Promote develop to main: Universe Explorer release - #2
Merged
Merged
Conversation
Four discovery passes across mempool explorers, protocol explorers, live visualizers, privacy explorers, alerting products, and agent data interfaces. The last two passes surfaced new product names but no new competitor category, capability class, user expectation, or trend, which is the closure condition recorded in market-lessons.md. Adds the deliverables that make the research accountable: the lessons that each produced a specific build decision, the patterns deliberately rejected and why, an emerging-trends brief, a user jobs and pain-point synthesis, a claim-to-source ledger, and the experience brief that turns all of it into one sentence the implementation can be measured against. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The explorer could already prove what one transaction did. This adds the surfaces that make that useful. Universal search classifies outpoints, inscription ids and numbers, rune names and ids, satoshis, and protocol names entirely in the browser, so a query never leaves the device to be understood. Base Bitcoin identifiers still belong to the upstream matcher, and a bare integer is still a block height: nothing here steals the most common search in the product. New pages: an output is addressable at /outpoint/:txid/:vout, because an output is the unit that actually carries protocol assets; /inscription, /rune, and /sat resolve through the overlay's asset lookups; /protocols/:id explains a protocol, its authority, and what is not shown about it and why; /pulse publishes live protocol activity with the denominator it was measured against; /saved is everything this browser remembers, with one button that erases it. The transaction page now opens with a sentence a non-expert can act on, built only from actions the authority reported and worded so it never implies a trade. Block and address pages gained protocol panels that state their own coverage rather than implying completeness. Personalization is local storage only: recent items, saved pages, pinned protocols. No account, no server-side list, no profile. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three checks that make rules enforceable instead of remembered, each runnable over the source tree and over a built bundle, because an unreachable lazy chunk still ships. The branding gate bans the upstream product marks outside a path allowlist that records why each exemption exists, plus two attribution sentences that are permitted anywhere because publishing them is the point of them. The origin gate bans hosted blockchain APIs, public explorers, analytics, and asset CDNs, matching on host boundaries so a filename like btc.component.html is not read as a host and a specification citation is not read as a data source. The text gate bans U+2014. All three run in CI, and the frontend job runs the first two again against frontend/dist after the production build. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…ty origins The fork keeps the upstream code and the upstream licence. It does not keep the upstream product identity, and until now it still shipped it. Deleted, because they exist only to present the upstream product: the about and trademark policy pages, the acceleration checkout, the upstream Lightning node group page, the sponsor index variants, and the upstream node fleet scripts. Rewritten as this deployment's own documents: the privacy and terms pages, and the API documentation. Two runtime behaviours mattered more than the copy. The hosted analytics path is now a documented no-op, so no tracker loads on any hostname. The enterprise service no longer redirects an unknown subdomain to a third-party site, and no hostname is treated as an enterprise subdomain. Cross-network links, the services API, the Electrum failover Host header, and the sample and proxy configuration all point at this deployment or are unset, and the build no longer rewrites asset downloads to a hosted mirror. The Goggles trademark is gone from identifiers as well as from copy, so it does not survive minification. The remaining upstream references are the ones the licence requires: the preserved notices, the fork provenance record, and two attribution sentences. The 33 upstream translation catalogues still carry upstream copy. They are not built: the production script omits localization, so only the English build ships. Re-enabling locales means regenerating them from the rebranded sources. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…ripts The deployment document described an intended stack. This records the one that exists: the public origin, the forward-only tunnel restricted to a single destination port, the gateway, and the three units behind it, with the ports each listens on and the user each runs as. It also records two decisions honestly rather than silently. The database is disabled, because it only powers historical indexing and mining statistics while the live explorer and the whole protocol layer work without it, and the host it runs on is also rebuilding an index. Asset synchronization and localization are omitted from the production build, so nothing is fetched from a third party at build time. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The text gates could not see these: 24 upstream logo, sponsor, promo, and screenshot assets were still being copied into every build, along with the preview cards eight pages set as their own og:image. Those cards carry upstream trade dress, so pages now fall back to the Universe card defined once in the page head. The accelerated-transaction badge uses an icon from the bundled set instead of the upstream sparkles mark. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The Angular asset entry for src/resources is inert; the upstream build relies on the sync-assets step to copy it. The Universe build skipped that step to avoid downloading anything from a third party, and lost the resources directory with it, including the file that carries the release commit hash. It now copies the directory and still downloads nothing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…ng up Two things the first deployment made obvious. The backend bound every interface, so the API answered on the host's public address as well as the one the gateway uses. It now takes MEMPOOL.HTTP_HOST and defaults to loopback; a deployment that wants a wider bind has to ask for it. The docker template keeps the previous behaviour so a container still works. More importantly, the node this explorer reads can be behind the chain that exists, and until now the explorer would have presented its tip as the present without comment. That is the one failure this whole product is built to avoid. The backend now publishes the node's own sync state on /api/v1/backend-info, and a notice at the top of every page states how far behind the data is and by how many blocks. Ordinary propagation lag stays quiet; a node in initial block download does not. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The explorer backend registers every route under its own /api/v1/ prefix, while the frontend and the Esplora-compatible surface address them as /api/. Upstream resolves that in nginx. The gateway did not, so the whole chain API returned 404 while the site itself loaded perfectly, which is precisely the kind of failure that survives a smoke test. Rewritten, tested, and the routing table is now covered directly: overlay paths pass through, explicit v1 paths pass through, unprefixed paths are rewritten with the query string intact, and anything outside the API tree is left for the static handler. WebSocket upgrades go to the backend unchanged, because its socket server attaches to the whole HTTP server rather than to one path. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The guard compared process.argv[1] against the module's own resolved path so a test could import the routing table without opening a socket. The release directory is reached through a symlink, so those two paths never match there: the gateway started, decided it was being imported, and exited cleanly. A clean exit with no error is the worst possible way for a service to fail. Listening is now the default and a test opts out with an environment flag. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…erflowing Found by auditing the deployed site rather than the source. Eight navigation links and the search submit button were icon only with no accessible name, so a screen reader announced them as bare links and buttons. Each now carries a name and a tooltip. Target spacing was measured at the same time and clears the 24 pixel minimum through the spacing exception, so no layout changed for that. On a 375 pixel viewport the protocol directory pushed its state chips past the edge of the screen with nothing to scroll, which hid the very thing the page exists to show. The state column wraps now, and below 576 pixels the row stacks. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The origin sent no content policy, no frame protection, and no referrer policy. A page URL here contains an address, a transaction, or an output, so a default referrer would have leaked exactly the thing a visitor came here to look up privately, to every site they clicked through to. That is now no-referrer, alongside a content policy that permits nothing off this origin. The build injects one inline script whose content changes every build, so the gateway hashes it at start-up and allows it by name rather than opening the policy to inline script. Two more things the deployed site showed. The blockchain toggles had no accessible name. And the incoming transactions panel rendered a heading over empty space, because this deployment keeps no historical statistics; it now says so, which is what every other empty state on the site already does. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
UPSTREAM.md listed the overlay and the first branding pass. It now lists the backend binding and sync-state additions, the analytics and redirect removals, the deleted and rewritten upstream surfaces, the renamed identifiers, and the accessibility fixes, along with the two new classes of merge conflict a sync should expect. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Measured against the deployed site: the largest layout shift on every page was the footer, 0.48 on the protocols page. The column holding the page has a spacer meant to push the footer down, but with no floor on that column the spacer had nothing to expand into, so the footer rendered just under a nearly empty page and jumped the moment data arrived. Worse on the pages that load fastest, because their content arrives later relative to first paint. A viewport-height floor holds the footer at the bottom until real content exceeds it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A floor on the page column parked the footer exactly on the viewport edge, which is the worst starting point: when the page grew, the region the footer swept through covered almost the whole screen, and the protocols page still scored 0.45. Giving the router outlet its own viewport-height floor starts the footer below the fold, so later growth moves it entirely outside the viewport. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Audited both projects with development dependencies excluded. The frontend carried 3 critical and 40 high advisories, the backend 4 high. Frontend: the Angular toolchain moves within 20.3.x to the patched releases, which clears the two sanitization bypasses in core and compiler, the date formatting denial of service in common, and the server-side request forgery in platform-server. 62 advisories become 11, with no criticals and no highs that reach the browser; the six that remain are build tooling whose only offered fix is a major downgrade. Backend: axios and ws move to their patched releases, which clears the request forgery bypass and the uninitialized memory disclosure in the WebSocket server. The backend now audits clean. Every version change is patch level within the same major. The frontend builds and its suite passes; the backend compiles. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…I gates (#1) Universe Explorer: protocol product surfaces, branding cleanup, and CI gates
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Promotes
developtomainso the release branch matches what is deployed atexplorer.bitcoinuniverse.io.Deployed commit:
67e968b0b, reported by/api/v1/backend-infoand by thepublic
/sourcepage.Universe CI passed on that commit: frontend build, backend build and test, and
the documentation, branding, origin, text, and gateway routing gates.
Full change summary and the known limitation are in #1.