#334 v0-detect: canonical data inventory + release manifest + boot cross-check#338
Merged
Merged
Conversation
…story
Answers RY's 2026-07-22 question ('can we simplify our stable of
supplementary parquet files?'): the live set is minimal (17 files +
search index); the accumulation is superseded suffix versions (each
bump a real fix that left its predecessor served — the isamplesorg#326 class)
and two orphan snapshot generations (202604/202606). Suffix-bump
policy + post-grant attic plan included. Chunk 1 of the isamplesorg#334 v0 work.
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
…v0, chunk 2) Machine-readable twin of CANONICAL.md: 20 canonical objects (16 parquet + 4 search-index sidecars), size/etag via 1-byte ranged GET against the live origin (the Worker 403s HEAD and the default Python UA — script identifies honestly). Fail-closed: any missing file aborts emission. Manifest is committed into the site repo so it deploys atomically with the code whose pinned URLs it describes; the boot cross-check (chunk 3) fetches it same-origin. Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
…the release manifest
New OJS cell (depends only on the URL cells; non-blocking, never throws):
fetches the same-origin release manifest and verifies every pinned data
URL appears in it. Mismatch -> console.error + dismissible amber banner
('data release could not be validated'); missing manifest -> console
warning only. Converts the isamplesorg#326 silent-divergence class into
caught-at-next-boot. window.__manifestCheck exposes the result for specs.
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
…rch-index gap - _quarto.yml: manifest added to project.resources (it would never have deployed — detection dead on arrival) - search_index_base: single top-level cell shared by the runtime substrate path AND the manifest check; check also compares manifest.search_index.path so a _v1->_v2 bump with a stale manifest trips the banner - check pins hot_topk.parquet (runtime-loaded), drops df.parquet (offline-only per contract); audits <link rel=preload> hrefs from the DOM - builder: strict probe (206 + exact Content-Range + 1 byte + ETag; --permissive for dev), shard inventory derived from shard_sizes.json (256 shards, not the hard-coded 852) - banner idempotent across cell re-evaluations - CANONICAL.md: 202604 correctly marked as backing the live /current/ alias (NOT movable); 202606 as documented pipeline intermediate; suffix history corrected to actual commits; nonexistent oc_sidecar row removed Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
… + doc accuracy - Shard inventory: base=256 (shard_sizes.json) + hot sub-files=591 (build_stats.json shard_files=847 physical vs shard_count=256 logical - the distinction round 2 surfaced) = 852 objects with sidecars; the builder cross-validates logical count between both artifacts - Probe validates status + Content-Range (strict fullmatch) BEFORE reading the body (bounded read(2)) so a Range-ignoring origin can't force a 300 MB download - Header text now matches reality (ranged GET, not HEAD); CANONICAL's attic sentence names its targets instead of a stale section ref Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
…counts hard-required - hot_shard_files now summed from hot_tokens.json tokens[*].sub_files (their authoritative declaration), not derived by subtraction - three-way cross-check: shard_sizes base + hot_tokens hot == build_stats shard_files, any disagreement or missing count refuses emission - last stale HEAD wording + shard_count/shard_files confusion in note and error text corrected Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
rdhyee
added
the
CC+Codex+LGTM
Contributor
Author
Staging (fork Pages, branch build): PASS. The manifest deploys ( Ready for merge on Raymond's call. |
rdhyee
added a commit
to rdhyee/isamplesorg.github.io
that referenced
this pull request
Jul 24, 2026
- Every stale reference to the deleted tutorial updated (navbar+sidebar, explorer footer, how-to-use, README, SERIALIZATIONS consumer note, narrow_vs_wide + archive links, 3 test files now assert 'Guided Tour') - Deep-link honesty: intro promise softened to what URLs actually carry; Anatolia links 300km -> 150km so mode=point is real (point-mode threshold is 180km); densest-cell claim scoped to the res4 aggregate - CSV claim corrected to the 50,000-row cap; place-name search mechanism corrected to place metadata (not coordinates) - Source bars use the real SOURCE_COLORS palette; vocab join constrained to lang='en' - Note: CANONICAL.md link resolves on merge (PR isamplesorg#338 landed on main after this branch was cut — merge brings them together) Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y
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.
BLUF (plain English)
The Explorer's live data set is already minimal — 16 named parquets + the search index. The "accidental history" is what surrounds it: superseded suffix versions (each bump a real fix that left its predecessor served — the #326 failure class) and old snapshot generations. This PR neutralizes the accident three ways: a human inventory (CANONICAL.md), a machine manifest (generated from the live origin, deployed with the site), and a boot cross-check — if the page ever pins a data file the manifest doesn't bless, users see an amber "data release could not be validated" banner instead of silently inconsistent results.
What's in it
project.resourcesso it deploys same-origin with the code whose expectations it recordsmanifestCheckcell — non-blocking, never throws; pins all 16 URLs + 4 runtime sidecars + whatever<link rel=preload>tags are actually in the DOM; compares the sharedsearch_index_basecell againstmanifest.search_index.path; mismatch → console.error + dismissible banner; missing manifest → warn onlyReview trail (4 Codex rounds)
Round 1 caught the two big ones: the manifest wouldn't have deployed at all (absent from
project.resources— detection dead on arrival) and CANONICAL mislabeled 202604 as orphan when it backs the live/current/alias tutorials use. Rounds 2–3 tightened the shard inventory (my "852 was wrong" correction was itself wrong — 256 logical vs 847 physical files) and probe strictness. Round 4: LGTM, no findings.Issue #334 stays open until this merges + the v1 (URL-derivation) disposition is recorded there. v1 is explicitly post-grant.
🤖 Generated with Claude Code
https://claude.ai/code/session_01Nohddj7oXLnkJ4P9Cibg8Y