexpected connections between works.
Default analysis runs on Cloudflare Workers AI (LLaVA 1.5). Bring your own key for Gemini, Claude, OpenAI, or Ollama if you want to override.
Select any image and press `i`. InspoSearch cross-references it across every source — related works, provenance, and surrounding context in one pass.
Pin images to a freeform canvas. Arrange, group, compare side by side. Persisted locally; optionally shared via 30-day links.
Every action has a shortcut. `/` focus, `m` mode, `i` interpret, `b` board, `s` sources, `k` keys. Built for fast, fluid research workflows.
6 hand-translated base locales plus 95 generated. Locale persisted locally; the app rebinds text on switch without a reload.
AGPL-3.0. Vanilla JS. No framework, no runtime dependencies, no tracking. Add sources via JSON manifest + PR. Self-host in one static folder.