Repository navigation
Split the dashboard's Environments table and history by service, with docs and /status links and more status detail, and fit every page on a phone - #160
Merged
Conversation
The report kept one /status shape for both services and nothing to link to. The generator now also whitelists the service's own version, the Solr index's lastModified and startTime, NameRes ES's top-level numDocs and size (as 'index', since it isn't Solr), and a summary of NodeNorm ES's Elasticsearch nodes: how many, their version, and the fullest disk. It records how long each /status took to answer. Each target section also gains the configured backend for each service and its docs and /status URLs, built from targets.ini. The Elasticsearch services serve Swagger UI at their root and answer /docs with a 404, so their docs link is the base URL. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NameRes v1.7 moved its Solr core's stats from under 'solr' to the top level of /status, so test, ci, dev and exp were reporting no document count, and the history line's solr_docs/solr_size had been null for them. The top-level block also carries the index's own 'version' counter (68618), which the previous commit would have taken for the service's version. The index stats now go into one 'index' key whether they came from 'solr' (prod, NameRes v1.5), the top level (v1.7) or NameRes ES; 'version' is kept only when there is no top-level numDocs; and the history line reads 'index'. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NodeNorm and NameRes are loaded separately and can be on different Babel releases: test had NodeNorm on 2025sep1 and NameRes on 2026jul22, but the table's one Babel row, and the environment cards, read only NodeNorm. Each service now has its own section: a header row of docs and /status links per environment, then status, Babel version, Biolink model, version, backend, and its own index and health rows (Redis records and memory, Elasticsearch nodes and disk, NameRes documents, index size and dates, latency, /status response time). A row no environment has a value for is hidden. The test counts become a third section. A card shows both Babel versions, a line each, when the two services disagree. NodeNorm's Version row is not shaded for disagreement: Redis 2.x and ES 1.x are separate codebases, and comparing them shaded every environment. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NodeNorm ES from 1.2.0 can serve several Babel releases, chosen per request, and lists them under babel.available_versions in /status: on 2026-11-09 ci served 2025sep1 and 2026jul22 while test served only 2025sep1. The generator keeps the names that are real Babel releases, oldest first, and the NodeNorm section gains a 'Babel releases served' row, hidden while no environment reports one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The environment card fell back to NameRes's release, unlabelled, when NodeNorm reported none (NodeNorm down, say), so it read as the whole environment's release. Now a single unlabelled version appears only when both services report the same one; otherwise each reported release is prefixed NN or NR. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NameRes Solr and NameRes ES are separate codebases with unrelated version numbers, the same reason NodeNorm's Version row is already not compared. Comparing them would mark ci-nameres-es amber every day. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes to what trim_status() and fetch_status() keep from a /status:
- babel.available_versions was cut to its first 50 keys before being
validated and sorted, so on a long or oldest-first list the current
releases could be the ones dropped. Every key is now checked, then the
newest 50 kept.
- The Elasticsearch summary read the version of the first node only, and
counted at most 50 nodes, so a cluster mid rolling upgrade looked
uniform. It now counts every node and reports up to five distinct
versions ("versions", a list, replacing "version").
- The top-level "version" was kept whenever the response had no numDocs,
so a NameRes error body without one would pass the Solr index's
version counter through as the service's release. trim_status() now
takes the service and keeps "version" for NodeNorm only.
- response_ms used requests' elapsed, which stops when the headers
arrive. It is now timed with time.monotonic() around the request and
the JSON parse, which is what "/status time" claims to show.
- fetch_status() took the base URL and rebuilt the /status URL with a
dummy backend; it now takes the URL build_report() already computed,
so the fetched and linked URLs cannot drift apart.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The backend defaulted to redis or solr even when targets.ini gives that service no URL, so the always-shown Backend row named "solr" beside a red NoURL on test. It is now None there, and the empty-row filter can hide it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
build_history_line() recorded only NodeNorm's babel_version, so the history diff never saw NameRes move to a new release on its own, which is the case the split Environments table exists to show. It now records nameres_babel_version too, and History.vue diffs it. A field the previous run did not record is not reported as a change, so the first run after this lands does not list every environment. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The history now records nameres_babel_version; the table shows it beside NodeNorm's on any run where the two disagree. Runs recorded before the field existed render as before. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A node mid-restart can report "version": null. sanitize() keeps None as
None, and sorting a set holding None and strings raised TypeError, which
fetch_status caught and turned into {"error": "TypeError"} for the whole
NodeNorm status: the card lost its Babel version, records and every
other row over one field.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The generator records {nodes: 0} for a cluster whose nodes have not
joined, but the Elasticsearch row's `!es?.nodes` read 0 as missing data
and showed an em dash. With only one ES environment, the row-hiding
rule then removed the whole row, so the most important failure state
vanished. The cell now says "no nodes" and is shaded red.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The generator keeps up to 50 release names and the Environments table's cells don't wrap, so a NodeNorm ES serving twenty releases stretched the table far past the viewport. The cell now shows the newest three and "+N older"; a new `title` option on STATUS_SECTIONS rows puts the full list in the cell's tooltip. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Older NameRes Solr releases, and a NameRes that timed out, report no babel_version, so every such card had changed from "2026sep24" to "NN 2026sep24" though nothing differed. Only NodeNorm's release alone is unlabelled again, as the card showed before NameRes's was read; the tooltip says NameRes reported none. NameRes's release alone keeps its "NR" label, so with NodeNorm down it is still not read as the environment's. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fetch_status checked `"error" not in status`, which trim_status only sets for a body that is not a JSON object; test that directly. Runs before this branch recorded solr_docs and solr_size as null for NameRes v1.7+ environments, which moved the index to the top level of /status. The first run after the deploy fills them in. Nothing compares them across runs today, but a comment now warns that a null followed by a number there is missing data, not a change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The History page's change list walks every environment in the whole history file, so one dropped weeks ago (ci-es) was missing from both of the runs being compared and was announced as "no longer reported" every day. A target absent from both runs is now no change at all. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
In the Environments table's compared rows, the majority is found among reported values, but an em dash was still shaded amber whenever the others disagreed, and left plain when they agreed. With NameRes's own Babel and Biolink rows, NameRes ES (which reports neither) was shaded on every day prod lagged behind. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
At 390px the whole page scrolled sideways: the Environments table reached 904px (711px before NodeNorm and NameRes got a section each), the history table 733px, and the navbar's links ran to 484px. The two tables now scroll sideways inside their cards, and the navbar wraps its page links onto a second line. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The results table is 700-odd pixels wide at 390px: six environment columns of outcome badges come to about 410px alone, and the Category column already goes below lg. Dropping columns, which theme.css asked for, can't get it onto a phone, and wrapping it in .table-responsive would silently break its sticky header, as that comment warns. So below md the card scrolls the table sideways and the header stops sticking; from md up nothing changes. The filter bar is outside the card and still sticks. Two other things widened the page: the sources dropdown, which w-auto sized to its longest option, a 580px URL; and the six outcome buttons, about 365px, which a btn-group never wraps. The dropdowns are capped at the bar's width and the outcome buttons wrap. The environment headers no longer break at each hyphen, which tripled the header's height. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ci-nameres-es is ci's NodeNorm against the NameRes ES under validation. At the end of the table its NodeNorm cells were far from the ci ones they match, and its NameRes ones far from the Solr NameRes they are meant to replace. A new SIDE_BRANCHES map in deploymentOrder.js names the target each side branch comes from, and sortBesideParents() moves it there for that table only. Every other table keeps the pipeline order, where ci-nameres-es stays after prod: they are read column against left neighbour, and between ci and test it would make every Solr-vs-ES difference look like a ci→test promotion (see #157). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Environments table shows a whitelisted part of each /status response; clicking a service's Status value now opens the whole thing, error responses included. With /status linked there, each service's header row carries just an "API docs" link, to the Swagger UI the generator located from targets.ini. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bootstrap's .table-warning and .table-danger, and .table-light on the results table's section headings, are fixed light colours: in dark mode they were near-white blocks in an otherwise muted table. The shaded cells now use new .cell-differs and .cell-broken classes built on the warning and danger subtle/emphasis pairs, which flip with the theme, as theme.css asks; light mode looks the same as before. The "no results" badge uses the same subtle pair, and the results headings use .table-group, like the Environments table's section rows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"Docs" read as the written API documentation each service keeps in its repository (NodeNormalization's documentation/API.md, say), which this does not link to: it opens the service's own Swagger UI. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t their root The comment said they "answer the API itself at the root", which contradicted service_links(), and a review of #160 took the "API" link for ES services to be broken. On 2026-11-11 the roots of nodenorm-es and namelookup-es returned Swagger UI and /docs returned 404, so the link was right and the comment was wrong. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
sorted() compared them as strings, so 8.10.0 came before 8.9.0, and the cut to five kept the lexically smallest versions rather than the newest ones. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
trim_status read the list of Babel releases a NodeNorm ES serves only when it was an object, and dropped any other shape without a word. That hides the "Babel releases served" row, which looks exactly like a NodeNorm serving only its default release. A plain list of names is now read the same way as the object NodeNorm ES 1.2.0 sends; any other shape, or a list in which no name is a Babel release, logs a warning in the dashboard workflow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The index block was chosen by the response's shape ("numDocs" in it), so a
NodeNorm body carrying a numDocs would have been given a NameRes index. Gate it
on the service as well, as "version" already is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…site deploymentOrder.js kept SIDE_BRANCHES (ci-nameres-es → ci) by hand: a third place to register a target, after targets.ini and DEPLOYMENT_ORDER, and unlike DEPLOYMENT_ORDER nothing warned when a new side branch was missing from it. It would just have sat at the far end of the Environments table. The parent is now SideBranchOf in the branch's targets.ini section, beside the shared URL that makes it a side branch. generate_report checks it names another deployed target and emits it as each target's side_branch_of, and sortBesideParents takes a lookup instead of the map. test_targets_ini fails when two sections share a service URL and neither names the other. Deriving the parent from the shared URL alone was considered and not done: two sections sharing a URL do not say which of them is the branch. Until the dashboard regenerates, the published report.json has no side_branch_of, and ci-nameres-es sorts last, as it does in every other table. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The diff since the previous run labelled NodeNorm's release "babel version", beside "nameres babel version", so a NodeNorm promotion read as the whole environment changing release. The labels are now "NodeNorm Babel version", "NameRes Babel version" and "NameRes version". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The card template called babelVersion(target) twice, once for its tooltip and once for its lines, building the same object twice on every render. A computed map gives both from one result. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gaurav
marked this pull request as ready for review
November 11, 2026 07:23
gaurav
added a commit
that referenced
this pull request
Nov 11, 2026
main has moved on 260 commits since this branch was cut, including #159 (CLAUDE.md renamed to AGENTS.md) and #160. Conflicts: - CLAUDE.md: kept deleted, as on main. This branch's edits to it (ruff in place of black, the -m unit note, babel_validation imports) are carried into AGENTS.md in the commits after this merge. - pyproject.toml: this branch's ruff dependency and settings, pythonpath and marker comment; main's pyyaml. black stays dropped. - Ten source and test files: main's side. This branch changed them only by the src. prefix rename, ruff's lint fixes and the ruff format sweep, all mechanical, so they are re-applied over the whole tree in the commits after this one rather than hand-merged here. main already had deselected_by_markexpr, which this branch also used. tests/unit/test_gsheet_parsing.py stubbed requests.get inside google_sheet_test_cases, which main replaced with fetch_sheet_csv(): the fixture now stubs that instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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.
The dashboard's Environments table mixed NodeNorm and NameRes, and its one Babel version row, like the cards above it and the History page, was NodeNorm's. The two services are loaded separately and can be on different releases: today
testhas NodeNorm on 2025sep1 and NameRes on 2026jul22, and the site didn't show it. This PR gives each service its own section, with its own Babel version, Biolink model, version and backend, a link to each environment's Swagger UI, its Status values linked to the full/status, and a few more/statusfields worth watching.A card shows one unlabelled Babel release when both services report the same one, or when only NodeNorm reports one; otherwise it labels each service's (NN/NR), so NameRes's release is never read as the environment's while NodeNorm is down. The history now records NameRes's Babel release alongside NodeNorm's, and its list of changes names the service each version belongs to. Closes #161.
Every page now also fits a phone and suits dark mode, and two existing dashboard bugs are fixed: History announced
ci-es, dropped weeks ago, as "no longer reported" on every run, and an environment reporting nothing was shaded as an odd one out.The calls worth overruling
/docs; the ES services serve it at their root and answer/docswith a 404 (checked on 2026-11-11). The generator builds both URLs fromtargets.ini, so the website links only to checked-in URLs. Backend rows show the configured value; NameRes/statushas none.indexkey, wherever they arrive: undersolron v1.5 (prod), at the top level on v1.7 and NameRes ES, so the old report had no document count for four of six targets.history.jsonl'ssolr_docs/solr_sizefill in from the first run after this merges; a comment inbuild_history_linesays a null-to-number step there is missing data, not a change.versionis its index's counter (68618), not a release, and an error body need not carry the fields that would give that away. Soversionis kept for NodeNorm only, andindexis read for NameRes only.babel_versionas NodeNorm's and addsnameres_babel_version, since a rename would blank the Babel version of every earlier run inhistory.jsonl. A field an older run didn't record isn't a change.ci-nameres-essits besideci, whose NodeNorm it shares. Every other view keeps it after prod, where Put ci-nameres-es after prod on the dashboard, and test NameRes /synonyms alongside /reverse_lookup #157 put it so it isn't read as a promotion. The parent comes from a newSideBranchOf = ciintargets.ini, not a map in the website; it is stated rather than derived from the shared URL, because two sections sharing a URL don't say which is the branch. A unit test fails if two sections share a service URL and neither names the other. Until the dashboard regenerates, the publishedreport.jsonhas noside_branch_of, andci-nameres-essorts last there too.theme.cssasked for dropping columns instead, but six environment columns alone are about 410px. Below 768px only; the filter bar still sticks; see Re-evaluate the Results table on phones: it scrolls sideways and loses its sticky header #166.What's here
generate_report.pyadditionally whitelists the service'sversion, the index'slastModified/startTime, and a summary of NodeNorm ES's Elasticsearch nodes: how many, their distinct versions (ordered as numbers, newest five kept), and the fullest disk. It also keeps the Babel releases a NodeNorm ES from 1.2.0 can serve (babel.available_versions, an object or a list, each name validated as a release, newest 50 kept; any other shape logs a warning rather than hiding the row). The table shows the newest three, the rest in a tooltip. It also times each/statusrequest. All of it goes through the existingsanitize/_trimmed_*helpers.STATUS_SECTIONSinreportData.jsreplacesSTATUS_ROWS; a row may now give adangershading and atitletooltip.StatusMatrix.vuerenders a section per service, then the test counts. Shaded cells use theme-aware.cell-differs/.cell-broken, since Bootstrap's.table-warning/.table-dangerwere near-white in dark mode.What it deliberately doesn't do
biolink_model_toolkit_versionisn't in the Biolink row: it's a library, not the model.nn_eq_records,solr_docs) keep their names: renaming them, as Record NodeNorm's and NameRes's Babel versions separately in the dashboard history #161 suggested, would cost the same old-run compatibility for nothing.Verification
On 018b4c7, the unit suite (290-odd tests) and the website tests (60-odd) pass and
npm run buildsucceeds; the generator run against live/statusgaveci-nameres-esits parent, listed ci's two served releases and prod's ES nodes, and gave no NodeNorm an index. On 06c5ad5, with the deployedreport.jsonin Chromium, no page is wider than the viewport at 360, 390, 768 or 1280px, and the Results header sticks from 768px. On bbdfb0d, a site built from live/status, a short live run and the deployed history showed test's NN/NR labels, ci-nameres-es's unlabelled release, prod's ES nodes and disk, and no false History changes; a zero-node cluster and a twenty-release list were checked with a substituted report.Nothing is blocking.
🤖 Generated with Claude Code