Repository navigation
feat(portal): load Google Tag Manager and GA4 from portal.analytics - #473
Merged
Merged
Conversation
sohail2721
pushed a commit
that referenced
this pull request
Nov 11, 2026
The list is only ever asked whether it holds a name (SonarCloud on #473). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
sohail2721
pushed a commit
that referenced
this pull request
Nov 12, 2026
The list is only ever asked whether it holds a name (SonarCloud on #473). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
sohail2721
force-pushed
the
sohail/portal-analytics
branch
from
November 12, 2026 11:01
cd8de5e to
0231cfc
Compare
Step 1.1 of .ai/plans/portal-analytics.md: pageLocation(), holdsOAuthAnswer() and the PortalEvents map with docs_page_view, tested without a DOM. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The list is only ever asked whether it holds a name (SonarCloud on #473). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Step 1.2 of .ai/plans/portal-analytics.md: the GTM and GA4 adapters, analyticsFor() with startAnalytics(), report() and reportPageViews(), and the shared adapter suite. The vendor settings join portal-types.ts. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…answer Step 1.3 of .ai/plans/portal-analytics.md: insertWhenClean() checks every 100 ms for up to 5 seconds, inserts once the answer is gone, and leaves a page never cleaned to the next page change. Every analyticsFor() test now runs on a fake clock. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ribe Nothing reads it before PR 2 adds window.apimatic.analytics.subscribe(), and the plan now builds top-down: nothing is written before the code that calls it. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Step 1.4 of .ai/plans/portal-analytics.md, the browser entry point: the root route subscribes the loader to onRendered and reports page views; portal-analytics.ts builds the one instance from portal.analytics and starts it at module evaluation; the shared type and PortalIdentity carry analytics, written as the default until the parser lands. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
analyticsFor() and its inner functions become the Analytics class: start(), insertWhenClean(), report() and reportPageView(). The router subscription moves into the root effect, which already holds the router, and the one-use helpers are inlined. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Removes the wait while the address holds an OAuth answer (step 1.3, 1c616d5): holdsOAuthAnswer(), the nine-name list, the timing constants and insertWhenClean(). start() appends the vendor tags directly and the root reports page views only. fuma-nama/fumadocs#3809 is the PR's merge prerequisite instead; until it ships, a Google tag can read the answer after a playground sign-in. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…out of replays The root route's head() writes Google's site verification meta when the identity file holds a value. The identity file gains `seo`, written as null until the config parser reads it. The API page wraps the playground in one element carrying the opt-out markers of the session replay tools a GTM tag may load (PostHog, FullStory, rrweb, Amplitude, Hotjar, Clarity, LogRocket), so a replay records the rest of the page and leaves the playground's inputs blank. The playground client is lazy-loaded from Fumadocs' public export, as Fumadocs itself loads it. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
`AnalyticsConfig` parses `googleTagManager` (with its optional `environment`), `googleAnalytics` and `keepQueryParams`, with every rule and message from the table in apimatic/apimatic-io#2348; an OAuth answer's names are refused in the kept list. `SeoConfig` parses `googleSiteVerification`. Both reach the identity file through `identity()` and are serialised only when set, so quickstart's block gains nothing. `portal generate` warns, before the build, when Google Analytics is set: the portal sends its own page views, so the stream's history-change page views must be turned off or every page change counts twice. GA4's `debugMode` is written as false until the delivery reaches `identity()` in the next step. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A container ID, a measurement ID, an environment's auth and preview values, the verification token and a kept query parameter name are any non-empty string: a wrong value fails at the vendor, not in the CLI, so the parser checks presence and type only and holds no pattern. The kept-name list no longer refuses an OAuth answer's names either. Decided 2026-11-12; it replaces the locked rule and the format rows of the table in apimatic/apimatic-io#2348, and is to be recorded in #2347. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ntity file `portal generate` and `portal serve` take `--no-analytics`, for a staging build that must not count. The flag reaches `identity()` as one boolean, passed through the prepare action and the project service and never stored: with it off, both vendors are written as null and the browser loads no tag, while the kept query names still go through. `portal serve` hands the same value to the rewrite when the block changes. `portal serve` is a hosted portal in every way: GA4's debug mode and the preview/build distinction are gone, so the sketched delivery class is not built. Decided 2026-11-12, to be recorded in apimatic/apimatic-io#2347. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Editors now offer both blocks. Every ID, environment value, kept name and the verification token is any non-blank string, as the parser reads them, so the schema carries no pattern and refuses nothing the parser accepts; the schema test holds the two to the same verdict on each case, including the formats the parser stopped checking. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The default end-to-end build reads a copy of its fixture with both vendors and a verification value set, and checks the output: a browser script holds both IDs while no prerendered page names the tag host, every prerendered page carries the verification meta, and the operation page holds the playground's prerendered header inside the element session replays skip. The fixture the other suites read stays as it is. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The README's generated sections carry --no-analytics on both portal commands. The docs page draft for docs.apimatic.io/v4 travels in .ai/plans/portal-analytics-docs.md: the settings, the page view event, the GA4 and GTM setup, previews and staging, Search Console, consent, the CSP domains, container access, session replay and other tag managers, with the CLI reference row. It is the reviewer's to take to the docs repository. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Three more events through the same report(): docs_markdown_copy when the copy succeeds, docs_plugin_install as the install command's copy button is clicked, through an Actions wrapper around Fumadocs' button, and docs_sdk_download with the SDK's language from every download link, the SDK page's own included. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Fumadocs' copy button has no callback, so the document hears every click: one inside a code block's figure reports docs_code_copy with the block's caption, or the selected tab's name in a tabbed set. An untitled block reports nothing, and the install command's block, which reports its own copy, is marked and skipped. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
One docs_search when the dialog closes after the reader typed in this open: the term trimmed, the result picked or '' for none, and how many results the list showed. Fumadocs closes the dialog before it hands over the pick, so the report waits a tick for it; a reopen closed without typing reports nothing. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
docs_not_found from the not-found page's mount: page_referrer is the portal page the reader left for it after an in-portal link, else the document's referrer without its query and # part on a first load, else empty. The page's own docs_page_view still arrives. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
docs_api_request once per result: the endpoint as the spec writes it, the HTTP status or client_error, and the duration, the response's own or timed from the request's start for one that never arrived. Never the request's address, headers or body. The reporting wraps Fumadocs' result display inside the playground's lazy chunk, so the split from the page bundle holds. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The docs draft's table lists every event with its keys and when it fires, the GA4 custom dimensions and metrics to register, and the GTM trigger and variable pattern for the events beyond the page view. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The playground's dialog, which Fumadocs portals outside the marked wrapper, is marked for session replay as it appears, until fuma-nama/fumadocs#3824 lets the markers be passed to it directly; the markers themselves move to one module the wrapper shares. docs_search carries the page it was made on, captured when the dialog opens, since a pick changes the address before the close is reported, and a pick without typing counts as a search. docs_api_request reports a result object once however often the display mounts, which happens on every reopen and twice under StrictMode, and a request that failed before it was sent has no duration. docs_not_found reports once per missing page and again for the next, keyed on the path. The code-copy listener ignores dropdowns and the playground's dialog, and prefers a block's caption over its tab. Vendor values are escaped in the tag addresses, since the parser checks presence only. The two empty events take no keys at report(). The docs draft's consent, CSP, replay, search and GA4 lines are corrected: direct GA4 cannot wait for a banner, search_term is GA4's own parameter, and the redact setting sits under the tag settings. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The warning that a GA4 stream counts page changes twice fires only when the build loads GA4, so a staging build built with --no-analytics is not told to change a stream it does not report to. AnalyticsConfig holds GA4 as a GoogleAnalyticsSetting, named as the template names it, instead of a bare measurement ID rebuilt in two places. Quickstart names the two booleans it passes to serve, the prompt builds its message before logging it and links with a scheme, and the toJSON() comment says what the code does. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The step-by-step plan the analytics PR was built from, with each step's commit, the branch table, the questions each step raised and their answers, and the review of 2026-11-12. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
sohail2721
force-pushed
the
sohail/portal-analytics
branch
from
November 12, 2026 11:25
0231cfc to
b8a13e6
Compare
In `portal serve` the root effect runs while the router is still pending, before the page's own head has applied, so the first docs_page_view carried the site name as its title; the address dedupe then dropped the onRendered report that had the right one. A prerendered page mounts with the router idle and the right title, and raises no onRendered for its first page, so the direct report stays for that case: the effect now reports directly only when the router is idle at mount, and leaves the first page to onRendered otherwise. Seen with headless Chrome against both a served and a built portal, each sending one first page view with the page's title after the change. The plan records the end-to-end run. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…rtup Mutation observers are told in the order they were made. One made as the bundle starts, before any vendor tag, marks the dialog before a replay tool's observer records it, which one made per mount did not. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The test hook no longer marks the cURL sample and the dialog, which carry the replay markers alone; the e2e test checks the one element again, with its markers. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…age ID Copies report through each block's Actions slot instead of a document listener: the MDX blocks, which the build names the language of, the API reference's blocks, its Markdown and the playground's, Example Response, and every command block, an SDK card's install command included. language is the ID SDK downloads also send, csharp or shellscript; the IDs and aliases move from shiki-bundle.ts to languages.ts, so a page can read them without loading the highlighter. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…quences searchOutcome() takes the dialog's open, loading, typed and picked state; the dialog keeps the refs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A first client-side visit to an operation page showed nothing in the playground's place until its chunk loaded; Fumadocs shows its endpoint bar there, which it does not export, so the template draws the same. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Comments that restated the code go; the rest say why in one line. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The build plan, its review log and the docs draft for apimatic-docs stay on the author's machine, so the squash brings none of them into dev. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fumadocs-openapi 12.6.0 reads a returned sign-in's answer from the address, so the callback sends the reader back with it attached, and gtm.js and gtag.js read the address for themselves. The tags now wait while the address carries the callback's path or a state, which Fumadocs sends with every authorization, and the playground loads them once Fumadocs has taken the answer out, in its effects; an answer nothing takes out waits for the next page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mehnoorsiddiqui
left a comment
Contributor
There was a problem hiding this comment.
I tested this branch end to end: the unit and e2e suites, plus headless Chrome against built portals and portal serve. Everything works apart from the not-found bug I commented on earlier. Most of the notes below are about things that will be hard to change once this ships.
A few general ones:
- Most of Waseem's and Saeed's threads look fixed in the code now, but 18 of them are still open with no replies. Could you reply to or resolve them, and re-request review from Saeed?
- The analytics logic is unit-tested, but the components that report events aren't, which is how the not-found bug got through. A lot of it also depends on Fumadocs internals (markup, storage keys, render order), so a Fumadocs update could stop events or the tag hold without any test failing. A browser smoke test that fires each of the 8 events once and checks their keys would catch both. I can share the Playwright scripts I used as a starting point.
- The manual GTM Preview / GA4 DebugView checks in the description are still unticked. My headless run doesn't replace those.
…n it The not-found page keyed its report on the router's location, which already names the page a reader is leaving for while the not-found page is still mounted. Leaving it by Back, the Home link or any other link sent a second docs_not_found, for the page the reader went to, with the missing page as its referrer. The report is now keyed on the path the router has resolved, and nothing is reported in between. Seen in headless Chrome on portal serve and on a built portal, with in-app navigations only: leaving the missing page reports nothing more, and a second visit and a second missing page each report once. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
languageId() looked a name up as written, so a fence written TS or TypeScript missed the alias table, and a language the bundle does not know went out in the author's capitals: several values for one language in the one GA4 dimension. It now reads the name in lowercase first. docs_sdk_download goes through the same function, so the two events share one table instead of two sets that happen to match. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
… term result was the picked page's path under the base path, while every page_location is a whole address, so the two did not join in a report. searchOutcome() now passes the pick through pageLocation(), which also drops the # part of a heading hit. The term is cut at 100 characters, which is what GA4 keeps of a value, so no other tag gets more of a key or a token pasted into search. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Every other built-in name is one an ad or mail platform reads from the address itself. Nothing reads ref, sites use it for referral and invite codes of their own, and a customer can add a name through keepQueryParams but cannot take one out, so once shipped it could not have been removed. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Customers build GTM triggers and GA4 reports on them. The typed table follows the interface through a rename, so a rename passed CI. The names and keys are now also listed as plain text, and a change has to be made in that list on purpose. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
resolve() and resolveSettings() defaulted to loading the vendor tags, so a new caller that forgot the argument got them without noticing. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…response copy button The check that the cURL sample sits inside an element session replays skip went out with 95c8187, though the sample follows what the reader types. It is back. A second check expects Fumadocs' copy button on "Example Response" to be the button[aria-live] the report listens for, so a Fumadocs update that changes either fails the build test instead of silently dropping the markers or the event. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
… only Its two values are public in the built portal, and with them anyone can load that environment's version of the container. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
sohail2721
commented
Nov 16, 2026
sohail2721
left a comment
Contributor
Author
There was a problem hiding this comment.
Thanks for the end-to-end run. Each note is answered in its thread; the three general ones:
- Open threads. Waseem's and Saeed's are all answered, and resolved but one: a playground result that lands after the dialog closed, which needs a result callback from Fumadocs. Saeed's review is re-requested.
- Browser smoke test. Agreed, and it belongs in apimatic/v4-test-automation, which already drives a browser against this CLI's build on every push to
dev. Yes please to the Playwright scripts. It is a follow-up item in the description; in this repository the new guards are the pinned event table and the two built-page checks. - Manual checks. GTM Preview and GA4 DebugView are still to do before merge.
…short name a block uses A language the bundle does not ship went out by the name the author gave it, so kt and kotlin were two values in the one GA4 dimension. A small report-only map beside the bundle's aliases names the common short names by Shiki's ID, and a test holds each entry to the ID and alias Shiki lists. A block in no language reported plaintext from a page and text from a description; both are plaintext now. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…rror docs_api_request sends status as the HTTP status in digits or as client_error, in one field customers register as one GA4 dimension. The choice is now written on the event's type and held by a test, so a change to either form is made on purpose. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
|
mehnoorsiddiqui
approved these changes
Nov 16, 2026
This was referenced Nov 16, 2026
Merged
Contributor
|
🎉 This PR is included in version 2.3.0-beta.1 🎉 The release is available on: Your semantic-release bot 📦🚀 |
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.



PR 1 of apimatic/apimatic-io#2348, with PR 3 (the other seven events) following in this same PR since 2026-11-12, so GA4 and GTM can be tested with every event before anything merges. PR 1:
portal.analyticsloads Google Tag Manager and GA4 fromapimatic.json, and the template reports each page view once asdocs_page_view. The detail and the code sketches are in apimatic/apimatic-io#2349. The build plan and the docs draft are kept out of the repository, so the squash brings neither intodev.Draft until the gate:
pnpm testgreen, the manual list in apimatic/apimatic-io#2349 ("Manual, on a test portal") checked, and each of PR 3's events seen once in GTM Preview with all of its keys. The squash title then covers both:feat(portal): load Google Tag Manager and GA4 from portal.analytics and report the portal events.What is on the branch
Read it from the entry point down.
portal-template/src/routes/__root.tsx) reports the first page's view on mount and one peronRendered;head()lists no script, and writes thegoogle-site-verificationmeta when the identity file holds a value.portal-template/src/components/api-page.tsx) renders the playground's header inside one element carrying the opt-out markers of the session replay tools a GTM tag may load (PostHog, FullStory, rrweb, Amplitude, Hotjar, Clarity, LogRocket) and thedata-portal-playgroundtest hook, with Fumadocs' endpoint bar while the playground's chunk loads. The cURL sample carries the same markers, since it follows what the reader types. The client is lazy-loaded from Fumadocs' public export, as Fumadocs itself loads it.portal-template/src/lib/portal-analytics.ts) runs at module evaluation, before React and before any vendor tag, never during prerendering: firstmarkPlaygroundDialogs()(lib/replay-markers.ts), one observer that marks the playground's dialog as Fumadocs adds it, ahead of any replay tool's observer (until openapi: let PlaygroundClient put attributes on the playground dialog fuma-nama/fumadocs#3824 takes the markers itself), thennew Analytics(portal.analytics).start().portal-template/src/lib/analytics.ts) holds the event table and theAnalyticsclass:start()createswindow.dataLayer, pushes every vendor's setup entries and inserts the tags, once per page load, and underportal servereloads the page when a config save changes the vendors;report()adds the cleanedpage_location(pageLocation()), pushes{ docs: null }then{ event, docs }, and calls each adapter that sends for itself;reportPageView()andreportNotFound()keep the page history the two events need, and the not-found page reports only once the router has settled on it (components/not-found.tsx), so leaving it reports nothing for the page the reader goes to. Nothing is reported on/oauth/callback, and the tags wait while the address holds an OAuth answer (holdsOAuthAnswer()inlib/oauth-callback.ts): on the callback, and on the page a sign-in returns to until the playground has taken the answer out, which loads them (analytics.loadTags()fromcomponents/playground.tsx).portal-template/src/lib/vendor-tags.ts): theadaptersregistry is keyed by both vendors, so a vendor without an adapter does not compile. GTM pushesgtm.startand loadsgtm.js(with theenvironmentsuffix when set), and reads the data layer itself. GA4 defineswindow.gtagonly when missing, pushesjsandconfigwithsend_page_view: false, loadsgtag/js, sends each event under GA4's own name, and setspage_locationon each page view, so GA4's own scroll and click events carry the cleaned address too.insertTags()appends each tag once, none while the address holds an OAuth answer.components/code-copy.tsx, through Fumadocs'Actionsslot) with the block's language ID, the IDs SDK downloads send too (csharp,shellscript), both throughlanguageId(), which reads a name in lowercase and knows the common short names of languages outside the bundle (ktiskotlin), so one language is one value; MDX blocks get theirs at build time (lib/transformer-language.ts), the API reference's at render, and "Example Response", whose button Fumadocs keeps to itself, through a click handler on its card. The playground times its requests in a ref of its own and reports each result once (components/playground.tsx,lib/api-request.ts). The search dialog's report issearchOutcome()(lib/search-report.ts): the result picked goes out as a whole address in the form ofpage_location, and the term at most 100 characters long, which is what GA4 keeps.portal-template/src/lib/portal-types.ts,PortalIdentityinsrc/types/portal/portal-config.ts) carriesanalyticsandseo, filled byAnalyticsConfigandSeoConfigfrom theportalblock, with #2348's messages less its format rules: an ID, an environment value, the verification token and a kept name are any non-empty string, and no name is refused (decided 2026-11-12, to record in #2347);portal generateandportal servewarn when GA4 is set.--no-analyticsonportal generateandportal serve, defined once inFlagsProvider) reaches the actions asPortalBuildOptions, andPortalSourceContextapplies it where the settings are read, for a build and for each config save underportal serve: both vendors are dropped, the kept query names stay. The options have no default, so every caller chooses.portal serveloads exactly what a hosted portal loads: no debug mode, no preview distinction, so the sketchedPortalDeliveryis not built (decided 2026-11-12, to record in #2347).apimatic.schema.json) describes both blocks with no pattern, as the parser reads them; the schema test holds the two to one verdict per case.test/e2e/portal-build.test.ts, behindAPIMATIC_E2E=1) builds the default fixture with both vendors and a verification value set: the IDs sit in a browser script, no prerendered page names the tag host, every page carries the meta, and the operation page holds the playground's prerendered header and the cURL sample each inside an element replays skip, and the copy button of "Example Response" as thebutton[aria-live]its copies are reported by, so a Fumadocs update that changes either fails here.test/commands/portal/*-flags.test.tsrun both commands to see the flag reach the settings reader.Steps, as the plan numbers them: 1.1 the pure functions, 1.2 adapters and page views, 1.3 the loader's wait (built, then removed, see below), 1.4 the bundle wiring, 1.5 the verification meta and the playground out of replays, 1.6 the parser and the GA4 warning, 1.7
--no-analytics, 1.8 the schema, 1.9 the end-to-end proof, 1.10 README and docs. PR 1 is complete. In this PR, PR 3: 3.1 the three click reporters, 3.2docs_code_copy, 3.3docs_search, 3.4docs_not_found, 3.5docs_api_request, 3.6 the docs (done). PR 1 and PR 3 are complete; the gate below is what remains. PR 2 (portal.scripts) follows in its own PR.Before merging
1bb6dfbb): with 12.6.0 the playground reads a returned sign-in's answer from the address, so the tags wait until Fumadocs has taken it out, and an answer nothing takes out waits for the next page. Seen in headless Chrome againstportal servewith both vendors: no tag at#access_token=…&state=…, both once the address was clean, and no GA4 hit carrying the token.createOAuthHandler()does, so the answer never reaches the address at all. After a playground sign-in the provider lands on/oauth/callback, and our own callback copies its query and#part (?code=…&state=…or#access_token=…) onto the starting page, where the vendor tags read the address for themselves; the portal's own events are clean, sincepageLocation()strips both. 12.6.2'sgetReturnedFlow()reads the answer fromsessionStorageunderfumadocs-openapi-oauth-returnedand falls back to the address; 12.6.0 reads the address only, so the callback cannot store it there before the bump. With the bump:oauthCallback()returns the bare starting page and the route storeslocation.hrefunder that key, withtest/portal-template/oauth-callback.test.tsand the comment atoauth-callback.ts:20following.devpins 12.6.0 since feat(portal): fumadocs-openapi 12.6 with PKCE for playground sign-ins #472 merged on 2026-11-12, under the seven-day release-age rule inpnpm-workspace.yaml; 12.6.2 clears it on 2026-11-19, and this branch takes it then. The 5-second wait that held the tags back (step 1.3, commit1c616d56) was removed on 2026-11-12.portal.scripts). PR 2 makes--no-analyticsalso leave out theanalyticsandmarketingscripts, and updates the flag's help. Released alone, this PR would have the flag drop more in a later release than it does in this one.searchOutcome(),apiRequest()and the built-page checks.docs_plugin_install, nodocs_code_copy), an SDK download, a code copy from a page's block, the API reference's cURL sample and "Example Response", each with its language ID, a search with a pick and one without, a reopen without typing (nothing), a broken portal link (docs_not_foundwith the page left, plus its page view), and a playground request with a response and one that fails (client_error), withdocs_api_requestnever carrying the request address.Base
Was stacked on #472 (fumadocs-openapi 12.6) while it was open, as the plan's section 3 says, so the diff showed only this work. #472 merged on 2026-11-12 as
f7fe7dff, and this branch was rebased ontodevand retargeted the same day.An end-to-end run on 2026-11-13 (headless Chrome against
portal serveand a built portal, with a real GA4 property) saw seven of the eight events arrive once each with their keys, and the playground dialog marked in the browser; the plugin event needs a project with apluginblock. It also found the first page view inportal servecarrying the site name as its title, fixed in__root.tsxby reporting directly only when the router is idle at mount.Decisions on record: the e2e list's "a
--no-analyticsbuild holds no ID" check is not intest/e2e/portal-build.test.ts;test/infrastructure/portal-project-service.test.tschecks the identity file written from a config without vendors instead, which saves a full build in CI (plan step 1.9). The code review of 2026-11-13 and the reviewer's findings of the same day are addressed frombe8cc928on; the review of 2026-11-13 requesting changes from09a22ba3to96885d4d, with the OAuth return page in1bb6dfbb, but for a playground result that arrives after the reader closed the dialog, which Fumadocs gives no callback for in 12.6.0 or 12.6.2.The review of 2026-11-16 is addressed from
f54b30efto02200bbc, and settled four things customers will build on. The event names and keys are pinned by a test, as plain text apart from the interface, so a rename is a change made there on purpose.statusstays one text value, the HTTP status as digits orclient_error. A search'sresultis a whole address in the form ofpage_location.refleft the built-in kept names: no platform reads it from the address, sites use it for codes of their own, and a customer can add a name throughkeepQueryParamsbut cannot take one out. The same review found the not-found page reportingdocs_not_foundfor the page a reader left it for; fixed, and seen in headless Chrome on a served and a built portal.🤖 Generated with Claude Code