Updates on the Logseq DB Alpha

Logseq DB Alpha is moving fast—so fast that a serious crash tied to creating a new property was reported and patched within hours—while the project deliberately limits risk by running the database in-browser and restricting export to SQL light. The testing window also brought a flood of feedback from heavy Logseq users, especially plugin developers and power users who translate day-to-day workflow into actionable bug reports. Communication shifted from Discord chat toward a forum-and-chat hybrid so each fix can be tracked per issue, with bugs logged in GitHub and patches landing quickly during “office hours.”

Beyond the pace, the most consequential change is structural: everything becomes a node. In the “clean graph” demo, pages, journals, tags, and tasks are treated as the same underlying building blocks (with blocks also behaving like nodes). Tags act like a parent-child, inheritable relationship—anchored by a “root tag” that can propagate changes across the system. That uniformity is meant to simplify storage and enable new compositions, such as introducing a “long form page” type for writing without the outliner, or building richer dashboards by combining properties and types rather than relying on separate mechanisms.

The interface is also being cleaned up, but that comes with a tradeoff in discoverability. A “configure” control governs how tag pages display properties, and many key actions are tucked behind that button. The demo shows how properties can be defined with types and constraints (including multiple values), positioned within a block (beginning vs end), and decorated with icons. Those property definitions then flow into tag pages—for example, adding an “author” property to a “book” set and having author pages appear under the book filter. Status becomes a reusable concept: tasks carry a default status set (e.g., backlog, to do, doing), and the same status can be applied to books and projects so filtering works across categories.

Task management remains flexible but raises usability questions. Priority is now handled via status values (low/medium/high), and tasks can be navigated through states using keyboard shortcuts like control enter, with quick entry via slash commands (e.g., “/low” and “/deadline”). Yet the system doesn’t auto-detect deadlines from markdown, so newcomers may struggle to discover that task features exist and how to set them.

Import/export is still rough around the edges and primarily intended for testing. The alpha supports SQL light export and an import flow that can convert tags into links and pages, producing side-by-side error messages while edge cases are refined. Performance tests suggest the foundation is viable—loading a large text like the first 13 chapters of Moby Dick eventually works, though slowly, while inserting around 1,300 notes into a single page reportedly holds up better. The next bottlenecks are expected to be query improvements and making editable, table-like views on top of the new backend.

Overall, the alpha’s core bet is that a stable data structure unlocks faster development and safer plugin ecosystems: plugin developers can store meaningful structure in properties and tags so content remains usable even if a plugin is disabled. Desktop remains the focus for now, with mobile improvements deferred until after the release, when touchscreen-specific UX can be tackled separately.