iframe-proxy

ArnabChatterjee20k · 2026-10-16T06:28:58Z

Includes #14586, merged into this branch: targetless users reach users/<userId> through the enabled Appwrite provider whatever the default, and message.providers lists only providers a send used. See #14586 for its description and tests.

What does this PR do?

Topic campaigns sent through the Appwrite push (MQTT) provider were published on each subscribed target's identifier instead of the topic ID.

streamRecipients keyed topic pages by target identifier and dropped the topic ID. Topic sends skip the per-user branch, so recipients = array_keys($identifiers). The Appwrite adapter then treats each to as a topic ID (Push/Appwrite.php persist() → getDocument('topics', $to)). The result:

Delivery only worked when the target was created with identifier = <topicId>, which is what the e2e fixture did.
Any other identifier, such as ID.unique(), failed with Topic <identifier> does not exist and counted as a failed delivery.
A user in N topics needed N Appwrite targets.

Fix

streamRecipients now yields each page with its topic ID (null for user, target or email addressed pages) instead of a perUser bool.
A topic page on an Appwrite push provider sends to [$topicId]. The subscribed Appwrite targets only mark that the topic has MQTT subscribers.
Each topic is published once per message, across pages and across multiple Appwrite providers.

What each recipient source yields

$perUser is now $topicId === null. Only topic pages carry an ID:

For an Appwrite provider page:

$perUser true: sends to users/<userId>, unchanged.
$perUser false: sends to [$topicId] (new).

FCM and APNs pages still send to device tokens. A mixed send such as topics: ['news'], users: ['U'] routes each page by its own value.

One Appwrite target with any identifier now covers every topic it is subscribed to. Targets that already use identifier = topicId keep working.

Test Plan

E2E in tests/e2e/Services/Mqtt/MqttServerTest.php:

The shared setupPushTopic fixture now uses a placeholder identifier, so every existing campaign, replay and QoS test runs against the fixed routing.
testCampaignReachesEveryTopicOfOneTarget: one target subscribed to two topics gets the message exactly once on each topic ID, and the message ends sent.
testCampaignPublishesTopicOncePerMessage: two users on two different Appwrite providers each get the message exactly once.
testCampaignIsNotDeliveredOnTargetIdentifier: subscribing to the target identifier is refused, and the only delivery is on the topic ID.

.github/ci/e2e.json: Workers/Messaging.php and Messaging/Adapter/Push/ now also trigger the Mqtt lane.

Locally I ran php -l, Pint and PHPStan on the changed files. PHPStan reports no new errors; the only ones are the Utopia\Database\Id not-found errors that main also shows because the local vendor is stale. E2E was not run locally; the CI e2e lanes ran it.

CI

The earlier rust-toolchain repin is superseded by #14577 on main. The merge from main takes its v1 pin, so this PR no longer changes that workflow.

Related PRs and Issues

Follow-up to Appwrite Native Push Broker #13372 and feat(console): Appwrite push provider behind appwritePush flag #14532

🤖 Generated with Claude Code

A topic campaign sent through the Appwrite push provider was published on each subscribed target's identifier rather than the topic id, so it only reached devices when the target identifier happened to equal the topic id. Any other identifier failed with "Topic <identifier> does not exist". Carry the topic id through the recipient stream and publish an Appwrite topic page on it, once per message, so one Appwrite target with any identifier covers every topic it subscribes to. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tenki-reviewer · 2026-10-16T06:29:05Z

Review complete. No blocking issues — approved ✅; 1 nitpick below.

🧹 Nitpicks (1) — 🟢 1 low

🟢 expected+1 consume always burns full timeout (MqttServerTest.php:869) — MqttSubscriber::consume() loops until $limit is reached or the deadline expires, so asking for one more than the expected deliveries makes every successful run block the whole timeout: testCampaignReachesEveryTopicOfOneTarget waits a guaranteed 10s (tests/e2e/Services/Mqtt/MqttServerTest.php:869) and testCampaignPublishesTopicOncePerMessage waits 10s plus 5s sequentially (lines 913-914).

This PR reworks recipient streaming in the messaging worker so each yielded page carries its topic id, routing per-user and topic campaigns to the correct MQTT wire topic with a dedup map guarding against double publishes, and adds MQTT e2e tests plus CI lane wiring.

Files	Change
`src/Appwrite/Platform/Workers/Messaging.php`	Generator contract changes from bool to `?string` topic id; per-topic publish dedup across streamed recipient pages.
`tests/e2e/Services/Mqtt/MqttServerTest.php`	New e2e helpers (topic, target, subscriber) and duplicate-delivery assertions against a live MQTT broker.
`.github/ci/e2e.json`	Registers the MQTT service so its e2e lane runs in CI.

_{Reviewed commit: f864fa7}

hansi-codes · 2026-10-16T06:30:32Z

🟢 Tier S · Ready to merge

The incremental changes have no actionable issues.

Messaging recipient pages retain topic IDs so Appwrite push campaigns publish to each topic ID once per message rather than to target identifiers. The worker also routes targetless users through an enabled Appwrite provider when another push provider is the default, and builds message search terms from providers actually used. MQTT e2e coverage exercises topic routing and provider selection, the CI path map includes messaging and push-adapter changes, and the console snippets workflow repins its Rust toolchain action.

Latest changes: The new commits route targetless user sends through enabled Appwrite even when it is not the default, track only providers actually used in message search, and add MQTT e2e cases for enabled and disabled Appwrite-provider selection.

Verdict	New comments	Fixed	Still open
✅ Approved	0	0	0

📂 Walkthrough · 4

File	Change
`.github/ci/e2e.json`	Routes messaging worker and push-adapter changes to the MQTT e2e lane.
`.github/workflows/console-snippets-qa-server.yml`	Repins the Rust toolchain action to the current stable-branch commit.
`src/Appwrite/Platform/Workers/Messaging.php`	Routes topic campaigns by topic ID, handles targetless users with a non-default enabled Appwrite provider, and records only providers used by delivery batches.
`tests/e2e/Services/Mqtt/MqttServerTest.php`	Covers topic-ID delivery, multi-topic and multi-provider deduplication, and targetless-user routing with and without an enabled Appwrite provider.

_{Reviewed the commits since ea41069 · Details · Comment @hansi-codes review to re-run, or mention @hansi-codes with a question.}

hansi-codes

🟢 Tier S · Looks good to merge. Summary

github-actions · 2026-10-16T06:42:56Z

Security rules

No new WARNING or ERROR findings from security rules.

19 existing findings tracked in .semgrep/baseline.json

php.appwrite.guest-write-without-abuse-limit (12)
php.appwrite.secret-compare-timing (5)
php.appwrite.weak-secret-env-default (2)

Posted by Checks / Rules. Re-runs update this comment in place. Rule details and baseline: .semgrep/README.md.

github-actions · 2026-10-16T06:50:49Z

✨ Benchmark results

Comparing main (before) → fix/mqtt-topic-send-by-topic-id (after).

Metric	Before	After	Change
🚀 Requests/sec	`205.24`	`222.75`	🟢 `+8.5%`
⏱️ Latency P50	`82.64 ms`	`78.04 ms`	🟢 `-5.6%`
⏱️ Latency P95	`192.76 ms`	`179.82 ms`	🟢 `-6.7%`

Per-scenario breakdown & investigation details

Metrics below reflect the current branch (after). Δ P95 compares against the base.

Scenario	P50 (ms)	P95 (ms)	Requests	RPS	Δ P95 (ms)
API total	`78.04`	`179.82`	`13,680`	`222.75`	`-12.94`
Account	`145.5`	`281.96`	`720`	`12.38`	`-25.64`
TablesDB	`77.18`	`148.35`	`7,440`	`123.37`	`-6.14`
Storage	`67.83`	`157.65`	`3,600`	`61.67`	`-23.99`
Functions	`107.97`	`204.05`	`1,920`	`33.61`	`-12.62`

Top API waits (after)

dtolnay/rust-toolchain rewrote its stable branch, leaving the pinned commit outside the repo's history, which fails the zizmor workflow audit on main. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…isn't the default A users[] push reached a user with no push target on users/<userId> only when the default (first enabled) push provider was Appwrite. With FCM or APNs set up first, the message failed with "No valid recipients found." even though an Appwrite provider was enabled. Resolve the enabled Appwrite push provider once per send and use it for the targetless fallback, whichever provider is the default. Users holding other push targets are unchanged, so no notification is sent twice. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The worker appended every provider in its resolved-provider cache to the message's search text and the message.providers span. That cache is seeded with the default provider, and now the enabled Appwrite provider, so a send delivered entirely through another provider was still searchable by providers it never used. Record the provider of each batch result, wake-ups included, and build the search and span from that set. The cache keeps its seeding. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…akes unobservable The messageSearch attribute filter recomputes a message's search text on every write from its id, description, status, title and type, so provider names are never stored and cannot be asserted through the search API. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…te-provider fix(messaging): reach targetless users through Appwrite push when it isn't the default

…y-topic-id # Conflicts: # .github/workflows/console-snippets-qa-server.yml

… into fix/mqtt-topic-send-by-topic-id

hansi-codes Bot approved these changes Oct 16, 2026

View reviewed changes

tenki-reviewer Bot approved these changes Oct 16, 2026

View reviewed changes

fix(ci): repin rust-toolchain to its rewritten stable branch

ea41069

dtolnay/rust-toolchain rewrote its stable branch, leaving the pinned commit outside the repo's history, which fails the zizmor workflow audit on main. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

ArnabChatterjee20k mentioned this pull request Oct 16, 2026

fix(messaging): reach targetless users through Appwrite push when it isn't the default #14586

Merged

ArnabChatterjee20k and others added 2 commits October 16, 2026 13:16

ArnabChatterjee20k mentioned this pull request Oct 16, 2026

docs: announce Appwrite Push #14588

Draft

4 tasks done

ArnabChatterjee20k and others added 2 commits October 16, 2026 13:34

Merge pull request #14586 from appwrite/fix/mqtt-user-fallback-appwri…

59046f8

…te-provider fix(messaging): reach targetless users through Appwrite push when it isn't the default

ArnabChatterjee20k mentioned this pull request Oct 16, 2026

fix(account): route APNs push targets by bundle ID and environment #14589

Merged

ArnabChatterjee20k added 2 commits October 16, 2026 14:22

Merge remote-tracking branch 'origin/main' into fix/mqtt-topic-send-b…

4325e54

…y-topic-id # Conflicts: # .github/workflows/console-snippets-qa-server.yml

Merge remote-tracking branch 'origin/fix/mqtt-topic-send-by-topic-id'…

96beb0c

… into fix/mqtt-topic-send-by-topic-id

ArnabChatterjee20k merged commit c3abfba into main Oct 16, 2026
66 of 67 checks passed

ChiragAgg5k mentioned this pull request Oct 21, 2026

chore: prepare 2.6.0 release #14730

Merged

6 tasks

Recipient source in `streamRecipients`	Yields	`$perUser`
`topics: [...]`: subscriber pages for each topic	`[$page, $topic->getId()]`	false
`users: [...]`: those users' push targets	`[$page, null]`	true
`users: [...]` with no live target: fallback onto `users/<id>` (only when the default provider is Appwrite)	`[[$default->getId() => $identifiers], null]`	true
`targets: [...]`: explicitly named targets	`[$page, null]`	true
Raw `emails` (email messages only)	`[[$default->getId() => ...], null]`	true

Sunbelt Computer Software

PL/B Language Development and Support

Navigation Menu

Repository navigation

fix(messaging): deliver Appwrite push topic campaigns on the topic id - #14580

fix(messaging): deliver Appwrite push topic campaigns on the topic id#14580
ArnabChatterjee20k merged 8 commits into
mainfrom
fix/mqtt-topic-send-by-topic-id

ArnabChatterjee20k commented Oct 16, 2026 •

edited

Loading

Uh oh!

tenki-reviewer Bot commented Oct 16, 2026 •

edited

Loading

Uh oh!

hansi-codes Bot commented Oct 16, 2026 •

edited

Loading

Uh oh!

hansi-codes Bot left a comment

Uh oh!

github-actions Bot commented Oct 16, 2026

Uh oh!

github-actions Bot commented Oct 16, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Sunbelt Computer Software

PL/B Language Development and Support

Conversation

ArnabChatterjee20k commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Fix

What each recipient source yields

Test Plan

CI

Related PRs and Issues

Uh oh!

tenki-reviewer Bot commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hansi-codes Bot commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🟢 Tier S · Ready to merge

Uh oh!

hansi-codes Bot left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Oct 16, 2026

Security rules

Uh oh!

github-actions Bot commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✨ Benchmark results

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

ArnabChatterjee20k commented Oct 16, 2026 •

edited

Loading

tenki-reviewer Bot commented Oct 16, 2026 •

edited

Loading

hansi-codes Bot commented Oct 16, 2026 •

edited

Loading

github-actions Bot commented Oct 16, 2026 •

edited

Loading