docs: "source cell" describes 6% of the corpus — say what we actually keep - #32
Merged
Merged
Conversation
… keep
Sampled 1,453 rows across nine key prefixes at data version 2026.192:
84 cite a spreadsheet cell. That is 5.8% — 39.5% under fuels, 1.5% under
grid, 0% under materials, freight, spend, refrigerants, gwp and waste.
Rows carrying no source reference at all: zero.
So the traceability claim was right and the noun was wrong. DEFRA ships a
spreadsheet; the IPCC ships chapter tables, CBAM an Annex with CN codes,
ECCC a model database, the Dutch factors a URL. check-mb-shape has
required "the exact cell/table/page" since v1.26 and the platform made
this same correction once already (plugin v6.685.0, Factor Watch alerts)
without sweeping it. Site side landed in plugin v6.1088.0; this is the
SDK half.
Where a cell genuinely is the reference it stays — the DEFRA worked
examples and the E25 assertion in the LangChain behaviour test are
untouched.
TWO DEPLOY ACTIONS THIS DOES NOT PERFORM, both the owner's call:
1. sheets/src/{core,Code}.js and Sidebar.html carry user-visible
strings. They reach installers only on a `clasp push` plus an App
Configuration version bump, and the live listing currently pins
script v3 — same reason the duplicate Help entry is still open.
2. sheets/store/card-220x140.svg is a published store asset; the repo
copy is now correct and the listing still shows the old card until
it is re-uploaded.
The Postman per-request description ("the exact source cell behind each
value") is generated from the LIVE OpenAPI spec, so it is gateway-owned
and unchanged here. The collection-level copy in description.md is
fixed; collection regenerated, check-collection and
check-collection-parity both pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LeaxFttn3GPPVq1yj3Jjuk
This was referenced Sep 13, 2026
jeremiahsay
added a commit
that referenced
this pull request
Sep 13, 2026
…e spec
The per-request description on GET /v1/sources/{id} said "the exact
source cell behind each value". It is generated from the live OpenAPI
spec, so PR #32 could not reach it; gateway PR #124 fixed the spec and
deployed. This picks it up.
check-collection: 28 requests, 0 failing, 3 skipped. Parity: PASS.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LeaxFttn3GPPVq1yj3Jjuk
jeremiahsay
added a commit
that referenced
this pull request
Sep 13, 2026
…ative step Pushed the "source cell" -> "source reference" wording (PR #32) to the script and cut version 5; moved the bound deployment @3 -> @5. Verified before pushing that the remote HEAD differed from the repo by exactly those strings and nothing else, so nobody's Apps Script UI edit was clobbered. appsscript.json untouched: no new scopes, no re-review. What an installer sees still has not changed. This is an EDITOR add-on, so Marketplace SDK -> App Configuration pins a script VERSION NUMBER, not a deployment id — the README already said so and the redeploy is only housekeeping. Setting App Configuration to 5 is a console step. Also: `clasp version list` is not a subcommand in clasp 3.x. It reads "list" as the version description and silently creates a version, which is where the junk v4 came from. The listing commands are `clasp list-versions` / `clasp list-deployments`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LeaxFttn3GPPVq1yj3Jjuk
jeremiahsay
added a commit
that referenced
this pull request
Sep 13, 2026
… the PNG PR #32 corrected sheets/store/card-220x140.svg and stopped there. The Marketplace store listing uploads the PNG, and that was still the raster built from the pre-correction SVG, so the repo said one thing and the file anyone would actually upload said the other. Regenerated from the corrected SVG at the same 220x140: rsvg-convert -w 220 -h 140 card-220x140.svg pngquant --force --skip-if-larger --strip --quality 75-92 Compared old against new before committing: identical but for the one line, "value · source cell · data version · citation" -> "value · source ref · data version · citation". 14,713 -> 4,251 bytes, because the previous build was 8-bit RGB and this one is palettised; text is clean at that size. All four store screenshots checked and left alone — none of them show the corrected string. screenshot-2 happens to show a factor card citing "ECCC_EF_2025 · Table 4.1", which is the argument for the wording change rather than against it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LeaxFttn3GPPVq1yj3Jjuk
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.
Site-side half landed in plugin v6.1088.0; this is the SDK/Sheets/Postman half.
The measurement
Sampled 1,453 rows across nine
key_prefixqueries at data version 2026.192: 84 cite a spreadsheet cell.Rows carrying no source reference at all: zero. The traceability claim was right; the noun was wrong. DEFRA ships a spreadsheet, so a DEFRA row cites a cell. The IPCC ships chapter tables, CBAM an Annex with CN codes, ECCC a model database, the Dutch factors a URL — all exact, all checkable, none of them cells.
Not a new judgement:
check-mb-shapehas required "the exact cell/table/page, not just the document" since v1.26, and plugin v6.685.0 already made this correction for the Factor Watch alerts and was never swept.Kept as "cell" on purpose
sheets/README.mdand the Postman sample — a cell genuinely is the reference there.core.js'scellfield description now says what the field actually holds: "a workbook cell, a table row or an annex clause".Gates
sheetstest suite: 21/21 passnode --checkoncore.jsandCode.js: cleanpostman/check-collection.mjs: 28 requests, 0 failing, 3 skippedpostman/check-collection-parity.mjs: PASSdescription.md.Two deploy actions this PR does NOT perform — your call
sheets/src/{core,Code}.js+Sidebar.htmlcarry user-visible strings. They reach installers only on aclasp pushplus an App Configuration version bump. The live listing currently pins script v3 — the same reason the duplicate Help menu entry is still open. Merging this does not change what an installed add-on says.sheets/store/card-220x140.svgis a published store asset. The repo copy is now correct; the listing keeps showing the old card until it is re-uploaded.Out of scope, flagged
The Postman per-request description ("the exact source cell behind each value", on the source-feed endpoint) is generated from the live OpenAPI spec, so it is gateway-owned. Fixing it means a
gc-api-gatewaychange and deploy.Separately,
scripts/drift/run.mjsin the plugin repo is red onorigin/mainfor an unrelated reason worth a look: gateway PR #117 documented/v1/account/watch,/watch/removeand/watch/digestwith required fields (key,key,cadence) the origin never reads — that breaks the first call a developer makes against those paths.🤖 Generated with Claude Code
https://claude.ai/code/session_01LeaxFttn3GPPVq1yj3Jjuk