Logseq organizes knowledge as graphs made of pages, journals and addressable blocks. Each page is an outline rather than one uninterrupted document. Indenting a block changes its parent-child position in that outline; it does more than alter the left margin. References, backlinks, tasks and queries use this structure to find related blocks.
A reference points to content without copying it
A page reference connects the current block to another page. A block reference targets one block directly. Neither operation creates an independent duplicate. An embed displays referenced content in another context, while later edits still belong to the original target.
Indentation affects the hierarchy that queries and task workflows read. Moving a task under another block can change its context or which parent a query returns. A pasted outline that looks visually similar can therefore behave differently when one level of indentation changes.
File graphs and database graphs store different models
A file graph stores pages through Markdown or Org files and represents many properties as text syntax. A database graph stores its structure in SQLite and treats tags or properties as database entities with schemas. Plugins or workflows that expect file-property syntax do not automatically understand the database representation.
The file-to-database importer converts the old material rather than attaching the database interface to the original files. It can remap task statuses, detect property types and handle tags according to import choices. Assets receive copied files with unique identifiers. The imported graph therefore needs inspection before the old and new workspaces are treated as interchangeable.
Export choice determines what survives
A database graph can export SQLite with or without assets, editable EDN, standard Markdown or selected public pages. Standard Markdown creates readable files but omits block properties and does not preserve every timestamp or database field. It is unsuitable as a full-fidelity reconstruction of the database graph.
Build EDN captures graph data, properties and tags in an editable Logseq representation. The official documentation does not recommend EDN as the only backup. SQLite plus the required assets serves another recovery purpose, while Markdown serves portability and reading outside Logseq. Choosing one format for every purpose discards a different part of the workspace.
Publishing creates a separate remote result
Publishing a page creates a remote page from the selected local state. Later edits inside Logseq do not update that publication automatically; the user must publish again. A public page also does not turn the entire graph into a synchronized collaboration space.
Templates, queries and flashcards operate on the graph’s blocks and properties. Their output can depend on the storage model and hierarchy underneath them. Before changing graph type or reorganizing an outline, an export and a test copy give the user a place to inspect changed statuses, links and property behavior without relying on appearance alone.
