The DELETE_TYPE_TRANSACTION constant was 'transaction' (singular), so
the worker's inner switch on $document->getCollection() (which returns
'transactions') never matched and every transaction deletion fell to
the default branch, logging "No lazy delete operation available for
document of type: transactions". Renamed to DELETE_TYPE_TRANSACTIONS
to align with every other constant used in that switch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The sampler reads project.id from the span; realtime and VCS were storing it
under namespaced keys (realtime.project.id, vcs.github.event.repo.{id}.project.id),
causing all those spans to be dropped when _APP_TRACE_PROJECT_ID was set.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Span attribute keys are now snake_case with dots only for child
relationships. Worker span lifecycle moved to app/worker.php; selective
trace filtering moved to the exporter sampler in app/init/span.php so
handlers only call Span::add.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The prefix length 700 exceeded the resourceId/parentResourceId column
size of 255 (Database::LENGTH_KEY), crashing the container at startup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add skipFilters to Reports/Get.php (was the only endpoint still
triggering the full N+1 subquery cascade)
- Scale CTA batch limit dynamically (insightCount * MAX_CTA_COUNT)
instead of fixed APP_LIMIT_SUBQUERY to prevent silent truncation
- Revert deleteReport to callback-based pagination so CTAs are not
orphaned when a report has more than APP_LIMIT_SUBQUERY insights
- Add explicit prefix lengths (700) to _key_project_resource and
_key_project_parent_resource indexes to stay under InnoDB 3072-byte limit
- Validate CTA service/method against ADVISOR_CTA_SERVICES and
ADVISOR_CTA_METHODS enums in the CTAs validator
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix N+1 in Reports/XList (51→4 queries) via skipFilters + batch fetch
- Add skipFilters to Reports/Delete and cursor fetch (avoid loading all
nested insights/CTAs just for ownership check)
- Fix N+1 in deleteReport worker (flat CTA deletion instead of per-insight)
- Add advisor entity cleanup on project deletion (reports, insights, CTAs)
- Remove resourceInternalId, parentResourceInternalId, $permissions from
Insight response model (internal IDs leak DB internals, permissions unused)
- Remove dead subQueryInsightCTAs filter registration
- Remove stale enum-value comments from platform schema
- Fix _key_dismissedAt index to include projectInternalId
- Fix scope category from 'Other' to 'Advisor'
- Switch action base class from Utopia\Platform\Action to Appwrite\Platform\Action
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The resources/context migration replaced the resolver closure with a
first-class callable from `$swoole->context()->get(...)`. That
captures the result of `$swoole->context()` once -- on the first
dispatch -- and binds the bus to the first request's per-request
container forever after.
Wrap the lookup in a fresh closure so each dispatch re-resolves
`context()` against the current Swoole coroutine. Restores the
behavior of the pre-migration `getContainer()` form.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
EVENT_START and EVENT_TASK were each constructing `new Http($swoole, 'UTC')`
purely as a vehicle to reach DI -- never configuring routing, never calling
->run(). Now that the Swoole adapter exposes resources() directly, the
Http instance is no longer needed: the resources container we passed into
the Server constructor is already in scope as $container.
Drop the throwaway Http construction, pass $container straight through, and
update createDatabase() to take Container instead of Http to match. Also
rename $swooleAdapter to $swoole for consistency with the rest of the file.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adopts the new split DI containers in utopia-php/http: `resources()` for
boot-time wiring (shared across requests) and `context()` for per-request
state. Replaces the removed `getResource()`/`setResource()`/`getContainer()`
helpers throughout the HTTP entry point, controllers, GraphQL layer, and
installer.
Bumps the dependency chain accordingly: utopia-php/http to the dev branch
(aliased to 0.34.25 to satisfy platform's exact pin), servers 0.4.*,
queue 0.18.*, and pulled-along cli/platform/database upgrades.
Also tightens app/init/resources/request.php by collapsing single-return
factories to arrow functions.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The migrations worker (Appwrite→Appwrite migrations + CSV/JSON imports
& exports) reads `_APP_MIGRATION_HOST` to call back into this instance's
API. `_APP_MIGRATION_HOST` was introduced in #11229 (1.8.x / 1.9.x) but
was never added to `app/config/variables.php`, so `appwrite install` /
`appwrite upgrade` never write it into the generated `.env`. With the
var unset, the migrations worker fails — on a fresh self-hosted install
the first export hangs with no error (#11853). (Contributors and CI
don't hit it because the repo's hand-maintained `.env` already has
`_APP_MIGRATION_HOST=appwrite`; only the *installer-generated* `.env` is
missing it.)
Add `_APP_MIGRATION_HOST` to `app/config/variables.php` with default
`appwrite` — the API service name in the standard Docker Compose setup,
which is what the repo's own `.env` and the cloud Helm charts already
use, and what the migration endpoint was hardcoded to before #11229.
`appwrite install` and `appwrite upgrade` now write it into the
generated `.env`, so fresh installs and upgrades have it set and the
migration/import/export flows work.
Scope: this PR fixes the install & upgrade paths only — it deliberately
doesn't change the worker code.
Fixes#11853
The module's namespace and directory now match the top-level service
name (`advisor`) instead of one of its resource names (`insights`):
- src/Appwrite/Platform/Modules/Insights -> .../Modules/Advisor
- src/Appwrite/Insights -> src/Appwrite/Advisor
- tests/unit/Insights -> tests/unit/Advisor
- Route group label flipped from `'insights'` to `'advisor'`
- Section-header comments aligned
Resource names (`insights`, `reports`, `insightCTAs`) and the
`Insight*`/`Report` response models stay — those are the resources the
service exposes, not the service itself.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`getDatabasesDB` used `??` to fall back from a database doc's `database`
attribute to the project DSN, but `??` only triggers on null. Migration
destinations end up with an empty-string `database` (the value is copied
from the source DB but isn't a valid DSN on the destination's pool),
which slipped past the fallback and surfaced as a 500 with
`new DSN('mysql://')` in the catch block. Use elvis so empty strings
fall back too.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Resolves conflict in app/init/constants.php (kept 4327 cache buster, took 1.9.4 stable version).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
target stores free-form URLs or resource IDs. URLs in the wild can
exceed the prior 2048-char cap, so switch the column to VAR_TEXT
(65535). The _key_project_target index already declares an explicit
700-char prefix length, so indexing still works on both MariaDB and
MongoDB.
Bump APP_CACHE_BUSTER for the schema change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Switch markdown summaries, JSON CTA params, and the embedded
insights/ctas subQuery payloads from sized VAR_STRING to VAR_TEXT.
None of these fields are indexed and they all hold free-form data, so
removing the artificial size cap means we never have to revisit the
limit when an analyzer surfaces a longer report.
- reports.summary, insights.summary: markdown, no upper bound enforced
- reports.insights, insights.ctas: subQuery virtuals carrying full
child rows
- insightCTAs.params: JSON parameter blob for arbitrary CTA actions
Bump APP_CACHE_BUSTER for the schema change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Register dedicated reports.write scope and switch deleteReport to it
so cloud can issue narrowly-scoped delete keys without granting
insights.write.
- Make insights.parentResourceInternalId optional with null default to
match its companion parentResourceType/parentResourceId fields and
unblock insights with no parent (e.g. database-level performance
insights).
- Tighten Insight.reportId model description: insights always belong to
a report, ad-hoc insights are not supported.
- Add reports.write to default test API key and admin role so existing
e2e tests using serverHeaders() can hit the delete endpoint.
- Bump APP_CACHE_BUSTER for the schema change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Jake Barnby
Chirag Aggarwal
loks0n
Matej Bačo
copilot-swe-agent[bot]
Atharva Deosthale
Prem Palanisamy
Matej Bačo