Spin up ephemeral, API-triggered containers for run-once workloads, from batch jobs to AI agents.
Run-to-completion work, on demand.
Task containers bring a new way to run work on Upsun: ephemeral containers that spin up on an API trigger, run a command with full access to your environment's services, then shut down when the job is done. They handle any run-to-completion workload, from data migrations and report generation to batch processing, and they give background AI agents the isolated, on-demand compute they need.
Billed only while a task runs, with CLI and Console support. [Read the documentation to get started](https://developer.upsun.com/docs/configure-apps/tasks).
You can now deploy Meilisearch on Upsun using its official public Docker image.
[This is the first verified example](https://developer.upsun.com/docs/add-services/docker-images/docker-meilisearch) for self-hosted services from public Docker images. Upsun pulls the image from a public registry and runs it as-is, without requiring you to write a Dockerfile or maintain a build process.
You can run any public Docker image as a self-hosted service, making it easy to experiment with other tools and services beyond Meilisearch.
You remain responsible for the image version, maintenance, and security updates.
You can now view configuration changes directly in the activity log. Open a commit activity and select **Config changes** to see the exact changes to your Upsun configuration files.
Compare added and removed lines without leaving the Console or digging through Git, making it easier to understand what changed between deployments and troubleshoot issues faster.
Task containers are a new infrastructure primitive: ephemeral containers that spin up on an API trigger, run a command with full access to your environment's services, and shut down when done. They're built for any run-to-completion workload, from data migrations and report generation to batch processing.
Running isolated, on-demand code with scoped environment access, they're also a natural fit for background AI agents.
Available on Upsun Flex as a prerelease. Read [the documentation](https://developer.upsun.com/docs/configure-apps/tasks) to learn more and request access.
Upgrade to RabbitMQ 4.2 for more efficient messaging and better operational visibility:
Quorum queues now support single active consumer, improving message ordering and simplifying consumer coordination.
Stream performance and stability improvements help handle high-throughput workloads more reliably.
Enhanced observability with additional metrics and monitoring capabilities makes it easier to understand system behavior in production.
With RabbitMQ 4.2, you can process messages more reliably while gaining better insight into your system’s performance.
Upgrade to Solr 10.0 for improved search performance, scalability, and modern platform support:
Expanded vector search capabilities, including scalar and binary quantized dense vectors, improve performance and reduce memory usage.
Optional GPU acceleration via cuVS-Lucene enables faster vector operations for advanced workloads.
New indexing and query improvements, including enhanced KNN tuning and optimized merge policies, increase throughput and search efficiency.
Modernized platform with Java 21 support, Jetty 12 upgrade, and full migration to OpenTelemetry for observability.
Stability improvements with Apache Curator managing ZooKeeper interactions.
This is a major release with breaking changes and removed legacy components, so review upgrade notes before migrating.
We’ve upgraded our status page to make it even easier for you to stay informed. Not only does the Upsun status page have a new look, but it also has a new home at status.upsun.com. Resubscribe now to continue receiving status notifications for region and system health, as well as upcoming maintenance events.
Visit [status.upsun.com](http://status.upsun.com) to resubscribe, grab the new RSS feed, or see the status of your region or upcoming maintenance you need to be aware of.
Managing worker instances manually often leads to a trade-off between over-provisioning for peak loads or risking performance bottlenecks. To simplify this, we are introducing **horizontal autoscaling for workers** on Upsun Flex.
You can now configure your worker instances to scale automatically based on **CPU and RAM utilization**. This initial release provides a critical foundation for developers who need to handle resource-intensive processing without manual intervention.
This feature is available immediately via the **Console, CLI, and API** for all projects on Upsun Flex.
Upgrade to the latest Enterprise release to leverage advanced search capabilities and significant resource optimizations:
* **ES|QL:** A pipe-based query language to transform, enrich, and simplify data investigation in a single request.
* **Semantic Search:** Use ELSER (Elastic Learned Sparse Encoder) for out-of-the-box relevance without model training.
* **Efficiency at Scale:** Reduce memory usage for vectors by up to 96% with BBQ and cut log storage costs by 65% using `logsdb` mode.
* **Universal Profiling:** Gain whole-system visibility with zero-instrumentation monitoring.
* **Learning to Rank:** Optimize search relevance using machine learning models.
*This version is available as a Premium Service.*
FrankenPHP is now supported on Upsun, providing a high-performance alternative to traditional PHP-FPM setups. By embedding the PHP runtime directly into the server, it reduces orchestration overhead and simplifies your application architecture.
**Key benefits:**
* **Simpler Infrastructure:** Combines the web server and PHP runtime into a single service, reducing the number of moving parts to maintain.
* **Flexible Execution:** Choose **Classic mode** for drop-in, stateless compatibility, or **Worker mode** for maximum throughput by keeping your application in memory.
* **Integrated Observability:** Full support for Blackfire profiling in both execution modes.
The .NET 10 runtime is now available. This LTS release focuses on performance optimizations and new C# 14 language features.
* **C# 14 Features:** Use the new `field` keyword for auto-properties and enhanced extension blocks.
* **Runtime Performance:** Improvements in JIT inlining, method devirtualization, and AVX10.2 support.
* **ASP.NET Core 10:** Built-in Minimal API validation, Server-Sent Events (SSE) support, and OpenAPI 3.1.
* **Security:** Expanded support for post-quantum cryptography (ML-DSA, ML-KEM).
* **Native AOT:** Faster compilation times and reduced binary footprints.
You can now deploy Mercure Hub as a managed service on Upsun. Mercure allows you to push server-to-client updates via Server-Sent Events (SSE), eliminating the complexity of managing custom WebSocket servers.
* Native SSE: Works with any web browser without extra libraries.
* Secure: Built-in support for JWT authorization.
* Integrated: Add the Mercure image to your configuration to start broadcasting updates immediately.
You can now use the latest Nix stable channel, 25.11, in your Composable Image configurations. This update provides access to the newest package versions and improvements from the Nix ecosystem while maintaining the reproducibility and efficiency of your build process.
To upgrade, update the `type` or channel reference in your `.upsun/config.yaml`.
The Ruby 4.0 image is now available. This release introduces major performance and concurrency features:
* **Ruby Box (Experimental):** Provides isolation for definitions, monkey patches, and libraries within a process.
* **ZJIT Compiler:** A new JIT compiler developed as the next generation of YJIT.
* **Ractor Enhancements:** Improved parallel execution with reduced global lock contention. Includes the new `Ractor::Port` class for improved messaging and `Ractor.shareable_proc` for sharing objects between Ractors.
The Python 3.14 image is now available. Key improvements include:
* Concurrent subinterpreters: Achieve true parallelism by running code in separate interpreters without GIL interference.
* UUID6-8: Support for time-ordered, sortable unique identifiers.
* t-strings: Introduced deferred-evaluation template strings.
* Enhanced Terminal UX: Syntax highlighting is now enabled for the REPL, PDB, argparse help, and unittest.