Roadmap
What we're building next, why it matters, and how to influence it.
Priority reflects current thinking — not a commitment. Dates are targets, not promises. File an issue or email us if something here is wrong for your use case.
Now (shipping in the next 4–6 weeks)
Source health dashboard
Live status page showing which adapters are responding, average latency, and recent error rate. Lets contributors know when a source needs attention before users file a ticket. Why it matters: with 2,400+ sources, silent breakage is the main failure mode.
Metadata quality pass
Fill the most common empty fields (artist, date, medium) across the top 50 sources by traffic. Currently some adapters drop useful fields because of shallow response mapping. Why it matters: metadata is what turns a thumbnail into research.
Institutional outreach
Direct partnership conversations with museums and archives interested in being indexed. See institutions page. Why it matters: the quickest path to better metadata is asking the source.
Next (Q2 / Q3 2026)
Saved collections
Bookmark images into named collections (localStorage first, then optional sync via Cloudflare KV). Export to CSV with attribution.
Download queue with attribution
Batch download selected images alongside a credits file — title, creator, date, source URL, licence. For research use where citations are required.
Advanced filters in the search bar
Date range, medium, source category — toggleable without leaving the grid.
Similar image search
Reverse lookup via perceptual hash. Select an image → find visually related works across the full corpus.
PWA / offline mode
Installable, with offline browsing of previously cached searches.
Later (exploring — may or may not happen)
Public read-only API
Programmatic access to /search and /sources with a rate-limited public key. Already exists internally via api/worker.js — the work is documenting, rate-limiting, and supporting it.
Browser extension
Right-click any web image → "Search InspoSearch."
Embed widget
Museums and archives can embed a narrow-scope InspoSearch on their own domain.
Collaborative boards via CRDTs
Shared boards without a server — already partially wired through the BOARDS KV namespace; the research question is how to do multi-user edits without a coordinator.
InspoEnrich (commercial, separate product)
A metadata enrichment service for institutions, built on the same engine. Keeps the InspoSearch app free and AGPL; funds sustainability. Early-stage — more in 2027.
Recently shipped
See the changelog for dated details. Highlights from 1.1.0:
- Workers AI as default vision provider (LLaVA 1.5) — analyse works without keys
- 3-lane load-more dispatcher — infinite scroll survives past page 5
- RRF fusion + MMR diversification — fewer near-duplicates, more visual variety
- Deep-link URL state for shareable queries
- Opt-in community metadata contribution
Influencing the roadmap
- Request a feature: open an issue with the
feature-requestlabel. Tell us the use case, not just the feature name — we prioritise by how many real workflows something unblocks - Vote: thumbs-up reactions on existing issues count
- Fund specific work: email us — directed contributions can reshuffle priority
- Ship it yourself: see Contributing