Blocks API: insert, move, delete, and render blocks
Manage blocks in the editor: create, delete, update, and reorder content.
Reaching the editor instance
The methods below run on the editor you created with new Blok(). They are available once editor.isReady resolves.
// You already hold the instance returned by the constructor.
const editor = new Blok({ holder: 'editor' });
await editor.isReady;
// Call any API method on it.
editor.caret.setToLastBlock('end');Methods
blocks.clear()
Promise<void>Remove all blocks from the editor.
When to use
Same as the top-level clear(), exposed on the blocks module. Use it when you already hold editor.blocks.
await editor.blocks.clear();
// All content removed; editor keeps one empty paragraphblocks.render(data)
Promise<void>Render passed JSON data as blocks, replacing the current document. It is echo-safe: when the incoming document is structurally equal to the current saved content (time/version are ignored), the call is a caret-preserving no-op. So the data → render → onSave → setState → data round-trip needs no dedupe on your side.
It accepts the loose wire shape (LooseOutputData). The editor deep-clones the data, so the passed object is never mutated or retained. You can pass frozen store state (Redux, Immer) directly.
A render that replaces the document also clears undo history, so undo cannot bring the previous document back.
When to use
Replaces the whole document. Echoing the editor's own output back is a caret-preserving no-op, so you need no dedupe on your side. To add blocks without clearing, use insert() or insertMany().
const data = {
blocks: [
{ id: '1', type: 'paragraph', data: { text: 'Hello World' } },
{ id: '2', type: 'header', data: { text: 'Title', level: 1 } }
]
};
await editor.blocks.render(data);blocks.renderFromHTML(data)
Promise<void>Render HTML string as blocks by converting it to block format.
When to use
Useful for importing legacy HTML, but the result is only as good as the tools' paste handlers. When you control the source, JSON via render() keeps more fidelity.
const html = '<h1>Title</h1><p>Hello World</p>';
await editor.blocks.renderFromHTML(html);
// HTML is converted to appropriate blocksblocks.importMarkdown(md, options?)
Promise<OutputData>Converts a Markdown string to blocks and renders them. This REPLACES the current document, because it calls blocks.render() internally.
The converter is lazy-loaded on the first call. The resolved OutputData is the document that was rendered. options is a MarkdownImportConfig (tool mapping, GFM toggle, micromark/mdast extensions).
To add blocks instead of replacing them, use markdownToBlocks() from the standalone @bloklabs/core/markdown subpath together with blocks.insertMany().
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
md | string | Required | — | Markdown source string. |
options | MarkdownImportConfig | — | undefined | Tool mapping, GFM toggle, and micromark/mdast extensions. |
const data = await editor.blocks.importMarkdown('# Title\n\n- one\n- two');
// The whole document is replaced; data is the rendered OutputDatablocks.exportMarkdown()
Promise<string>Serializes the current document to Markdown. It is the outbound twin of importMarkdown. Blocks are read through the Saver, so the output reflects the saved (validated) document, not the raw DOM.
The promise resolves to '' when there is nothing to save. Blocks owned by a table cell are serialized inside the pipe table, not repeated as loose lines.
What Markdown cannot express is degraded.
- Table
colspan/rowspanand heading columns are dropped. - A table with no heading row gets an empty header row, because GFM requires one.
const md = await editor.blocks.exportMarkdown();
// → '# Title\n\n- one\n- two'blocks.delete(index?, setCaret?)
Promise<void>Remove the block at the specified index, or current block if no index provided.
When to use
Without an index it deletes the focused block. Deleting the last remaining block automatically inserts a fresh empty paragraph, so the editor is never left without a caret target.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
index | number | — | current block index | Index of the block to delete. |
setCaret | boolean | — | true | Whether to move the caret to the surviving block after deletion. Pass false to avoid stealing the user's caret during programmatic deletion. |
// Delete current block
await editor.blocks.delete();
// Delete block at index 0
await editor.blocks.delete(0);
// Delete without moving the user's caret (programmatic deletion)
await editor.blocks.delete(0, false);blocks.move(toIndex, fromIndex?)
voidMoves a block to a new position. If fromIndex is not provided, moves the current block.
When to use
toIndex is in post-removal index space. For a forward move, subtract 1 or the block lands one slot too far.
// Move current block to top
editor.blocks.move(0);
// Move block from index 2 to index 0
editor.blocks.move(0, 2);blocks.getBlockByIndex(index)
BlockAPI | undefinedGet the BlockAPI object for the block at the specified index.
When to use
Returns undefined for out-of-range indexes. Calling a method on an undefined result throws, so check the return value first.
const block = editor.blocks.getBlockByIndex(0);
if (block) {
console.log(block.id, block.name);
}blocks.getById(id)
BlockAPI | nullGet the BlockAPI object for the block with the specified ID.
When to use
The stable way to reference a block across edits, since indexes shift. Returns null if the id no longer exists.
const block = editor.blocks.getById('block-123');
if (block) {
await editor.blocks.update(block.id, { text: 'New content' });
}blocks.getCurrentBlockIndex()
numberGet the index of the currently focused block.
When to use
Reflects the block holding the caret. Returns -1 when nothing is focused, so check the value before using it as an index.
const index = editor.blocks.getCurrentBlockIndex();
console.log('Current block index:', index);blocks.getBlockIndex(blockId)
number | undefinedGet the index of a block by its ID.
When to use
Finds where a known block sits right now. Pair it with getById() when you have an id but need a position.
const index = editor.blocks.getBlockIndex('block-123');
if (index !== undefined) {
console.log('Block is at index:', index);
}blocks.getBlockByElement(element)
BlockAPI | undefinedGet the BlockAPI object for the block containing the given HTML element.
When to use
Bridges from a raw DOM node (e.g. a click target) back to its block. Useful in custom event handlers.
document.addEventListener('click', (e) => {
const block = editor.blocks.getBlockByElement(e.target);
if (block) {
console.log('Clicked on block:', block.id);
}
});blocks.scrollToBlock(id)
voidScrolls a block into view, selects it, pulses the arrival highlight and announces the navigation to assistive tech. It is the public counterpart of the boot-time URL-hash scroll. It does nothing when no block with that id is in the document.
A new jump replaces any blocks that are already selected. Pass { select: false } as the second argument to jump without selecting, as in-page navigation such as a table of contents does: a selected block is deleted by the next Backspace.
Framework adapters that mount into a detached holder (React/Vue/Angular) render seeded content before it joins the page, so the boot hash scroll defers. @bloklabs/react drains it automatically once the holder connects.
For deep linking, call this method yourself after the editor is ready.
editor.blocks.scrollToBlock(nodeId);blocks.getChildren(parentId)
BlockAPI[]Get all child blocks of a parent container block.
When to use
Returns the direct children of a container block (a column, toggle, or database). Empty array if it has none.
const children = editor.blocks.getChildren('parent-block-id');
children.forEach(child => {
console.log('Child:', child.id);
});blocks.setBlockParent(blockId, parentId)
voidReparents a block. It updates the block's parentId and the parent's contentIds through core's single reparent chokepoint.
- Pass
nullto move the block back to the root level. - An unknown
blockIddoes nothing and logs a warning. - A
parentIdthat would make the block a descendant of itself throws.
The block's place in the document order follows its new parent.
- If the block already sits inside the parent's subtree, it does not move. That means it comes right after the parent, or right after one of the parent's descendants.
- Otherwise the block moves to the end of the parent's subtree. Its own children move with it.
- A table or a database places its own children, so a block that joins one does not move.
- A block that leaves for the root or for one of its ancestors moves, with its children, to just after the subtree it left, unless it already sits there. To move several siblings out of a container and keep their order, move the last one first.
// Move a block into a container block
editor.blocks.setBlockParent('child-block-id', 'parent-block-id');
// Move it back to the root level
editor.blocks.setBlockParent('child-block-id', null);blocks.insertInsideParent(parentId, insertIndex, childData?, toolName?)
BlockAPIInserts a block as a child of parentId atomically. The creation and the parent assignment are grouped into ONE undo entry, so a single Cmd+Z removes it completely. That is better than insert() followed by a reparent, which is two entries.
insertIndexis the FLAT document index where the child should appear. It is clamped into the parent's subtree.- An index at or before the parent becomes the slot right after the parent.
- An index past the end of the subtree becomes the end of the subtree.
toolNamepicks the child's block tool and defaults toconfig.defaultBlock.- A tool that is restricted inside table cells is demoted to the default block when the new child would land inside one.
- An unregistered name throws before anything is written.
childDatadefaults to{ text: '' }for the default block.- For any other
toolNameit defaults to{}, which lets the tool apply its own defaults.
- For any other
const parentIndex = editor.blocks.getBlockIndex('parent-block-id') ?? 0;
const child = editor.blocks.insertInsideParent(
'parent-block-id',
parentIndex + 1,
{ text: 'Nested content' },
);
// A TYPED child in the same single undo entry — no insert-then-reparent
editor.blocks.insertInsideParent(
'parent-block-id',
parentIndex + 1,
{ text: 'Nested heading', level: 2 },
'header',
);blocks.insertAt(type?, data?, options?)
BlockAPIInserts a new block at a place named by its parent and its siblings, not by a flat index. It is ONE undo entry.
positionis'start','end'(the default),{ before: id }or{ after: id }.{ after: id }lands after that sibling's whole subtree, not among its children.
parentIdnames the parent;nullis the root.- When you omit it,
before/afteruse the sibling's parent, and'start'/'end'use the root.
- When you omit it,
replace: idputs the new block in that block's place. It keeps the parent; the children follow the turn-into rules. It cannot be combined withparentIdorposition.focusdefaults tofalse, so a programmatic insert does not move the current block.- A tool the parent does not allow is demoted, as with
insertInsideParent.
It throws, and changes nothing, when the parent, the sibling or the replaced block is not found, or when the sibling is not a child of parentId.
// Last child of a toggle
editor.blocks.insertAt('paragraph', { text: 'Inside' }, { parentId: toggleId });
// Right after a block's whole subtree, at the block's own level
editor.blocks.insertAt('header', { text: 'Next', level: 2 }, { position: { after: blockId } });
// First block of the document
editor.blocks.insertAt('paragraph', { text: 'Top' }, { parentId: null, position: 'start' });blocks.moveTo(id, target)
voidMoves a block, with its whole subtree, to a place named by its parent and its siblings. It is ONE undo entry, even when the parent changes.
positionworks as ininsertAt.{ after: id }lands after that sibling's whole subtree.parentIdnames the new parent;nullis the root. When you omit it,before/afteruse the sibling's parent, and'start'/'end'use the root.
It throws, and changes nothing, when:
- a block is not found, or the sibling is not a child of
parentId; - the target is the block itself or inside its own subtree;
- the move would leave or enter a container that owns its children (a table, a column list), a column, or a table cell;
- the new parent does not allow the tool (
childTools).
// Into a toggle, as its first child
editor.blocks.moveTo(blockId, { parentId: toggleId, position: 'start' });
// Out of a container, right after it
editor.blocks.moveTo(childId, { position: { after: containerId } });blocks.getBlocksCount()
numberGet the total number of blocks in the editor.
When to use
Counts every block including nested children (columns, toggles, table cells), not just top-level blocks. Use it for bounds before an index-based insert() or move().
const count = editor.blocks.getBlocksCount();
console.log('Total blocks:', count);blocks.insert(type?, data?, config?, index?, needToFocus?, replace?, id?, tunes?, origin?)
BlockAPIInsert a new block with full control over its properties and position.
When to use
Every parameter is optional. Call insert() for an empty paragraph, or pass type/data/index for full control. Pass replace: true to overwrite the block at index.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
type | string | — | config.defaultBlock | Tool name to instantiate. |
data | BlockToolData | — | {} | Initial tool data for the new block. |
config | ToolConfig | — | {} | Ignored. It is accepted only to keep the positional signature stable. The editor binds it to Pass |
index | number | — | current block index + 1 | Position to insert the block at. The new block also takes the parent that this position implies. That holds when
A column row never takes the block this way. Right after a table or a database, or a block inside one, no parent is inferred. The block is a root block, unless it comes right after a block in a column. With The parent's |
needToFocus | boolean | — | true | Whether to move focus to the inserted block. Pass false to insert without moving the caret. |
replace | boolean | — | false | Replace the existing block at index instead of inserting around it. |
id | string | — | auto-generated | Custom id for the new block. |
tunes | { [name: string]: BlockTuneData } | — | undefined | Optional tune data applied at creation, keyed by tune name. |
origin | BlockOrigin | — | 'api' | Why the block is being created. It is handed to the tool constructor as
|
Errors
No
typeis given and nodefaultBlockis configured.Could not insert Block. Tool name is not specified.
Pass an explicit
type, or setdefaultBlockin the editor config.The resolved tool name is not registered in the editor.
Could not compose Block. Tool «<type>» not found.
Register the tool in the editor's
toolsconfig before inserting a block of that type.replace: trueis passed but no block exists atindex.Could not replace Block at index <index>. Block not found.
Check
indexagainstblocks.getBlocksCount()before calling withreplace: true.
// Insert after the current block with the default type
const block = editor.blocks.insert();
// → BlockAPI { id: 'kP3xQ...', name: 'paragraph', ... }
// Insert paragraph with data at index 0
const block = editor.blocks.insert('paragraph', { text: 'Hello' }, undefined, 0);
// Insert with custom ID
const block = editor.blocks.insert('header', { text: 'Title' }, undefined, undefined, undefined, undefined, 'custom-id');
// From your own slash menu / toolbar: declare the user gesture so a container
// tool seeds its default children (see BlockOrigin on the tool contract)
const block = editor.blocks.insert('column_list', undefined, undefined, index, undefined, undefined, undefined, undefined, 'user');blocks.insertMany(blocks, index?)
BlockAPI[]Insert multiple blocks at once. When the index is omitted, the blocks are appended at the end of the document.
When to use
Bulk-insert in one operation, and one undo step. That is far cheaper than looping insert() for large pastes or imports.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
blocks | OutputBlockData[] | LooseOutputBlockData[] | Required | — | The blocks to insert. Loose wire blocks are accepted. A |
index | number | — | end of document | Position to insert at. When omitted, defaults to appending at the end of the document. |
Errors
The provided
indexis negative.Index should be greater than or equal to 0
Pass an
indexof 0 or greater, or omit it to append.
const blocksToInsert = [
{ id: '1', type: 'paragraph', data: { text: 'First' } },
{ id: '2', type: 'paragraph', data: { text: 'Second' } }
];
const inserted = editor.blocks.insertMany(blocksToInsert, 0);
console.log('Inserted:', inserted.length, 'blocks');blocks.composeBlockData(toolName)
Promise<BlockToolData>Create empty block data for the specified tool type.
When to use
Builds a tool's default data without inserting anything. Useful when you need a valid empty payload for insert() or update().
Errors
toolNameis not a registered tool.Block Tool with type "<toolName>" not found
Register the tool in the editor's
toolsconfig first.
const emptyData = await editor.blocks.composeBlockData('paragraph');
// Returns: { text: '' } or appropriate empty state for the toolblocks.update(id, data?, tunes?)
Promise<BlockAPI>Updates a block's data and/or tunes. The data update is applied IN PLACE when the block's tool implements setData(newData) on its prototype. Every React/Vue/Angular block does, as do the built-in header, list, code, toggle and table tools.
In place means the same block instance, the same DOM holder and the same mounted component keep living. So ephemeral tool state, adopted child blocks and the caret all survive. That is what makes update() safe to call on every keystroke, for example when renaming a card while the user types.
Three cases fall back to recomposing the block:
- tools without
setData - a tool whose
setDatareturnsfalse(it needs a different DOM shape) - any call that passes
tunes
Recomposing means a fresh tool instance replaces the old one, which is destroyed.
When to use
Patches data/tunes in place, keeping the block id and type. To change the type, use convert() instead.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
id | string | Required | — | Id of the block to update. |
data | Partial<BlockToolData> | — | undefined | Partial data merged into the block's existing data. |
tunes | Record<string, BlockTuneData> | — | undefined | Tune data merged into the block's existing tunes. |
Errors
No block exists with the given
id.Block with id "<id>" not found
Confirm the id with
blocks.getById()before callingupdate().
// Update block data
const block = await editor.blocks.update('block-123', { text: 'New text' });
// Update with tunes
const block = await editor.blocks.update('block-123', undefined, { alignment: 'center' });
// Guard against an id that no longer exists (e.g. the block was deleted
// concurrently) instead of letting the rejection surface unhandled
try {
await editor.blocks.update(staleId, { text: 'New text' });
} catch {
console.warn('Block was already removed, skipping update');
}blocks.convert(id, newType, dataOverrides?)
Promise<BlockAPI>Convert a block to a different type. Both tools must support conversion config.
When to use
Both the source and target tools must declare conversionConfig, otherwise the call throws. Preserves text where possible.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
id | string | Required | — | Id of the block to convert. Its tool must declare conversionConfig.export. |
newType | string | Required | — | Name of the registered tool to convert to. Its tool must declare conversionConfig.import. |
dataOverrides | BlockToolData | — | undefined | Data fields to overwrite on the resulting block after conversion. |
Errors
No block exists with the given
id.Block with id "<id>" not found
Confirm the id with
blocks.getById()before callingconvert().newTypeis not a registered tool.Block Tool with type "<newType>" not found
Register the target tool in the editor's
toolsconfig.The source tool has no conversionConfig.export, the target tool has no conversionConfig.import, or neither does.
Conversion from "<sourceType>" to "<newType>" is not possible. <ToolName(s)> tool(s) should provide a "conversionConfig"
Add a conversionConfig to whichever tool is missing one, or convert through an intermediate tool that supports both directions.
// Convert paragraph to header
const headerBlock = await editor.blocks.convert('block-123', 'header', { level: 2 });
// Convert with data overrides
const headerBlock = await editor.blocks.convert('block-123', 'header', { text: 'New Title', level: 1 });
// Not every pair of tools supports conversion — guard it rather than
// assuming the target type is always convertible
try {
await editor.blocks.convert('block-123', 'table');
} catch (error) {
console.warn('Conversion not supported between these tools:', error);
}blocks.splitBlock(currentBlockId, currentBlockData, newBlockType, newBlockData, insertIndex)
BlockAPIAtomically split a block by updating the current block and inserting a new block. Both operations are grouped into a single undo entry.
When to use
Use it for caret-aware splits, for example Enter mid-text. The update and the insert land as a single undo entry.
// Split a paragraph at cursor position
const newBlock = editor.blocks.splitBlock(
'current-block-id',
{ text: 'First part' },
'paragraph',
{ text: 'Second part' },
1
);blocks.stopBlockMutationWatching(index)
voidStop mutation watching on a block at the specified index. Use this to prevent spurious block-changed events during block replacement operations.
When to use
A niche method. Call it before you replace a block's DOM programmatically, to suppress spurious block-changed events. The next render re-arms it.
// Replace a block without triggering change events
editor.blocks.stopBlockMutationWatching(0);
// Perform block replacement...
// Mutation observer will not fire for this blockblocks.startBlockMutationWatching(blockId)
voidRe-arms mutation watching on a block previously silenced by stopBlockMutationWatching. It takes an id, not an index, because inserts and replacements between the two calls shift indexes. An id that no longer exists is silently skipped: a block replaced in place was constructed with its own watcher.
const blockId = editor.blocks.getBlockByIndex(0)?.id;
editor.blocks.stopBlockMutationWatching(0);
// Perform block replacement...
if (blockId) {
editor.blocks.startBlockMutationWatching(blockId);
}blocks.transact(fn)
voidGroups every block operation performed inside fn into a single undo entry. fn must be SYNCHRONOUS. Operations that land after an await are no longer part of the group. Use it so structural edits are not partially undoable.
editor.blocks.transact(() => {
editor.blocks.insert('paragraph', { text: 'One' });
editor.blocks.insert('paragraph', { text: 'Two' });
});
// A single Cmd+Z removes both blocksblocks.beginTransaction()
voidOpens an undo group that stays open across async boundaries. Every block operation until endTransaction() lands in one undo entry.
Use it for pointer gestures that mutate the document continuously, such as dragging a table's corner to add rows. transact() cannot help there, because it only wraps a synchronous function. Pair every call with endTransaction().
editor.blocks.beginTransaction();
// ... continuous mutations across async boundaries ...
editor.blocks.endTransaction();blocks.endTransaction()
voidClose the undo group opened by beginTransaction().
editor.blocks.beginTransaction();
// ... continuous mutations across async boundaries ...
editor.blocks.endTransaction();blocks.transactWithoutCapture(fn)
voidRun block operations without recording anything in the undo history. Use it for auto-repair and normalization (e.g. ensuring an empty cell always has a block) that Cmd+Z should never step through.
Pass { derivedFrom: blockId, from: ['url'] } when the operations are data worked out from those keys of a block, such as a finished upload that replaces the block. They still add no undo step and keep redo, but they join the step that wrote those values: undoing it takes them back, and redo brings them back. An update() started inside fn counts too, although it finishes later. While a save of that block is still in flight, fn runs once it lands, after the call returns.
editor.blocks.transactWithoutCapture(() => {
editor.blocks.insert('paragraph', {}, undefined, 0);
});
// Nothing was added to the undo historyblocks.setPointerDragActive(active)
voidTell core that a pointer drag interaction started or ended. While it is active, DOM-mutation-triggered Yjs syncs are suppressed. That way browser DOM churn during the drag cannot corrupt Yjs state.
editor.blocks.setPointerDragActive(true);
// ... run the drag gesture ...
editor.blocks.setPointerDragActive(false);Properties
| Property | Type | Description |
|---|---|---|
isSyncingFromYjs | boolean | A readonly getter. It is true while a Yjs sync operation (undo/redo) is in progress. Tools read it to skip cleanup that would fight undo state. This is a PROPERTY on |
isPointerDragActive | boolean | A readonly getter. It is true while a pointer drag interaction is active. Framework adapters read it to defer a programmatic |
isApplyingRemoteChange | boolean | A readonly getter. It is true while the editor applies a change another client made. A peer's new block can arrive before the write that gives it a parent. A container tool reads this to tell such a block from one this client adds in its own sync window. |