Skip to content

Release notes

What changed in each release of Velonde, newest first. Each release lists only the groups that apply: Features (new things you can do), Changes (existing behaviour that is different now, including anything you have to act on), Bug fixes and Updates (new versions of bundled components: IDEs, R, Python, Claude Code, ShinyProxy, ...).

v1.0.0 (2026-11-15)

The first public release of Velonde.

Features

  • RStudio, VS Code and JupyterLab in the browser, behind your identity provider's login. Each user has one home directory that all three IDEs share, can run several sessions at once and picks a size for each. Claude Code is installed and configured in every IDE, and Posit Assistant in RStudio, with the workspace's skills and rules for deploying and the conventions of R and Python projects. VS Code has Quarto, the version the workspace renders documents with. A workspace can also be set up without any AI tools.
  • R and Python in both main IDEs. VS Code has R with the workspace's R stack: the R extension, an Interactive window that shows plots and data frames inline, and RStudio's shortcuts for the assignment arrow and the pipe. RStudio has Python: uv, the workspace's Python stack and reticulate, with R and Python in one session (py$x, r.x), also in Quarto and R Markdown.
  • VS Code has a Velonde view that lists what you and others deployed, with buttons to open an app, read its logs and roll it back, and a Deploy button in the status bar. RStudio has the same as two addins, "Deploy to the workspace" (a background job) and "Velonde: deployed apps".
  • The coding agents follow rules for secure apps in R and Python: SQL with parameters, no user input in unescaped HTML, no code or shell commands built from input, safe uploads and file paths, and access decided in the server. claude_login_method and claude_login_org in velonde.env restrict Claude Code's sign-in to your organisation's Anthropic accounts.
  • velonde new <name> --template <template> creates a project from the workspace's templates, with its dependencies, velonde.yaml, a CLAUDE.md for the coding agent and a first commit: R projects in RStudio (also velonde::new_app() and File > New Project > New Directory > Velonde project), Python projects in VS Code and JupyterLab.
  • velonde deploy publishes a project from the IDE: Shiny apps in R and Python, Streamlit and Dash apps, R Markdown and Quarto documents in R and Python, Jupyter notebooks (with the outputs saved in them, or run at build with execute: true), plumber, plumber2 and FastAPI APIs (vetiver models too), MCP servers in R and Python, and pipelines: targets, or a single R or Python script, with a store that persists between runs. The workspace detects the kind and the system libraries the packages need, and keeps the last ten builds of every app for velonde rollback. An app can also follow a branch of a git repository and is rebuilt on every push. A .velondeignore file leaves large data and whatever the app does not need out of the upload. velonde show says how an app is set up and whether it is up, velonde logs --follow goes on printing its log, and velonde apps, history and runs print JSON with --json.
  • Every deploy is checked before it goes live. Passwords, keys and tokens in the sources, packages with a known vulnerability that a newer version fixes, and failing tests or files that do not parse stop the deploy; risky code such as SQL pasted from text or a shell command is listed. The output names the file and line and how to fix it. Vulnerabilities are looked up in a copy of the OSV database that the workspace downloads daily, so no package list leaves the workspace. Admins decide per check whether it stops a deploy, only warns or is off (check_secrets, check_vulnerabilities, check_code, check_tests in velonde.env). velonde check (in R velonde::deploy_check()) runs the same checks without deploying.
  • An app's own Dockerfile is accepted only when an admin allows it (deploy_allow_dockerfile), if wanted only for some groups and only on approved base images, by default the workspace's own.
  • Pipelines run on a schedule, with velonde run, or after: other pipelines. A document with schedule: or after: is a scheduled report: it renders in runs with the app's secret variables and pipeline stores, every render becomes a version you can roll back to, params: and variants: set its parameters, and email: mails its readers the link.
  • Working on an app's code together: velonde clone <app> gets the code an app was deployed from, on its branch, with its packages installed, for everyone who may open the app. velonde git login stores your own token for your git server in your home directory, and velonde git create creates a private repository for a project on GitHub or GitLab. Commits carry your name and email address.
  • A package repository for your own R and Python packages. velonde deploy in a package directory builds it and publishes it: for R with its dependencies, R CMD check when velonde.yaml says check: true, and a binary for the workspace's R so that it installs in seconds; for Python with uv build and, with check: true, pytest. Every session and app build of the workspace then installs it like a package from CRAN or PyPI (install.packages(), uv add, pip install), and an app's renv.lock or uv.lock restores it on deploy. A published version is final: the workspace refuses a version that exists and a name that CRAN or PyPI has. The Packages tile on the landing page shows every package, its versions, documentation and owner, and velonde packages token gives a personal token for installing from computers outside the workspace. The repository is optional; it runs RDepot with a PostgreSQL database in the workspace's namespace.
  • Velonde admins add system libraries to the RStudio and JupyterLab images without a new release: velonde ide-packages add --for sf adds what the R package sf needs, or name Ubuntu packages directly. Sessions started after the rebuild of a few minutes have them.
  • Access by groups of your identity provider for apps and APIs, and for single people with access-users. An app has an owner and can name maintainers and maintainer groups who deploy and manage it with them; velonde owner hands it to someone else. When a scheduled run or the build of a pushed commit fails, the owner and maintainers get a mail with the end of the output.
  • An audit trail: the workspace records who deployed, deleted, rolled back or ran an app, who read or changed its secret variables or API key, who handed it over or bound it to a repository, what a deploy changed of who may open it, and every call of an API the workspace refused. Values of secrets are never recorded. Admins read it with velonde audit or in the Grafana dashboard "Velonde audit"; it is kept for a year.
  • Apps, documents, APIs and MCP servers can be published to people outside the company. audience: external in velonde.yaml puts them behind a second login, such as your customers' identity provider, and audience: public opens them to anyone without a login. Each audience has its own host name, which an admin adds to the workspace. A public app shares its containers between visitors and admits 50 at a time unless seats-per-container and max-users say otherwise. max-users limits how many people use any app at the same time.
  • APIs and MCP servers accept calls from sessions and apps in the workspace as the calling user, tokens of your identity provider from outside, or a key per API. MCP servers are registered in every permitted user's Claude Code and Posit Assistant.
  • Your environment goes into the images: R and Python packages and snapshot dates, package mirrors, CA certificates, system packages and project templates.
  • User code (sessions, apps, APIs, builds, pipeline runs) reaches no private addresses but the networks you name. With egress=proxy it reaches the internet only through a proxy and only the hosts on an allow-list (package repositories, GitHub, GitLab, the coding agents' providers and a few more), which you extend in the overlay; the proxy logs every host it allows or refuses.
  • Shared volumes: a team's network share (NFS, SMB, CephFS, EFS, Azure Files, or other storage the cluster can mount) appears in every IDE at /srv/volumes/<name>, for the users and groups the administrator names, read-only or writable, whole or one folder per user. Apps, APIs and pipelines read a share they name under volumes:.
  • Settings for a workspace whose users do not know each other, such as a trial for visitors: default_access=owner keeps an app without access lists to its owner and maintainers, app_max_per_user limits the apps per user, and ide_sizes=small offers only the Small size.
  • An installer for an Ubuntu server or your own Kubernetes cluster, with upgrade and rollback by release. Dashboards and mail alerts for usage, apps, resources and backups, and optional nightly backups to S3-compatible storage.
  • The workspace runs on the license that comes with your subscription. The installer fetches it, and install.sh license installs a renewal. Thirty days before the license runs out, the workspace shows a banner and every velonde command a notice. After its last day, new IDE sessions and deploys are blocked until a renewed license is installed, and thirty days later the sessions still open end. What is deployed keeps running: apps, documents, APIs, MCP servers, pipeline schedules, scheduled reports and the package repository. No data is removed.