Licensed to be used in conjunction with basebox, only.
// reference
basebox 1.9.0: Connectors That Fit Your Organisation
Control over what the assistant may reach, what it may say, and what it hands back
basebox 1.9.0 is about governance without friction. Administrators get precise control over which systems the assistant can use in which app, which words may never appear in a conversation, and how long uploaded documents live. Users get an assistant that produces real documents, shows how well its answers are backed by your sources, and offers more models to choose from. All of it stays where it belongs: inside your installation.
This release covers everything since 1.8.7.
🔌 Connectors, Exactly Where You Want Them
Connectors link the assistant to the systems your team already uses: wikis, ticket systems, mailboxes, document stores. In 1.9.0 they become far more precise.
Per-app connectors. An app author decides on a new Connectors tab which connectors their app offers, and the chat shows only those. Administrators decide per connector whether it is available in every app or only in apps that opted in. A support app can reach the ticket system while a writing app cannot, without anyone having to explain the difference to users.
Always-on connectors. Some connectors need nothing from a user: no login, no account, nothing to decide. The Document Generator is the prime example. Administrators can now mark such a connector as active for everyone. It disappears from the connector menu and simply works wherever it is available.
Plain-language errors. When a connector call fails, users no longer see a stack of technical output. They see what happened and what to do: the service is temporarily unavailable, the account needs connecting, access has expired. Technical details and a reference ID stay one click away for your support team.
A page for your own connectors. Every user has a Your connectors page listing the connectors available to them, which ones still need a sign-in, and when each was last reached. Signing in happens in one dialog wherever a connector is met, and connectors are switched on and off right from the chat.
Administration, redesigned. Connectors are shown as cards with a single settings panel, the settings are asked as a few plain questions, adding a connector takes fewer steps, and each connector lists what it can do. The interface says connector everywhere instead of MCP, and connector names and descriptions are translated.
New in the catalog. CentralStation CRM, roXtra and SharePoint.
🏢 Settings That Belong to Your Organisation
Where several organisations share one installation, each now keeps its own connectors, support address, welcome message and groundedness cut-offs, and each administrator manages only their own.
🛡 Blocked Words: Company Guidelines, Enforced
Instructions to the assistant are requests. Blocked words are rules.
Administrators maintain a list of words and phrases that must not appear in a chat. If one shows up in a question or an answer, the content is withheld and the user is told that it is not allowed under the company guidelines. This works independently of the model and independently of how a question is phrased, which makes it the right tool for the handful of terms that must never appear in writing.
📄 The Assistant Delivers Documents
The Document Generator, introduced as a preview in 1.8.9, is now complete.
Ask for a report and receive a Word file. Ask for a presentation and receive PowerPoint slides. Ask for a table and receive an Excel workbook. Prefer LibreOffice? The OpenDocument formats are there too. Need something for printing or signing? Request a PDF.
Files appear as a download link right next to their name in the answer, and the link keeps working when you return to the conversation later. Everything is built inside your installation from the content of the request. No document ever leaves your infrastructure to be rendered somewhere else.
📊 How Well Is an Answer Grounded?
Citations already told you where an answer came from. Now they tell you how strongly the source supports it.
Every citation carries a High, Medium or Low label for how closely the cited passage matches the statement, optionally with the exact percentage. The chat also says how many documents were searched and how many passages matched. Administrators tune the cut-offs after evaluating real answers, so the labels reflect your documents rather than a generic default.
🤝 More Models, Your Choice
An app is no longer tied to a single model. Each app can use the model that suits it, and administrators pick a default chat model and a default image model separately.
Several providers can be offered side by side, including EU-hosted ones, and each model shows the region it runs in, so data residency is visible right where the choice is made. Image generation gains a second route through Black Forest Labs' own API, with an EU endpoint.
🔒 Second Factor by Text Message
Two-factor authentication now offers a code by SMS as well as an authenticator app, and users choose the method that suits them. Two-factor authentication can also be required for individual groups only, on top of or instead of the organisation-wide setting.
🎨 Your Colour, Your Chat History
Administrators set the organisation's primary colour for buttons and highlights, and basebox warns when a shade would be hard to see in light or dark mode.
Past conversations get their own page. Search them, filter them by date, or delete them all at once. Screenshots and other images can be pasted straight into the chat, and if the assistant ever stops without answering, the chat says so and offers to try again.
🧹 Retention You Control
Uploaded documents can now be deleted automatically after a chosen number of days, keeping a server's storage from filling up. Audit log retention changes take effect immediately, without a restart. Administrators also decide where web search appears: as a toggle in every chat, or as a dedicated Web Search app in the sidebar.
🔧 Fixed Along the Way
- Saving a connector's settings no longer clears its title and description.
- Connectors that need nothing from the user no longer ask for credentials.
- Cited passages containing tables, images or links render as text rather than raw markup.
- The chat scrolls all the way to the end of a finished answer.
- Generated PowerPoint files open without repair prompts.
- The usage overview counts everything that was used, including speech-to-text and chats that call connectors.
- An organisation is no longer limited to 100 custom apps.
🚀 Upgrading
1.9.0 adds twelve database migrations, applied automatically at startup on installations with automatic migration enabled. Connectors registered before this release keep working unchanged, and existing connectors and organisation settings are carried over into every organisation. The new per-app and always-on switches default to today's behaviour, so nothing changes until an administrator decides it should.
Installations that run models locally get a newer inference engine. It supports more recent model architectures, but graphics cards older than the Turing generation are no longer supported.
Questions about upgrading your installation? Your basebox contact is happy to help.