Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Joe Li <joe@preset.io>
BigQuery rejects filter values containing apostrophes (e.g. O'Brien): the
sqlalchemy-bigquery dialect renders string literals via repr(), which switches
to double-quote delimiters that BigQuery parses as identifiers, causing a
syntax error.
Monkey-patch the dialect's colspecs with a TypeDecorator whose literal_processor
emits single-quoted literals using backslash escaping ('O\'Brien'). Doubled
single quotes ('O''Brien') are NOT valid in BigQuery (parsed as concatenated
literals). Control characters are emitted as named escapes with a \xhh fallback,
since BigQuery forbids literal control chars in quoted strings. Follows the
existing Databricks dialect pattern.
Fixes#35857
# zizmor: ignore[adhoc-packages] - supersetbot is a first-party Apache CLI (apache-superset/supersetbot) installed globally as a tool; a global CLI install has no application manifest/lockfile context
# zizmor: ignore[adhoc-packages] - installs the locally packed supersetbot tarball built from the trusted apache-superset/supersetbot checkout; no lockfile applies to a global CLI install
run:|
# simple trick to install globally with dependencies
--title "Scheduled Docker image refresh failed for ${LATEST_RELEASE}" \
--label "infra:container" \
--label "bug" \
--body "The weekly Docker base-image refresh failed for release \`${LATEST_RELEASE}\`. Published images may be missing upstream base-layer security patches until this is resolved.
description:'Use Cypress Dashboard (true/false) [paid service - trigger manually when needed]. You MUST provide a branch and/or PR number below for this to work.'
description:"Use Cypress Dashboard (true/false) [paid service - trigger manually when needed]. You MUST provide a branch and/or PR number below for this to work."
and please read our [Slack Community Guidelines](https://github.com/apache/superset/blob/master/CODE_OF_CONDUCT.md#slack-community-guidelines)
- [Join our dev@superset.apache.org Mailing list](https://lists.apache.org/list.html?dev@superset.apache.org). To join, simply send an email to [dev-subscribe@superset.apache.org](mailto:dev-subscribe@superset.apache.org)
- If you want to help troubleshoot GitHub Issues involving the numerous database drivers that Superset supports, please consider adding your name and the databases you have access to on the [Superset Database Familiarity Rolodex](https://docs.google.com/spreadsheets/d/1U1qxiLvOX0kBTUGME1AHHi6Ywel6ECF8xk_Qy-V9R8c/edit#gid=0)
- Join Superset's Town Hall and [Operational Model](https://preset.io/blog/the-superset-operational-model-wants-you/) recurring meetings. Meeting info is available on the [Superset Community Calendar](https://superset.apache.org/community)
@@ -242,16 +247,13 @@ Understanding the Superset Points of View
@@ -109,7 +109,7 @@ If yes, it is in scope. If no, it is out of scope. The lists below apply that te
- Any action an Admin role can perform through documented configuration, API, or UI. The Admin role is a trusted operational principal by policy. Per MITRE CNA Operational Rules 4.1, a qualifying vulnerability must violate a security policy; behavior within a documented trust boundary does not.
- Deployment or operator decisions: the values of secrets and tokens, whether internal networks are reachable from the server, which database connectors or cache backends are enabled, which feature flags are set, where notifications are delivered, and which third-party plugins are loaded.
- Compromise, modification, or malicious control of trusted backend infrastructure. Apache Superset assumes the integrity of its metastore, cache backends (for example Redis or Memcached), message brokers, secret stores, and other operator-managed infrastructure. Findings that require an attacker to read from, write to, or otherwise tamper with these systems, including injecting malicious state, serialized objects, cache entries, task metadata, configuration, or database records, are post-compromise scenarios and do not constitute vulnerabilities in Apache Superset itself. A finding remains in scope only if an unprivileged user can cause such modification through a vulnerability in Apache Superset.
-Code paths whose intended purpose is example data, demos, fixtures, local development, or documentation, rather than the production runtime.
-The continued presence of expired key-value or metastore-cache entries that have not yet beendeleted from the metadata database. Such entries are excluded from reads once expired, are purged opportunistically on write, and are removed in bulk by the scheduled `prune_key_value` maintenance task; their lingering until purged is an eventual-cleanup property, not a security boundary, and does not constitute a vulnerability.
- How a downstream application (spreadsheet program, email client, browser handling user-downloaded files) interprets output Apache Superset produced for it.
- Findings without a reproducible proof of concept against a supported release. The burden of demonstrating exploitability rests with the reporter; findings closed for lack of a proof of concept may be refiled if one is later produced.
- Brute force, rate limiting, denial of service, or resource exhaustion that does not bypass a documented control.
@@ -24,6 +24,293 @@ assists people when migrating to a new version.
## Next
- [39925](https://github.com/apache/superset/pull/39925): URL prefixing for `SUPERSET_APP_ROOT` subdirectory deployments is now handled automatically by helpers in `src/utils/navigationUtils` (`openInNewTab`, `redirect`, `getShareableUrl`, `<AppLink>`). Direct imports of `ensureAppRoot` / `makeUrl` from `src/utils/pathUtils` are forbidden outside `navigationUtils.ts` (enforced by a static-invariant test); contributors writing new code should use the focused helpers instead. No runtime behaviour change for existing callers — all 19 prior call sites have been migrated and four pre-existing double-prefix and missing-prefix bugs are fixed as part of the migration.
- [39925](https://github.com/apache/superset/pull/39925): `SupersetClient.getUrl()` now strips a single leading application-root segment from the supplied `endpoint` before building the request URL, so a caller that accidentally pre-prefixes its endpoint (for example by wrapping it with `ensureAppRoot` before passing it to the client) no longer produces a doubled `/superset/superset/...` URL under subdirectory deployment. The strip is **single-pass** — a genuine `/superset/superset/<slug>` route is preserved, not collapsed — and **silent** (no console warning); the static-invariant test remains the primary signal for pre-prefixing at the call site, and this runtime strip is a safety net beneath it. Code that intentionally targeted a literal `/<app_root>/<app_root>/...` endpoint through `getUrl` (a configuration that has no legitimate use under the prefixing model) would have its first redundant segment removed.
- **Breaking — `Superset` view class route prefix removed.** The `Superset` view in `superset/views/core.py` now declares `route_base = ""`, overriding Flask-AppBuilder's auto-derived `/superset` prefix. Routes that previously lived at `/superset/welcome/`, `/superset/dashboard/<id>/`, `/superset/dashboard/p/<key>/`, `/superset/explore/`, etc. now respond at `/welcome/`, `/dashboard/<id>/`, `/dashboard/p/<key>/`, `/explore/`, etc. Under subdirectory deployment (`SUPERSET_APP_ROOT=/superset`) the URLs are unchanged from end-user perspective — `AppRootMiddleware` re-applies the prefix via `SCRIPT_NAME`. Under root deployments, any external integration or bookmark that hard-codes `/superset/<endpoint>/` paths must be updated to drop the prefix. This fixes the doubled `/superset/superset/...` URLs that `url_for` emitted for these endpoints under subdirectory deployment and the related 404s on the routes themselves.
- **Breaking — Three sibling view classes route prefix removed.** Following the same rationale as the `Superset` class above, `ExplorePermalinkView` (`superset/views/explore.py`), `TagModelView`, and `TaggedObjectsModelView` (`superset/views/tags.py`, `superset/views/all_entities.py`) now mount at the application root rather than a hard-coded `/superset/...`. The user-visible URLs `/superset/explore/p/<key>/`, `/superset/tags/`, and `/superset/all_entities/` are unchanged under subdirectory deployment; under root deployments these views now serve `/explore/p/<key>/`, `/tags/`, and `/all_entities/`, so any external integration or bookmark must drop the `/superset/` prefix. `Dashboard.url` and `Dashboard.get_url` likewise return `/dashboard/<id>/` instead of the prior `/superset/dashboard/<id>/` literal so downstream consumers (DashboardList row hrefs, MCP service `dashboard_url`) emit a single, deployment-correct prefix.
- **Legacy `/superset/*` path support.** A new outermost WSGI middleware `LegacyPrefixRedirectMiddleware` (`superset/middleware/legacy_prefix_redirect.py`) 308-redirects every enumerated legacy `/superset/<canonical>` path to its post-`route_base=""` canonical location (e.g. `/superset/welcome/` → `/welcome/` under root; → `/superset/welcome/` under `SUPERSET_APP_ROOT=/superset`, because the canonical resolves through `AppRootMiddleware`). Bookmarks, email links, and external integrations survive the route-base collapse for one release cycle. POST against a GET-only canonical returns 410 Gone instead of 308 (308 would 405 on retry). The shim is removed at EOL `5.0.0`, matching the `@deprecated(eol_version="5.0.0")` gate on `Superset.explore` and `Superset.explore_json`.
- **PWA web app manifest served dynamically.** The PWA manifest is now served at `/pwa-manifest.json` (under `APPLICATION_ROOT`) by a new `PwaManifestView` (`superset/views/pwa_manifest.py`) instead of the static file at `/static/assets/pwa-manifest.json`. The legacy static source at `superset-frontend/src/pwa-manifest.json` has been removed (along with its `webpack.config.js``CopyPlugin` rule). The new endpoint resolves `APPLICATION_ROOT` and `STATIC_ASSETS_PREFIX` at request time so PWA install works under subdirectory deployments and split static-prefix / app-root deployments (where `STATIC_ASSETS_PREFIX` points to a CDN host while the Superset backend stays under `APPLICATION_ROOT`). The `<link rel="manifest">` href in `superset/templates/superset/spa.html` was updated correspondingly (using a new `application_root_rstrip` template global). Operators with a forked `spa.html` should switch any manifest `<link>` to `{{ application_root_rstrip }}/pwa-manifest.json`.
- **Hard re-bookmark break — `/superset/sql/<database_id>/`.** SQL Lab moved to its own blueprint at `/sqllab/`. The legacy `/superset/sql/<id>/` shape changed to a query-string form (`/sqllab/?dbid=<id>`); no 1:1 path mapping exists, so `LegacyPrefixRedirectMiddleware` does **not** redirect this route — it passes through and surfaces a 404. Users with bookmarks to `/superset/sql/<id>/` must update them to `/sqllab/?dbid=<id>`.
- **`SqlaTable.sql_url` query-string format.** `SqlaTable.sql_url` now URL-encodes `table_name` and joins it as a query parameter rather than concatenating a second `?`. Previously, with `Database.sql_url` returning `/sqllab/?dbid=<id>`, the concatenation produced `/sqllab/?dbid=<id>?table_name=<raw>` — a malformed second `?` that broke the query parser. External code that parsed the legacy `<base>?table_name=<raw>` shape now sees properly percent-encoded values (e.g. `/` → `%2F`, ` ` → `+` or `%20`); decode with `urllib.parse.parse_qsl`.
- **New config flag `EMBEDDED_DISABLE_PERMALINK_ORIGIN_REWRITE` (default `False`).** Share/permalink URLs now substitute `window.location.origin` for the backend-supplied origin so a proxied or subdirectory-deployed Superset never hands the user an unreachable internal hostname. Operators whose reverse proxy correctly forwards `X-Forwarded-Host`*and* who want permalinks to carry the backend's literal origin can opt out by setting `EMBEDDED_DISABLE_PERMALINK_ORIGIN_REWRITE = True` in `superset_config.py`. Default `False` (rewrite is on); flipping the default would regress the dominant proxied/subdir deployment to an unreachable host.
### SQL Lab denies large-object and information_schema access by default
`DISALLOWED_SQL_FUNCTIONS` and `DISALLOWED_SQL_TABLES` now ship with additional default entries, so SQL Lab and chart-data queries that reference them are rejected where they were previously allowed:
- PostgreSQL large-object routines (`lo_from_bytea`, `lo_export`, `lo_import`, `lo_put`, `lo_create`, `lo_creat`, `lowrite`, `lo_get`, `loread`, `lo_unlink`), which read and write bytes on the database server's filesystem.
- The SQL-standard `information_schema` views (`tables`, `columns`, `routines`, `views`, the privilege/grant views, etc.), which expose table, column, privilege, and view-definition metadata across the whole database.
Deployments that legitimately query these (for example tooling that introspects `information_schema`) can restore the previous behavior by overriding `DISALLOWED_SQL_FUNCTIONS` / `DISALLOWED_SQL_TABLES` in `superset_config.py` to drop the entries they need.
Because the denylist now resolves the effective schema through the query-aware path, PostgreSQL queries that change the `search_path` (e.g. `SET search_path = ...`) are rejected on the SQL Lab execution and cost-estimate paths whenever any `DISALLOWED_SQL_TABLES` entry is configured (the default for PostgreSQL), matching the behavior previously applied only when `RLS_IN_SQLLAB` was enabled.
### SQL parser input length cap (SQL_MAX_PARSE_LENGTH)
The SQL parser now rejects scripts whose UTF-8 byte length exceeds the new
`SQL_MAX_PARSE_LENGTH` config option (default `1_000_000` bytes) before they are
handed to sqlglot, which bounds parser memory and CPU usage. A single query
larger than the cap (for example a very large `IN (...)` list or a big
virtual-dataset SQL) raises a parse error in SQL Lab and dashboard-generated
queries. Deployments that legitimately run queries above this size should raise
the value, and `SQL_MAX_PARSE_LENGTH = None` disables the check entirely.
### Guest-token RLS rules reject unknown fields
The `rls` rules passed to `POST /api/v1/security/guest_token/` are now validated strictly: a rule may only contain `dataset` and `clause`. Previously unknown fields were silently dropped, so a mistyped or legacy scope key (most commonly `datasource` instead of `dataset`) produced a rule with no `dataset`, which is treated as a *global* rule applied to every dataset the embedded resource can reach. Such a request now returns HTTP 400 identifying the offending field instead of issuing a token with an unintended global rule. Integrators that were sending extra fields in RLS rules must remove them; valid dataset-scoped (`{"dataset": 41, "clause": "..."}`) and global (`{"clause": "..."}`) rules are unaffected.
### MCP service requires `MCP_JWT_AUDIENCE` when JWT auth is enabled
When the MCP service has JWT auth enabled (`MCP_AUTH_ENABLED = True`), an audience must be configured via `MCP_JWT_AUDIENCE` so issued tokens are bound to this service. The service now fails to start with a clear configuration error when the audience is unset, instead of starting with audience validation skipped. Deployments that enable MCP JWT auth must set `MCP_JWT_AUDIENCE` to the audience value their identity provider issues for the MCP service. API-key-only MCP deployments (JWT auth disabled) are unaffected.
### Swagger UI is opt-in (off by default)
`FAB_API_SWAGGER_UI` now defaults to `False` and is driven by the `SUPERSET_ENABLE_SWAGGER_UI` environment variable. The interactive Swagger UI / OpenAPI documentation endpoints (e.g. `/swagger/v1`) are therefore no longer exposed by default. To enable them, set `SUPERSET_ENABLE_SWAGGER_UI=true` (the bundled Docker development environment sets this) or override `FAB_API_SWAGGER_UI = True` in `superset_config.py`.
### Build details (git SHA / build number) are admin-only by default
The git SHA and build number surfaced in the "About" section, the bootstrap payload, and the public `/version` endpoint are now only included for admin users by default; the release version string is still shown to everyone. To expose the build details to all users (the previous behavior), set the `SUPERSET_EXPOSE_BUILD_DETAILS` environment variable (or `EXPOSE_BUILD_DETAILS_TO_USERS = True` in `superset_config.py`).
### Pivot table First/Last aggregations follow data order
The pivot table chart's `First` and `Last` aggregations now return the first and last value in data (query result) order, instead of effectively returning the minimum and maximum. Existing pivot tables that use these aggregations for totals/subtotals may show different values after upgrading. For deterministic results, ensure the underlying query has a stable sort order.
### `FetchRetryOptions` callback parameters widened to allow `null`
The `error` and `response` parameters of the `retryDelay` and `retryOn` callbacks in `FetchRetryOptions` (exported from `@superset-ui/core`) are now typed `Error | null` and `Response | null` to match the actual call-site signature provided by `fetch-retry`. Because these parameter types are contravariant, consumers who typed their callbacks with the non-nullable `(attempt: number, error: Error, response: Response) => number` will get a TypeScript compile error. Widen your callback signatures to accept `Error | null` / `Response | null`.
### `thumbnail_url` removed from dashboard list API response
The `thumbnail_url` field has been removed from `GET /api/v1/dashboard/` list responses. External consumers relying on this field must now construct the thumbnail URL client-side using `id` and `changed_on_utc`:
The thumbnail endpoint redirects to the current digest URL regardless of whether the supplied digest is exact. If the image is not yet cached, that digest URL may return `202` and trigger async generation. Using `changed_on_utc` as the digest is sufficient for cache-busting purposes.
### Tagging fix for `create_all`-bootstrapped schemas
Only affects deployments whose metadata schema was created with SQLAlchemy's `create_all` (rather than `superset db upgrade`) on a foreign-key-enforcing backend — PostgreSQL, or MySQL with `FOREIGN_KEY_CHECKS=1`. Such schemas carry three invalid foreign keys on `tagged_object.object_id` that break tagging (`TAGGING_SYSTEM = True`) with a `ForeignKeyViolation`. Schemas built via `superset db upgrade` are unaffected.
This release stops the ORM from emitting these constraints, but it cannot drop ones already present in your schema. If affected, drop them manually (names vary by backend, so look them up first):
```sql
-- PostgreSQL: names are typically tagged_object_object_id_fkey, _fkey1, _fkey2
### Webhook alerts/reports block private/internal hosts by default
Webhook alert/report dispatch (`WebhookNotification.send`) now validates the target URL's host against the same private/internal-IP block applied to dataset import URLs. If the resolved host is in a loopback, link-local, private (RFC-1918), shared-CGNAT, or multicast range, the webhook is rejected with `NotificationParamException`.
Deployments that intentionally point webhooks at internal targets (chatops bridges, internal automation servers, on-premises Mattermost/Rocket.Chat, etc.) can opt out by setting `ALERT_REPORTS_WEBHOOK_ALLOW_INTERNAL_HOSTS = True` in `superset_config.py`. This mirrors the existing `DATASET_IMPORT_ALLOW_INTERNAL_DATA_URLS` opt-out for dataset imports.
### Impala cancel_query blocks private/internal hosts by default
The Impala engine spec's `cancel_query` issues an HTTP request from the Superset backend to the host configured on the Impala database connection. That host is now validated before the request: if it resolves to a private/internal IP range, the cancel call is refused and a warning is logged. Operators whose Impala cluster runs on an internal network can opt out by setting `IMPALA_CANCEL_QUERY_ALLOW_INTERNAL_HOSTS = True` in `superset_config.py`. This mirrors the dataset-import and webhook opt-out flags.
### Map chart renderer and OpenStreetMap migration behavior
The MapLibre migration for deck.gl charts preserves saved non-Mapbox styles on
the MapLibre-compatible path. Saved styles such as OpenStreetMap, `tile://`
tile templates, generic HTTPS style URLs, and charts without a saved style are
not reclassified as Mapbox during migration and do not require
`MAPBOX_API_KEY` only because of the migration.
Saved true Mapbox styles whose value starts with `mapbox://` remain
Mapbox-backed. If a Superset deployment does not configure `MAPBOX_API_KEY`,
those saved Mapbox charts keep the existing missing-key message instead of
silently falling back to MapLibre or another provider. In Explore, deck.gl and
point-cluster renderer controls preserve saved Mapbox state, but the Mapbox
choice is not available as a new working renderer without a configured key.
The MapLibre style choices include `Streets (OSM)`, backed by
`https://tile.openstreetmap.org/{z}/{x}/{y}.png`. This OpenStreetMap tile
be used through normal browser map tile requests and caching; it is not intended
for bulk prefetch or offline tile downloads.
### Password complexity policy enabled by default
Superset now ships a default password-complexity policy, enforced (via Flask-AppBuilder) across self-registration, the user create/edit/reset forms, and the User REST API. The policy requires a minimum password length of 8 characters and rejects a built-in blocklist of common/guessable passwords.
This is enabled by default (`FAB_PASSWORD_COMPLEXITY_ENABLED = True`), so new or reset passwords that are too short or appear in the blocklist will be rejected where they were previously accepted. Existing stored passwords are unaffected until they are next changed.
Operators can tune or disable the policy via config:
### Data uploads bounded by UPLOAD_MAX_FILE_SIZE_BYTES
Single data-file uploads (CSV, Excel, columnar) are now bounded by the `UPLOAD_MAX_FILE_SIZE_BYTES` config option, which defaults to `100 * 1024 * 1024` (100 MB). Files larger than this are rejected with a `413` before their contents are buffered into memory. Set `UPLOAD_MAX_FILE_SIZE_BYTES = None` to disable the check and restore unbounded uploads.
### Currency symbol position follows the locale when unset
When a chart's currency control leaves the **Prefix or suffix** field empty, the currency symbol position is now derived from the deployment locale's own convention via `Intl.NumberFormat` instead of always defaulting to a suffix. For example, under the default `en-US` locale `USD`, `GBP`, and `EUR` render as a prefix (`$ 1,000`), while eurozone locales such as `fr-FR` render `EUR` as a suffix (`1 000 €`). An explicit Prefix/Suffix selection is always honored and is unaffected.
Charts that relied on the previous always-suffix default for an unset position will render the symbol on the locale-appropriate side instead; set the position explicitly on the metric's currency control to pin it.
### Duration formatter precision
The `DURATION` number formatter now uses `Intl.DurationFormat` for locale-aware output. By default, sub-second fields are omitted, so values that previously displayed fractional seconds with `pretty-ms`, such as `10500` milliseconds rendering as `10.5s`, now render as `10s`.
To preserve sub-second precision in custom duration formatters, enable `formatSubMilliseconds`.
### Cache warmup authenticates via SUPERSET_CACHE_WARMUP_USER
The `cache-warmup` Celery task now drives a real WebDriver session for reliable authentication and reads the user to authenticate as from the new `SUPERSET_CACHE_WARMUP_USER` config option. It no longer consults `CACHE_WARMUP_EXECUTORS` for the warmup path. `SUPERSET_CACHE_WARMUP_USER` defaults to `None`, so the task fails fast with a clear message until you set it. Operators who previously relied on `CACHE_WARMUP_EXECUTORS` for cache warmup must set `SUPERSET_CACHE_WARMUP_USER` to a dedicated least-privilege user with access to the dashboards they want warmed up before the next warmup run.
### YDB now uses a native sqlglot dialect
YDB SQL parsing now relies on the dedicated [`ydb-sqlglot-plugin`](https://pypi.org/project/ydb-sqlglot-plugin/) dialect, which registers itself with sqlglot automatically. YDB users must install this plugin (e.g., via `pip install "apache-superset[ydb]"`) to avoid a `ValueError` when Superset parses YDB queries.
### Embedded dashboards enforce configured Allowed Domains for postMessage
The embedded dashboard page now validates the origin of incoming `postMessage` events against the dashboard's configured **Allowed Domains**. The server-rendered embedded page exposes the configured domains in its bootstrap payload, and the frontend rejects message events whose origin is not in that list.
Enforcement only applies when the Allowed Domains list is non-empty. If the list is empty (the default), any origin is accepted, so there is no behavior change for embeds that did not configure Allowed Domains.
### Default guest/async JWT secrets are rejected at startup
Superset already refuses to start in production (non-debug, non-testing) when `SECRET_KEY` is left at its built-in default, and when `GUEST_TOKEN_JWT_SECRET` is left at its default while `EMBEDDED_SUPERSET` is enabled. This behavior is extended to `GLOBAL_ASYNC_QUERIES_JWT_SECRET`: if the `GLOBAL_ASYNC_QUERIES` feature flag is enabled and the secret is still the publicly known default (`test-secret-change-me`), Superset logs a clear error and refuses to start.
As with the existing `SECRET_KEY` check, this only fails in production. In debug mode, testing mode, or under the test runner, a warning is logged instead of exiting, so local development is unaffected.
To resolve the error, set a strong random value in `superset_config.py`:
The check is only active when the relevant feature is enabled, so deployments that do not use global async queries (or embedding) are not affected.
### Guest token revocation (opt-in)
Embedded guest tokens can be coarsely revoked at runtime via a new opt-in mechanism. A new config flag `GUEST_TOKEN_REVOCATION_ENABLED` (default `False`) gates the feature. When enabled, every minted guest token carries a revocation version, and tokens whose version is below the current expected version (stored in the metadata database) are rejected at validation time.
Bump the expected version with the new CLI command to invalidate all outstanding guest tokens:
```bash
superset revoke-guest-tokens
```
This change is backward compatible. The feature is off by default, and even when enabled nothing is revoked until an admin explicitly bumps the version: the expected version starts at `0`, and tokens minted before this change (which carry no version claim) are treated as version `0`. No database migration is required.
### Sessions are terminated when an account is disabled
Disabling a user account (setting `active` to `False`, via the admin UI, REST API, or CLI) now terminates that user's outstanding sessions on their next request, instead of relying on a passive check. This works for both client-side cookie sessions and server-side session stores via a per-user invalidation epoch (`user_attribute.sessions_invalidated_at`, added by a migration). The mechanism is inert for users that were never disabled (NULL epoch), so there is no behavior change for active users. Re-enabling an account and logging in again starts a fresh, valid session. The migration backfills the epoch for accounts that are already disabled at upgrade time, so re-enabling such an account does not revive a session that predates this feature.
### Opt-in SSH tunnel server host key verification
SSH tunnels can now optionally pin the expected SSH server host key as a defense-in-depth measure against man-in-the-middle attacks. paramiko's transport performs no known-hosts checking by default, so previously the SSH server's identity was not verified. This feature is opt-in and off by default; existing tunnels are unaffected.
- A new nullable `server_host_key` column on the `ssh_tunnels` table stores the expected host key in authorized-key form (e.g. `ssh-ed25519 AAAA...`). It is a public key and is stored in plaintext. It can be set via the SSH tunnel POST/PUT payloads (`ssh_tunnel.server_host_key`).
- When a tunnel has `server_host_key` set, Superset connects to the SSH server, reads the host key it presents, and rejects the tunnel if it does not match.
- A new config flag `SSH_TUNNEL_STRICT_HOST_KEY_CHECKING` (default `False`) controls fail-closed behavior. When `True`, every tunnel must declare a `server_host_key`; a tunnel without one is rejected.
Runbook to adopt:
1. Capture the SSH server's host key, e.g. `ssh-keyscan -t ed25519 ssh.example.com` (verify it out-of-band).
2. Set that value on the tunnel's `server_host_key` (via the database/SSH tunnel API or UI payload).
3. Optionally set `SSH_TUNNEL_STRICT_HOST_KEY_CHECKING = True` in `superset_config.py` to require host-key verification on all tunnels.
### SMTP server certificate validation enabled by default
`SMTP_SSL_SERVER_AUTH` now defaults to `True` (previously `False`). With this default, STARTTLS/SSL connections to the configured SMTP server validate the server's TLS certificate against the system trusted CA store. This makes outbound email (alerts and reports) verify the mail server's identity out of the box.
If your SMTP server presents a self-signed certificate, or a certificate that is not trusted by the system CA store, email delivery may now fail with a certificate verification error. To restore the previous behavior of skipping certificate validation, set the following in `superset_config.py`:
```python
SMTP_SSL_SERVER_AUTH=False
```
The recommended fix is to add the SMTP server's certificate (or its issuing CA) to the system trust store rather than disabling validation.
### Dataset import validates catalog against the target connection
Importing a dataset now validates the `catalog` field against the target database connection. When the connection has multi-catalog disabled (`allow_multi_catalog` off) and the dataset's catalog is not the connection's default catalog, the import fails instead of silently persisting the non-default catalog. This matches the validation already enforced on the dataset update path and prevents imported datasets from querying an unintended database.
If you relied on importing datasets with a non-default catalog, enable "Allow changing catalogs" on the target connection, or set the dataset's catalog to the connection's default before importing.
### Extension supply-chain controls (denylist + version policy)
Two opt-in static gates control which extensions are allowed to load:
-`EXTENSION_DENYLIST` refuses extensions matching an id (every version) or `id@version` (a single version), e.g. `["compromised-extension", "other-ext@1.2.3"]`.
-`EXTENSION_VERSION_POLICY` enforces a minimum version per extension id, e.g. `{"acme.widget": "1.2.0"}` (PEP 440 comparison); a release below the minimum is refused.
Both default to empty (no behavior change). They apply to both the `LOCAL_EXTENSIONS` and `EXTENSIONS_PATH` load paths.
### Dynamic Group By respects the sort toggle for display values
The Dynamic Group By chart customization now orders its display values according to the "Sort display control values" toggle: ascending (A–Z), descending (Z–A), or the dataset's source order when the toggle is unset. Previously the dropdown always sorted alphabetically. Existing dashboards where the toggle was never set will show options in source order instead of A–Z; open the customization and enable the toggle to restore alphabetical ordering.
### Selectable encryption engine for app-encrypted fields (AES-GCM)
App-encrypted fields (database passwords, SSH tunnel credentials, OAuth tokens, etc.) can now use authenticated **AES-GCM** encryption instead of the historical unauthenticated **AES-CBC**. A new config selects the engine for the default adapter:
```python
# "aes" (AES-CBC, historical default) | "aes-gcm" (authenticated, recommended for new installs)
SQLALCHEMY_ENCRYPTED_FIELD_ENGINE="aes"
```
**No action required / no behavior change:** the default remains `"aes"`, so existing installs are unaffected.
**Opting in on an existing install:** flipping the engine on a populated database without re-encrypting first will make stored secrets undecryptable, because the two ciphertext formats are not compatible. A migrator is provided. Recommended runbook:
1. Take a metadata-DB backup.
2. Re-encrypt existing secrets into the new engine (the `SECRET_KEY` is unchanged):
```bash
superset re-encrypt-secrets --engine aes-gcm
```
3. Set `SQLALCHEMY_ENCRYPTED_FIELD_ENGINE = "aes-gcm"` in your config.
4. Restart Superset.
5. Re-run the migrator once more after the restart:
```bash
superset re-encrypt-secrets --engine aes-gcm
```
A live instance keeps writing *new* secrets as AES-CBC during the window between step 2 and the restart in step 4; this second pass sweeps those up (it is idempotent, so already-migrated values are skipped).
Schedule the cutover in a quiet window. Runtime reads use only the single configured engine, so in a multi-worker deployment there is an unavoidable brief decrypt-outage between the migration commit and the last worker restarting with the new config — each migrator run is transactional, but the fleet-wide cutover is not zero-downtime.
The migration is transactional (all-or-nothing) and idempotent — it can be safely re-run or resumed. Note that AES-GCM, unlike AES-CBC, does not support querying directly over encrypted columns; audit any code that filters on an encrypted column before switching. See the SIP at `docs/sip/authenticated-encryption-at-rest.md` for details.
### Soft delete and restore for dashboards
**Everything in this section applies only when the `SOFT_DELETE` feature flag is enabled. The flag defaults to `False`** (`@lifecycle: development`), so on a default deployment `DELETE /api/v1/dashboard/<id>` continues to **hard-delete permanently** — nothing is recoverable. Enable `SOFT_DELETE` to get the behavior described below.
**Flag-toggle caveat:** the soft-delete visibility filter is evaluated per query while the flag is on. If dashboards are soft-deleted during a flag-on window and the flag is later turned **off**, those rows reappear as live dashboards in all lists and lookups (including slug lookups — if a soft-deleted dashboard's slug was reused while the flag was on, both rows become visible with the same slug). The `POST /<uuid>/restore` endpoint and the `dashboard_deleted_state` list filter remain functional regardless of the flag, deliberately, so rows soft-deleted during a flag-on window stay discoverable and restorable after a rollback of the flag.
With the flag enabled: `DELETE /api/v1/dashboard/<id>` no longer hard-deletes the dashboard (the bulk-delete endpoint behaves the same way). The row is marked with a `deleted_at` timestamp and hidden from the dashboard API's list, detail, and lookup endpoints, which return 404 for soft-deleted dashboards. The embedded-dashboard iframe URL (`/embedded/<uuid>`) keeps rendering because it reads only `embedded.allowed_domains` and `embedded.dashboard_id` (the FK column) without dereferencing the parent dashboard; the frontend's subsequent dashboard-API fetch is what sees the 404 and surfaces "dashboard not found" to the user.
**New endpoint** — `POST /api/v1/dashboard/<uuid>/restore` clears `deleted_at` and returns the dashboard to active state. Requires `can_write on Dashboard` and ownership of the row (or admin). Soft-deleted dashboards can also be surfaced in the list endpoint via the new `dashboard_deleted_state` rison filter: `include` returns both live and soft-deleted rows, `only` returns just the soft-deleted ones. Any other value is ignored. For non-admin users, soft-deleted rows are limited to dashboards they own — the same audience that can restore them.
**Permissions migration:** existing role grants of `can_write on Dashboard` cover the new restore endpoint automatically; no role migration is required.
**Schema migration:** the migration adds a nullable `deleted_at` column and an index on it (`ix_dashboards_deleted_at`) to the `dashboards` table, and **replaces the full unique constraint on `slug`** with a partial unique index (`ix_dashboards_active_slug`) enforcing slug uniqueness only among active (non-soft-deleted) rows. The column add is instant. On Postgres the constraint swap briefly blocks reads and writes during `ALTER TABLE ... DROP CONSTRAINT` (acquires `ACCESS EXCLUSIVE`), then blocks writes only during `CREATE UNIQUE INDEX` (acquires `ShareLock`); reads pass through during the index build. Both windows are sub-second on a typical `dashboards` table. MySQL InnoDB builds the functional index online (no blocking).
**Rollback note:** the downgrade restores the original full unique constraint on `slug`. If the partial-index window allowed slug reuse (a soft-deleted row and an active row holding the same slug), `ALTER TABLE ... ADD CONSTRAINT idx_unique_slug UNIQUE (slug)` will abort with a unique-constraint violation. Before downgrading, hard-delete the soft-deleted duplicates (or rename one side) so each slug appears at most once across all rows. Rolling back the application code while leaving the new migration in place is also possible but exposes soft-deleted rows to the older code path; pair the rollback with a data decision (restore, hard-delete, or migrate-down).
The partial-index replacement is dialect-dependent: PostgreSQL uses a native `WHERE deleted_at IS NULL` partial index; MySQL 8.0.13+ uses a functional index over `(CASE WHEN deleted_at IS NULL THEN slug END)` (8.0.13 is the first release with functional key parts). **MySQL <8.0.13, MariaDB, and SQLite keep the original full unique constraint** (functional indexes / column-level UNIQUE recreation aren't supported cleanly — MariaDB is excluded even at 10.x because its `CASE`-expression index semantics differ), so on those backends a soft-deleted dashboard continues to reserve its slug for the lifetime of the row.
**Slug semantics:** on PostgreSQL and MySQL 8.0.13+, the slug of a soft-deleted dashboard is **free for reuse**. A new active dashboard can claim it immediately. Restoring a soft-deleted dashboard whose slug has since been claimed returns **422 with a clean error** (`DashboardSlugConflictError`) — rename one of the dashboards and retry; the restore is not silently rejected by a database-level constraint violation.
**Importer behavior:** importing a dashboard YAML whose UUID matches an existing **soft-deleted** dashboard is treated as an implicit restore-with-update — **and this happens even when `overwrite` is not set**. This is a deliberate asymmetry with active rows: an active dashboard imported without `overwrite=true` is returned unchanged (the import never mutates it), but a soft-deleted UUID match is restored *and* has the upload's contents applied regardless of the `overwrite` argument, on the reasoning that re-importing a deleted dashboard's exact UUID is an explicit request to bring it back. The restore preserves the original PK and all pre-deletion relationship rows (`dashboard_slices` junctions, role grants, owners, tags) — including role grants that were implicitly revoked by the deletion. Callers whose imports must never mutate existing state should treat bundles that may contain previously deleted UUIDs accordingly. The operation is permission-gated: it requires `can_write` and ownership of the deleted row (or admin) — non-owners get `ImportFailedError`, and callers without `can_write` get `ImportFailedError` instead of silently receiving the soft-deleted row.
### Granular Export Controls
A new feature flag `GRANULAR_EXPORT_CONTROLS` introduces three fine-grained permissions that replace the legacy `can_csv` permission:
@@ -53,6 +340,9 @@ Added a new combined datasource list endpoint at `GET /api/v1/datasource/` to se
- The endpoint is available to users with at least one of `can_read` on `Dataset` or `SemanticView`.
- Semantic views are included only when the `SEMANTIC_LAYERS` feature flag is enabled.
- The endpoint enforces strict `order_column` validation and returns `400` for invalid sort columns.
## 6.1.0
### ClickHouse minimum driver version bump
The minimum required version of `clickhouse-connect` has been raised to `>=0.13.0`. If you are using the ClickHouse connector, please upgrade your `clickhouse-connect` package. The `_mutate_label` workaround that appended hash suffixes to column aliases has also been removed, as it is no longer needed with modern versions of the driver.
@@ -310,6 +600,29 @@ See `superset/mcp_service/PRODUCTION.md` for deployment guides.
}
```
### Composite primary keys on many-to-many association tables
Eight M:N association tables move from a synthetic `id INTEGER PRIMARY KEY` to a composite `PRIMARY KEY (fk1, fk2)` on their two foreign-key columns. The surrogate `id` is dropped, and the redundant `UNIQUE (fk1, fk2)` on the two tables that carried one is removed (now subsumed by the PK).
- The migration **deletes** two classes of pre-existing rows the composite PK cannot accommodate: duplicate `(fk1, fk2)` pairs (it keeps the lowest `id` and removes the rest) and rows with `NULL` in either FK column. Both are meaningless for `secondary=` association tables, but export the affected rows first if you need an audit record.
- External tooling (BI tools, backup scripts) that references the surrogate `id` on these tables will break; no application code references it.
- Downgrade restores the `id` column (and the original `UNIQUE` on the two tables that had it) but leaves the FK columns `NOT NULL` (intentional — a `NULL` FK in a junction row is meaningless).
For large `dashboard_slices` / `report_schedule_user` tables, see the operator runbook in [#39859](https://github.com/apache/superset/pull/39859) — pre-flight inventory queries, per-dialect lock-window sizing, and the duplicate / NULL-FK roll-up — to plan the maintenance window.
## 6.0.0
- [33055](https://github.com/apache/superset/pull/33055): Upgrades Flask-AppBuilder to 5.0.0. The AUTH_OID authentication type has been deprecated and is no longer available as an option in Flask-AppBuilder. OpenID (OID) is considered a deprecated authentication protocol - if you are using AUTH_OID, you will need to migrate to an alternative authentication method such as OAuth, LDAP, or database authentication before upgrading.
- [34871](https://github.com/apache/superset/pull/34871): Fixed Jest test hanging issue from Ant Design v5 upgrade. MessageChannel is now mocked in test environment to prevent rc-overflow from causing Jest to hang. Test environment only - no production impact.
max_attempts=300 # ~10 minutes at 2s intervals; first build can be slow
echo "Waiting for webpack dev server at $$url..."
attempt=0
until curl -sf --max-time 5 -H "Host: localhost" -o /dev/null "$$url"; do
attempt=$$((attempt + 1))
if [ "$$attempt" -ge "$$max_attempts" ]; then
echo "ERROR: webpack dev server did not serve $$url after $$max_attempts attempts." >&2
echo "Is the dev server running? With BUILD_SUPERSET_FRONTEND_IN_DOCKER=false you must start it on the host (e.g. 'npm run dev' in superset-frontend)." >&2
exit 1
fi
if [ $$((attempt % 15)) -eq 0 ]; then
echo "Still waiting for webpack dev server... ($$attempt/$$max_attempts)"
fi
sleep 2
done
echo "Webpack dev server is ready; starting nginx."
On a Slack Enterprise Grid org, an org-scoped token spans multiple workspaces, so
workspace-scoped methods such as `conversations.list` require a `team_id` to
indicate which workspace to target. Set `SLACK_TEAM_ID` to your workspace (team)
ID so Superset can list channels and deliver reports:
```python
# The workspace (team) ID to target, e.g. "T01234567"
SLACK_TEAM_ID = "T01234567"
```
This defaults to `None` and only needs to be set when using an org-scoped token;
it is accepted but ignored for standard workspace-level tokens.
### Webhook integration
Superset can send alert and report notifications to any HTTP endpoint — useful for chat platforms, incident management tools, or custom automation.
@@ -160,7 +175,7 @@ When enabled, Superset rejects webhook configurations that use `http://` URLs.
#### Retry Behavior
Superset automatically retries webhook deliveries on `429 Too Many Requests` and `5xx` server errors using exponential backoff.
Superset automatically retries webhook deliveries on `429 Too Many Requests` and `5xx` server errors using exponential backoff. Retries are bounded to roughly 120 seconds of cumulative wall-clock time (worst case ~210 seconds, because the bound is checked against the time elapsed before each attempt, so the final request can begin just under the limit and still run its full request timeout), after which the delivery is abandoned.
# Extend the default CeleryConfig to add cache warmup schedule
class CustomCeleryConfig(CeleryConfig):
beat_schedule = {
**CeleryConfig.beat_schedule,
'cache-warmup-hourly': {
'task': 'cache-warmup',
'schedule': crontab(minute=0, hour='*'), # hourly
'kwargs': {
'strategy_name': 'top_n_dashboards',
'top_n': 5,
'since': '7 days ago',
},
},
}
CELERY_CONFIG = CustomCeleryConfig
```
This will cache the top 5 most popular dashboards every hour. For other
strategies, check the `superset/tasks/cache.py` file.
## Caching Thumbnails
This is an optional feature that can be turned on by activating its [feature flag](/admin-docs/configuration/configuring-superset#feature-flags) on config:
The page served at `/` is rendered by an index view. By default Superset registers
`SupersetIndexView`, which redirects to `/superset/welcome/` and also adds the
`/lang/<locale>` locale handler. You can replace it with your own view, for example
to send users straight to a specific dashboard or to a chart list.
Set `FAB_INDEX_VIEW` to the **importable dotted path** of your view class. Flask-AppBuilder
resolves this during app initialization and uses it in place of the default:
```python
# my_overrides.py — must be importable on the PYTHONPATH
from flask import redirect
from superset.initialization import SupersetIndexView
from superset.superset_typing import FlaskResponse
from flask_appbuilder import expose
class MyIndexView(SupersetIndexView):
@expose("/")
def index(self) -> FlaskResponse:
return redirect("/chart/list/")
```
```python
# superset_config.py
FAB_INDEX_VIEW = "my_overrides.MyIndexView"
```
A few things that commonly trip people up:
- **Subclass `SupersetIndexView`, not Flask-AppBuilder's bare `IndexView`.** Subclassing
keeps Superset's `/lang/<locale>` locale handling; replacing it with a bare `IndexView`
silently drops that behavior.
- **The class must be importable as a real module.** `FAB_INDEX_VIEW` is resolved by
importing the dotted path, which is independent of how `superset_config.py` itself is
loaded. Superset only copies **uppercase** names out of `superset_config.py` into its
runtime config, so a `FAB_INDEX_VIEW = "superset_config.MyIndexView"` reference only works
if `superset_config` is itself importable by that name on the `PYTHONPATH`. If you load
config via `SUPERSET_CONFIG_PATH` (an arbitrary file path), put the view in a separate
importable module instead and reference that module.
- **Don't set `appbuilder.indexview` from `FLASK_APP_MUTATOR`.** The mutator runs after
routes are already registered, so the assignment has no effect on the `/` route. Use
`FAB_INDEX_VIEW` instead.
## Feature Flags
To support a diverse set of users, Superset has some features that are not enabled by default. For
@@ -504,6 +549,24 @@ CELERY_BEAT_SCHEDULE = {
Adjust `retention_period_days` to control how long query rows are kept. Companion opt-in tasks (`prune_logs`, `prune_tasks`) exist for pruning the logs and tasks tables; see the commented-out examples in `superset/config.py`. Without enabling these tasks, the metadata database will grow unbounded over time.
## Dashboard Layout Size Limit
Each dashboard stores its layout (the position, size, and nesting of every chart, row, and tab) as a JSON blob in the metadata database. Superset caps the length of this serialized blob with `SUPERSET_DASHBOARD_POSITION_DATA_LIMIT`, which defaults to `65535`:
```python
SUPERSET_DASHBOARD_POSITION_DATA_LIMIT = 65535
```
This is a Python-level cap (65535 is 2¹⁶ − 1), independent of the database column capacity — the `position_json` column is a `MEDIUMTEXT`, which holds far more. When the serialized layout reaches this limit, the editor blocks the save and reports the current length, the limit, and this setting's name. A warning is shown once the layout passes 90% of the limit.
Large dashboards — for example, many charts spread across nested tabs — can exceed the default. Because the underlying column comfortably stores larger values, you can safely raise the limit:
```python
SUPERSET_DASHBOARD_POSITION_DATA_LIMIT = 131072 # double the default
```
Alternatively, split a very large dashboard into several smaller ones. Note that this check is enforced when saving layout edits in the UI; a dashboard imported from a ZIP with an oversized layout will load and render, but cannot be edited and re-saved until the limit is raised.
:::resources
- [Blog: Feature Flags in Apache Superset](https://preset.io/blog/feature-flags-in-apache-superset-and-preset/)
Superset uses [Scarf Gateway](https://about.scarf.sh/scarf-gateway) to collect telemetry data. Knowing the installation counts for different Superset versions informs the project's decisions about patching and long-term support. Scarf purges personally identifiable information (PII) and provides only aggregated statistics.
To opt-out of this data collection in your Helm-based installation, edit the `repository:` line in your `helm/superset/values.yaml` file, replacing `apachesuperset.docker.scarf.sh/apache/superset` with `apache/superset` to pull the image directly from Docker Hub.
There are two independent telemetry channels:
- **Image pulls** (Scarf Gateway): to opt out, edit the `repository:` line in your `helm/superset/values.yaml` file, replacing `apachesuperset.docker.scarf.sh/apache/superset` with `apache/superset` to pull the image directly from Docker Hub.
- **The analytics pixel** rendered in the UI: to opt out, set the `SCARF_ANALYTICS` environment variable to `false` on the Superset containers via `extraEnv` in your `values.yaml`:
```yaml
extraEnv:
SCARF_ANALYTICS: "false"
```
This is read at runtime, so it takes effect on the pre-built images without rebuilding the frontend.
Ubuntu **24.04** uses python 3.12 per default, which currently is not supported by Superset. You need to add a second python installation of 3.11 and install the required additional dependencies.
In more recent versions of CentOS and Fedora, you may need to install a slightly different set of packages using `dnf`:
@@ -157,8 +150,15 @@ superset load_examples
superset init
# To start a development web server on port 8088, use -p to bind to another port
superset run -p 8088 --with-threads --reload --debugger
superset run -p 8088 --with-threads --reload
# For debugging with interactive console (⚠️ localhost only)
# superset run -p 8088 --with-threads --reload --debugger
```
:::warning Security Note
The `--debugger` flag enables Werkzeug's interactive console at `/console`. Only use this for local development and never bind to `0.0.0.0` or expose the server to networks when debugging is enabled.
:::
If everything worked, you should be able to navigate to `hostname:port` in your browser (e.g.
locally by default at `localhost:8088`) and login using the username and password you created.
@@ -161,6 +161,7 @@ Here's the documentation section how how to set up Talisman: https://superset.ap
- [ ] Regularly update to the latest major or minor versions of Superset. Those versions receive up-to-date security patches.
- [ ] Rotate the `SUPERSET_SECRET_KEY` periodically (e.g., quarterly) and after any potential security incident.
- [ ] Rotate the other security-critical secrets (guest-token and async-query JWT secrets, SMTP and database credentials) on the cadence in Appendix C, and after any potential security incident.
- [ ] Conduct quarterly access reviews for all users.
- [ ] Assuming logging and monitoring is in place, review security monitoring alerts weekly.
@@ -173,6 +174,24 @@ Rotating the `SUPERSET_SECRET_KEY` is a critical security procedure. It is manda
The procedure for safely rotating the SECRET_KEY must be followed precisely to avoid locking yourself out of your instance. The official Apache Superset documentation maintains the correct, up-to-date procedure. Please follow the official guide here:
### **Appendix C: Secrets Register and Rotation Schedule**
`SUPERSET_SECRET_KEY` is not the only security-critical secret in a Superset deployment. Maintain an inventory of all such secrets, store each in a secrets manager (not in `superset_config.py` or version control), assign an owner, and rotate them on a defined cadence as well as after any suspected compromise.
| Database connection passwords | Access to analytical databases and the metadata DB | Direct database access | Per organizational policy + post-incident |
Notes:
- Rotating `GUEST_TOKEN_JWT_SECRET` or `GLOBAL_ASYNC_QUERIES_JWT_SECRET` invalidates outstanding tokens of that type; schedule rotations accordingly.
- After a suspected compromise, rotate **all** of the above, not only `SUPERSET_SECRET_KEY`.
- Keep the register under change control so new secrets introduced by future features are added to the rotation schedule.
:::resources
- [Blog: Running Apache Superset on the Open Internet](https://preset.io/blog/running-apache-superset-on-the-open-internet-a-report-from-the-fireline/)
- [Blog: How Security Vulnerabilities are Reported & Handled in Apache Superset](https://preset.io/blog/how-security-vulnerabilities-are-reported-and-handled-in-apache-superset/)
# To start a development web server on port 8088, use -p to bind to another port
superset run -p 8088 --with-threads --reload --debugger
superset run -p 8088 --with-threads --reload
# For debugging with interactive console (⚠️ localhost only)
# superset run -p 8088 --with-threads --reload --debugger
```
:::warning Security Note
The `--debugger` flag enables Werkzeug's interactive console at `/console`. Only use this for local development and never bind to `0.0.0.0` or expose the server to networks when debugging is enabled.
:::
If everything worked, you should be able to navigate to `hostname:port` in your browser (e.g.
locally by default at `localhost:8088`) and login using the username and password you created.
@@ -34,15 +34,14 @@ Frontend contribution types allow extensions to extend Superset's user interface
Extensions can add new views or panels to the host application, such as custom SQL Lab panels, dashboards, or other UI components. Contribution areas are uniquely identified (e.g., `sqllab.panels` for SQL Lab panels), enabling seamless integration into specific parts of the application.
```tsx
importReactfrom'react';
```typescript
import{views}from'@apache-superset/core';
importMyPanelfrom'./MyPanel';
views.registerView(
{id:'my-extension.main',name:'My Panel Name'},
'sqllab.panels',
()=><MyPanel/>,
MyPanel,
);
```
@@ -112,6 +111,24 @@ editors.registerEditor(
See [Editors Extension Point](./extension-points/editors.md) for implementation details.
### Chat
Extensions can add a chat interface to Superset by registering a trigger component and a panel component. The host owns the layout, open/close state, and display mode — the extension only provides the UI. The panel can be displayed as a floating overlay or docked as a resizable sidebar beside the page content, and the user's preference is persisted across reloads.
```tsx
import{chat}from'@apache-superset/core';
importChatTriggerfrom'./ChatTrigger';
importChatPanelfrom'./ChatPanel';
chat.registerChat(
{id:'my-org.my-chat',name:'My Chat'},
ChatTrigger,
ChatPanel,
);
```
See [Chat](./extension-points/chat.md) for implementation details.
## Backend
Backend contribution types allow extensions to extend Superset's server-side capabilities. Backend contributions are registered at startup via classes and functions imported from the auto-discovered `entrypoint.py` file.
Licensed to the Apache Software Foundation (ASF) under one
or more contributor license agreements. See the NOTICE file
distributed with this work for additional information
regarding copyright ownership. The ASF licenses this file
to you under the Apache License, Version 2.0 (the
"License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
KIND, either express or implied. See the License for the
specific language governing permissions and limitations
under the License.
-->
# Chat Contributions
Extensions can add a chat interface to Superset by registering a trigger and a panel. The host owns the layout, open/close state, and display mode — the extension only needs to provide the UI components.
## Overview
A chat registration consists of two React components:
| Component | Role |
|-----------|------|
| **Trigger** | Always-visible entry point (e.g., a floating button). Rendered in the bottom-right corner in floating mode, or as a fixed overlay in panel mode. |
| **Panel** | The chat UI itself (message list, input, etc.). Mounted by the host in the active display mode. |
## Display Modes
The host supports two display modes, switchable by the user or the extension at runtime:
| Mode | Behavior |
|------|----------|
| `floating` | Panel floats above page content, anchored to the bottom-right corner. |
| `panel` | Panel is docked to the right side of the application as a resizable sidebar, sitting beside the page content. |
The user's last selected mode and open/closed state are persisted across page reloads.
## Registering a Chat
Call `chat.registerChat` from your extension's entry point with a descriptor, a trigger factory, and a panel factory:
```tsx
import{chat}from'@apache-superset/core';
importChatTriggerfrom'./ChatTrigger';
importChatPanelfrom'./ChatPanel';
chat.registerChat(
{id:'my-org.my-chat',name:'My Chat'},
ChatTrigger,
ChatPanel,
);
```
Only one chat registration is active at a time. If a second extension calls `registerChat`, it replaces the first and a warning is logged.
## Opening and Closing the Chat
The trigger component is responsible for toggling the panel. Use `chat.isOpen()`, `chat.open()`, and `chat.close()` to control visibility:
Call `chat.setDisplayMode` to switch between `'floating'` and `'panel'` modes. In your panel component, subscribe to `onDidChangeDisplayMode` to react to changes (including those triggered by the user):
All methods are available on the `chat` namespace from `@apache-superset/core`:
| Method / Event | Description |
|----------------|-------------|
| `registerChat(descriptor, trigger, panel)` | Register a chat extension. Returns a `Disposable` to unregister. |
| `open()` | Open the chat panel. No-op if already open or no registration. |
| `close()` | Close the chat panel. |
| `isOpen()` | Returns `true` if the panel is currently open. |
| `getDisplayMode()` | Returns the current display mode (`'floating'` or `'panel'`). |
| `setDisplayMode(mode)` | Switch between `'floating'` and `'panel'` mode. |
| `onDidOpen(listener)` | Subscribe to panel open events. Returns a `Disposable`. |
| `onDidClose(listener)` | Subscribe to panel close events. Returns a `Disposable`. |
| `onDidChangeDisplayMode(listener)` | Subscribe to display mode changes. Returns a `Disposable`. |
| `onDidRegisterChat(listener)` | Subscribe to registration events. |
| `onDidUnregisterChat(listener)` | Subscribe to unregistration events. |
| `onDidResizePanel(listener)` | Subscribe to panel resize events (panel mode only). Not all hosts provide a resizer — do not rely on this firing. Returns a `Disposable`. |
## Next Steps
- **[Contribution Types](../contribution-types.md)** — Explore other contribution types
- **[Development](../development.md)** — Set up your development environment
@@ -321,8 +321,8 @@ This can be used, for example, to convert UTC time to local time.
Superset uses [Scarf](https://about.scarf.sh/) by default to collect basic telemetry data upon installing and/or running Superset. This data helps the maintainers of Superset better understand which versions of Superset are being used, in order to prioritize patch/minor releases and security fixes.
We use the [Scarf Gateway](https://docs.scarf.sh/gateway/) to sit in front of container registries, the [scarf-js](https://about.scarf.sh/package-sdks) package to track `npm` installations, and a Scarf pixel to gather anonymous analytics on Superset page views.
Scarf purges PII and provides aggregated statistics. Superset users can easily opt out of analytics in various ways documented [here](https://docs.scarf.sh/gateway/#do-not-track) and [here](https://docs.scarf.sh/package-analytics/#as-a-user-of-a-package-using-scarf-js-how-can-i-opt-out-of-analytics).
Superset maintainers can also opt out of telemetry data collection by setting the `SCARF_ANALYTICS` environment variable to `false` in the Superset container (or anywhere Superset/webpack are run).
Additional opt-out instructions for Docker users are available on the [Docker Installation](/admin-docs/installation/docker-compose) page.
You can also opt out of the analytics pixel by setting the `SCARF_ANALYTICS` environment variable to `false`. This is read at runtime, so setting it on the Superset container (for example via `extraEnv` in the Helm chart, or `docker/.env` for Docker Compose) disables the pixel on the pre-built images without rebuilding the frontend. Note that this only disables the page-view pixel; the Scarf Gateway (container registry) and `scarf-js` (`npm`) channels are opted out separately, as described above.
Additional opt-out instructions are available on the [Docker Compose](/admin-docs/installation/docker-compose) and [Kubernetes](/admin-docs/installation/kubernetes) installation pages.
## Does Superset have an archive panel or trash bin from which a user can recover deleted assets?
<a href="https://x.com/apachesuperset" target="_blank" rel="noopener noreferrer" title="Follow us on X" aria-label="X">
<img src="/img/community/x-symbol.svg" alt="X" />
</a>
<a href="https://www.linkedin.com/company/apache-superset" target="_blank" rel="noopener noreferrer" title="Follow us on LinkedIn" aria-label="LinkedIn">
<a href="https://bsky.app/profile/apachesuperset.bsky.social" target="_blank" rel="noopener noreferrer" title="Follow us on Bluesky" aria-label="Bluesky">
# Additional domains required for docs site functionality:
# - widget.kapa.ai: AI chatbot widget (uses Google reCAPTCHA). Approval here: https://privacy.apache.org/faq/committers.html
# - *.googleapis.com, *.google.com, *.gstatic.com: Google Calendar embed, kapa.ai reCAPTCHA - all of these loaded with user consent, following policy laid out in https://privacy.apache.org/faq/committers.html
# - github.com, *.github.com, *.githubusercontent.com: GitHub user-attachment images in docs (apex github.com serves user-attachments/* assets). Discussed/resolved in this thread: https://issues.apache.org/jira/browse/INFRA-25701?filter=-2 (DPA in place with GitHub)
"description":"Enable semantic layers and show semantic views alongside datasets"
},
{
"name":"SOFT_DELETE",
"default":false,
"lifecycle":"development",
"description":"Temporary rollout / kill-switch gate for soft delete (default off = legacy hard delete). An emergency stop, not a clean rollback: flipping ON->OFF resurrects already-soft-deleted rows. Removed (along with its two gate points \u2014 BaseDAO.delete routing and the do_orm_execute visibility listener) once soft delete is stable."
@@ -1808,6 +1808,10 @@ If you enable DML in the meta database users will be able to run DML queries on
Second, you might want to change the value of `SUPERSET_META_DB_LIMIT`. The default value is 1000, and defines how many are read from each database before any aggregations and joins are executed. You can also set this value `None` if you only have small tables.
:::warning
`SUPERSET_META_DB_LIMIT` is applied to **each** underlying table *before* the in-memory join runs, not to the final result. If any table involved in a join has more rows than the limit, the meta database will read only the first `SUPERSET_META_DB_LIMIT` rows of that table, which means matching rows can be silently dropped and the join can return **incomplete or even empty** results with no error. If you join tables larger than the limit, raise `SUPERSET_META_DB_LIMIT` to comfortably exceed your largest joined table, or set it to `None` when working only with small tables, to get correct results.
:::
Additionally, you might want to restrict the databases to with the meta database has access to. This can be done in the database configuration, under "Advanced" -> "Other" -> "ENGINE PARAMETERS" and adding:
| supersetCeleryFlower.deploymentAdditionalPodSpec | object | `{}` | Custom pod spec to be added to supersetCeleryFlower deployment |
| supersetCeleryFlower.deploymentAnnotations | object | `{}` | Annotations to be added to supersetCeleryFlower deployment |
| supersetCeleryFlower.enabled | bool | `false` | Enables a Celery flower deployment (management UI to monitor celery jobs) WARNING: on superset 1.x, this requires a Superset image that has `flower<1.0.0` installed (which is NOT the case of the default images) flower>=1.0.0 requires Celery 5+ which Superset 1.5 does not support |
| supersetCeleryFlower.extraContainers | list | `[]` | Launch additional containers into supersetCeleryFlower pods |
@@ -207,18 +207,21 @@ On helm this can be set on `extraSecretEnv.SUPERSET_SECRET_KEY` or `configOverri
| supersetNode.connections.redis_host | string | `"{{ .Release.Name }}-redis-headless"` | Change in case of bringing your own redis and then also set redis.enabled:false |
| supersetNode.deploymentAdditionalPodSpec | object | `{}` | Custom pod spec to be added to supersetNode deployment |
| supersetNode.deploymentAnnotations | object | `{}` | Annotations to be added to supersetNode deployment |
| supersetNode.deploymentLabels | object | `{}` | Labels to be added to supersetNode deployment |
| supersetNode.env | object | `{}` | |
| supersetNode.extraContainers | list | `[]` | Launch additional containers into supersetNode pod |
| supersetNode.forceReload | bool | `false` | If true, forces deployment to reload on each upgrade |
| supersetNode.initContainers | list | a container waiting for postgres | Init containers |
| supersetNode.lifecycle | object | `{}` | Container lifecycle hooks, e.g. a preStop sleep so the Service/Ingress stops routing to the pod before gunicorn receives SIGTERM |
| supersetNode.livenessProbe.failureThreshold | int | `3` | |
@@ -251,11 +254,13 @@ On helm this can be set on `extraSecretEnv.SUPERSET_SECRET_KEY` or `configOverri
| supersetNode.startupProbe.successThreshold | int | `1` | |
| supersetNode.startupProbe.timeoutSeconds | int | `1` | |
| supersetNode.strategy | object | `{}` | |
| supersetNode.terminationGracePeriodSeconds | string | `nil` | Pod termination grace period (seconds). Set greater than GUNICORN_TIMEOUT so in-flight requests can drain before SIGKILL |
| supersetNode.topologySpreadConstraints | list | `[]` | TopologySpreadConstrains to be added to supersetNode deployments |
| supersetWebsockets.affinity | object | `{}` | Affinity to be added to supersetWebsockets deployment |
| supersetWebsockets.command | list | `[]` | |
| supersetWebsockets.config | object | see `values.yaml` | The config.json to pass to the server, see https://github.com/apache/superset/tree/master/superset-websocket Note that the configuration can also read from environment variables (which will have priority), see https://github.com/apache/superset/blob/master/superset-websocket/src/config.ts for a list of supported variables |
| supersetWebsockets.enabled | bool | `false` | This is only required if you intend to use `GLOBAL_ASYNC_QUERIES` in `ws` mode see https://superset.apache.org/docs/contributing/misc#async-chart-queries |
| supersetWebsockets.extraContainers | list | `[]` | Launch additional containers into supersetWebsockets pods |
@@ -309,11 +314,13 @@ On helm this can be set on `extraSecretEnv.SUPERSET_SECRET_KEY` or `configOverri
| supersetWorker.autoscaling.targetCPUUtilizationPercentage | int | `80` | |
| supersetWorker.command | list | a `celery worker` command | Worker startup command |
| supersetWorker.deploymentAdditionalPodSpec | object | `{}` | Custom pod spec to be added to supersetWorker deployment |
| supersetWorker.deploymentAnnotations | object | `{}` | Annotations to be added to supersetWorker deployment |
| supersetWorker.deploymentLabels | object | `{}` | Labels to be added to supersetWorker deployment |
| supersetWorker.extraContainers | list | `[]` | Launch additional containers into supersetWorker pod |
| supersetWorker.forceReload | bool | `false` | If true, forces deployment to reload on each upgrade |
| supersetWorker.initContainers | list | a container waiting for postgres and redis | Init container |
| supersetWorker.lifecycle | object | `{}` | Container lifecycle hooks for the worker pod |
| supersetWorker.livenessProbe.exec.command | list | a `celery inspect ping` command | Liveness probe command |
| supersetWorker.livenessProbe.failureThreshold | int | `3` | |
| supersetWorker.livenessProbe.initialDelaySeconds | int | `120` | |
@@ -334,6 +341,7 @@ On helm this can be set on `extraSecretEnv.SUPERSET_SECRET_KEY` or `configOverri
| supersetWorker.resources | object | `{}` | Resource settings for the supersetWorker pods - these settings overwrite might existing values from the global resources object defined above. |
| supersetWorker.startupProbe | object | `{}` | No startup/readiness probes by default since we don't really care about its startup time (it doesn't serve traffic) |
| supersetWorker.strategy | object | `{}` | |
| supersetWorker.terminationGracePeriodSeconds | string | `nil` | Pod termination grace period (seconds) for the worker pod so in-flight tasks can drain before SIGKILL |
| supersetWorker.topologySpreadConstraints | list | `[]` | TopologySpreadConstrains to be added to supersetWorker deployments |
| tolerations | list | `[]` | |
| topologySpreadConstraints | list | `[]` | TopologySpreadConstrains to be added to all deployments |
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
