lb_capture

Submits content for ingestion into the knowledge graph. The call returns immediately with a pending ID — processing (extraction, summarization, embedding, link detection) happens asynchronously and typically completes within one minute.

Use lb_add_node if you need the node to be available immediately.

Parameters

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | title | string | Yes | Title for the capture | | content_type | string | Yes | Type of content being captured (see values below) | | why | string | Yes | Reason for capturing — stored as context metadata | | url | string | No | Source URL of the content | | raw_content | string | No | Raw text or Markdown content (1–500,000 characters) | | captured_at | string | No | ISO 8601 timestamp override (defaults to now) |

content_type values

| Value | Description | |-------|-------------| | article | Web article or blog post | | tweet | Tweet or social media post | | repo | GitHub repository | | pdf | PDF document | | document | Word doc or other document format | | audio | Audio file or podcast | | note | Plain text note | | video | Video (YouTube, Loom, etc.) | | image | Image file (PNG, JPEG, etc.) |

Response

{
  "id": "...",
  "status": "pending"
}

The id is the capture ID. Once ingestion completes, a node will be created in your graph with a reference back to this capture.

Usage notes

• At least one of url or raw_content should be provided. If both are supplied, raw_content is used as the body and url is stored as the source reference.
• Indexing typically completes within one minute. Querying immediately after capture may not return the new node yet.
• The why field is required and stored alongside the node. It is used by retrieval tools to surface why you saved something, which improves context quality over time.