Repository navigation
release: 0x off benches 304 and 033, its key is dead - #3276
Merged
Merged
Conversation
… 0x for it (#3275) 0x's API answers 403 "You cannot consume this service". That is word for word the error zerox.env records as the reason the previous key was replaced on 19 October, so the replacement is dead the same way. The two benches told a reader two different lies about it. On 304 (quote output) 0x was declared and never published: ten providers on the board's list, nine in its data. A declared row that never appears reads as "this aggregator does not quote these pairs". On 033 (quote latency) it was worse, because the row DID publish: availability unavailable, unresponsive true, successRate 0, p50 0, on ethereum, base, arbitrum and bsc. The page printed 0x as the unresponsive provider in a latency table. The thing that is unresponsive is our credential. The harness adapter stays. Nothing is wrong with the code and deleting it would mean rewriting it when a key arrives; a provider the spec does not declare never reaches a board. The spec is what makes a claim to a reader. What the removal drags with it, because a provider set is not only a list: bench-unfiltered v99 -> v100 the set changed all-benchmarks v92 -> v93 lockstep bench-filters v40 -> v41 both specs declare a chain dimension providers v19 -> v20 product pages provider-set hash 9ca6a160 -> eecad3db 0x is now measured in no bench, so getProvider returns nothing and /products/0x 404s. It was a live 200 listed in the sitemap, so it goes in REMOVED_PRODUCT_SLUGS for the 410 and the sitemap exclusion, and in the middleware matcher, which is a literal list that its own test pins: that test is what caught the missing entry. Same shape as the goldrush note already there, and the reason it matters twice is that the worker's blob keeps a provider in its product list for a sweep or two after the spec changes, which is what rolled the Blast deploy back. /compare/0x-vs-paraswap also stops existing. It is noindex, follow and not in the sitemap, so there is nothing to retire. Three pieces of copy named 0x and are now wrong rather than stale: the gas-unit example on 304 (ParaSwap reports in wei too, so it carries the example), and the keyed-provider count on 033, which was four and is three. One number was already wrong before this change and is the same class of defect: the Ethereum tab said "across the 7 measured providers" while eleven published. It is now {{count:chain:ethereum}}. bench-template.test.ts says a hand-typed count made 29 specs contradict themselves on 2026-10-02. Checked rather than assumed: the lifi bullet claims its tool list includes "seven providers measured separately here". LI.FI's /v1/tools names exactly seven of our cohort (kyberswap, bebop, openocean, enso, paraswap, nordstern, route) and 0x is not among them, so that bullet is correct and untouched. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Flotapponnier
added a commit
that referenced
this pull request
Oct 24, 2026
release: remove the raw placeholder #3276 put on the Ethereum tab, and close the gate
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
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.
One change: 0x comes off benches 304 and 033 (#3275).
Its API answers
403 "You cannot consume this service"— word for word the errorzerox.envrecords as the reason the previous key was replaced on 19 October. The replacement is dead the same way, and each bench turned that into a claim about 0x:unresponsive: true,successRate 0,p50 0on ethereum, base, arbitrum and bsc. The page printed 0x as the unresponsive provider in a latency table. The thing that is unresponsive is our key.The harness adapter stays, so renewing the key brings the provider back with one spec block.
What ships with it
Five cache keys, because a provider set is not only a list:
bench-unfilteredv100,all-benchmarksv93,bench-filtersv41 (both specs declare achaindimension),providersv20, and the provider-set guard hash.A live URL retires. 0x is measured in no bench now, so
/products/0x— a 200 listed in the sitemap — goes intoREMOVED_PRODUCT_SLUGSfor the 410 and the sitemap exclusion, plus the middleware matcher, whose own test caught the missing entry. This is the same guard the Blast rollback argued for: the worker's blob keeps a provider in its product list for a sweep or two after the spec changes, so without it the sitemap advertises a URL that has started 404ing.Copy. The gas-unit example on 304 and the keyed-provider count on 033 named 0x and became wrong, not merely stale. And one number was already wrong before this change: the Ethereum tab claimed "7 measured providers" while eleven published; it is now
{{count:chain:ethereum}}.After the deploy
The worker clones
devat build, so the boards keep their old provider set until a rebuild plus roughly 72 minutes of display floor. Two things to watch then: bench 304's Robinhood tab should go from three aggregators to four (ParaSwap, #3272, already on main and waiting on the same rebuild), and 0x should be absent from every tab of both benches rather than present and empty.Enso is left declared: its key is exhausted the same way (
429 credit_quota_exceeded) and it still readsunavailableon 033, but the disclaimer names it explicitly, so the page is not making a silent claim.Gates on the branch:
pnpm validate253 specs,tsc --noEmitclean,bun test src/719 pass.