Superset 7.0 release candidates are out + latest (extremely spooky) Apache Superset news
It's Spooky Season, Supersetters! 🎃
There's a lot to cover since our last edition, so read on for the highlights. It's all treats 🍬, no tricks 🧻!
Since our last newsletter, the community merged about 1,600 PRs, over 100 new committers landed their first PR, and the repo surpassed 75k stars. Thanks for being a part of growing this Apache Superset™ community.
🔮 We see conferences in your future
On the heels of stints at Airflow summit, dbt Summit, and ASF's own Community over Code, Superset committers are still hitting the road!
Next up: OSA CON 2026
Join us at the Open Source Analytics Conference. It's free and you can register here to join us online or in personin San Francisco.
As always, there will be more updates on Superset's Community Calendar and on Preset's Events page.
🧟 It's aliiiiive! 7.0 release candidates are here!
7.0 is (as always) our most ambitious release yet, and chock full of features and fixes to make your Superset experience better than ever. You can expect a full blog post upon release with all the gory details🔪.
This candidate addresses issues found in the first, with fixes around the public role, embedded guest tokens, improved Redis/Valkey support for websockets, and improved error states.
Status:
We're now on 7.0.0rc2 as of October 16th, which open for community testing while the release vote runs. You can download and test the rc2 binary here.
Make sure the next one doesn't suck 🧛🏻♂️
We encourage you to continue testing even if others downvote the release over any given issue. The more issues we find now, the fewer release candidates we'll have to build and test.
To vote:
Please subscribe to the Superset dev email list by sending an email to [email protected]. Then you can then reply to the thread.
Upgrading from 6.1?
Read UPDATING.md and the first, since 7.0 changes behavior in places like version history retention, SQL execution limits and MCP response sizes. Try it against a staging copy and tell us what breaks. Check out the full changelog for more details.
More on the release roadmap 📍🗺️
- 6.1.0 remains the current stable release until 7.0 passes its vote. You can continue to open new issues against it.
- 5.x and prior are no longer supported. ⚰️ 🪦
- Using Kubernetes? The Helm chart moved from 0.16.0 to 0.22.8 over the summer, if you run Superset on Kubernetes. That said, the helm chart is now deprecated, and you should be migrating to the new Kubernetes Operator.
- Help shape what ships. Want a say in 7.0 and beyond? The dev mailing list is where scoping happens, and the community cal has every Town Hall and working group session on it.
😱 Superset gets screaming fast with Global Task Framework!
The Global Task Framework (GTF) has landed, and gives long-running work a place to live (and die). Zombie queries, meet your doooom 💀
Chart-data queries are cancellable now, with a Task List you can now view and filter
- Distributed locking kills duplicate runs
Task status fans out per browser tab, so closing one doesn't strand the others
Orphaned warehouse queries get reaped by a dedicated job, workers fence themselves off when they lose the metastore
Global Async Queries moved onto the framework too, retiring bespoke async plumbing that's haunted us for years 👻
Also shipped 🛳️
Superset now supports soft deletion of charts, dashboards and datasets behind the SOFT_DELETE flag. Delete something and it's timestamped and hidden rather than gone, and you can put it back. The MCP service picked up restore_chart and restore_dashboard to match, so an agent can undo itself too.
Superset now has a native Candlestick chart for OHLC and range-style time series. Bar labels pick inside or outside based on fit, time axes stop colliding, and a new headless query API lets you query datasets and semantic views by name over REST.
Preview any past version of a chart or dashboard and restore it, and a restore adds a new version instead of overwriting history. Soft delete reached general availability, so deleted charts, dashboards and datasets sit in "Recently Archived" for 30 days by default.
Dashboard exports now produce real multi-sheet workbooks, carrying chart queries and live filter state, with delivery to S3. Alerts and Reports can send as Excel over email now, and charts export to PNG and PDF alongside JPEG. Your finance stakeholders can stop asking.
Since 6.1 shipped, the service has picked up dashboard automation, dataset and report tools, semantic layer discovery, theme management, guest token auth, delete and restore, and chart-type plugins for histogram, box plot and waterfall. August was spent hardening and documenting.
Romanian and Serbian arrived in both Cyrillic and Latin, Japanese got a backfill with native corrections, French picked up numerous improvements, and Arabic, Catalan, Persian, Finnish, Thai and Turkish are now exposed in the UI. There's also a do-not-translate registry, so product nouns stop getting localized into nonsense.
One way to say who can do what 🤝
The new Subject model collapses Users, Roles and Groups into a single idea, so granting a capability and assigning a resource stop being the same operation.
Security fixes galore...
We don't mean to be cryptic but we will say we've cleared the project of numerous dependency CVEs, and completed several rounds of security audits with ASF and other external auditors who shall remain nameless. The AI security scan era is here, and we're patching things as fast as the world is finding them. You will want to upgrade to 7.0, it'll be our most secure release yet. 🧙🏻Speaking of witch...
Getting your hands on these changes:
Already on Preset? There's nothing to upgrade. We manage Superset for you, so new releases reach your workspace on our schedule.
Running open source? 6.1 is still the stable release and the floor for everything above. Grab it below, then try 7.0.0rc2 in a staging environment with SEMANTIC_LAYERS and SOFT_DELETE turned on, and tell us what your team makes of it before the final vote.
As always, feel free to file any issues on github, so we can put any bugs to rest. 🪦
Apache Distribution | PyPi | Docker | Changelog
Stay in the know
Superset News and Updates
Evan Rusackas' latest recap covers updates across 330 PRs from 48 contributors, including PWA support, the Global Task Framework, and a wave of embedded performance fixes.
A first-time user puts Apache Superset head to head with Power BI. Their take: Superset wins on SQL-native workflows and no per-viewer pricing.
Evan Rusackas articulates why open source velocity isn't the same as community health, and the ten roles, like the "Greeter" and the "Matchmaker," that people still do best.
108 contributors merged 727 PRs in August, and the theme was ~recoverability~. Highlights: Version history, soft delete, a mobile dashboard mode, a native Butterfly chart, and Snowflake OAuth.
Preset Founder & CEO Maxime Beauchemin on Agor Cloud, the hosted service for bringing your team and agents together with multiplayer AI.
Keep up with the repo
Star it, tell a friend, rep some swag, and add your company to In The Wild, to build trust for anyone evaluating Superset.
THANK YOU from the bottom of my heart🫀for shaping and growing the Superset community.
Evan Rusackas
ASF Member, Apache Superset PMC &
Preset Developer Advocate
Sent on behalf of Apache Superset by your friends, Committers, and PMC Members at
Copyright © 2026 Preset Inc. All rights reserved.
Preset, Inc, 548 Market St, PMB 51897, San Francisco, California 94104-5401, United States