Knowledge
Knowledge graph
See the people, projects and things your documents describe, and declare your own in a document.
# Knowledge graph The knowledge graph links the things your documents talk about: people, projects, customers, tasks. Instead of reading documents one by one, the brain can follow a link from one thing to the next to answer questions about how things relate. ## Where you see it The graph is the second view behind the brain on the skill tree home. Click the brain, then use the switch to choose **Graph**. If the graph is not turned on for your workspace, or has nothing in it yet, the switch is not there. You can search with "Find an entity", filter by kind in the legend, click a node for details and use **Fit to View**. Very large graphs show the most connected nodes only and say so. ## Where the graph comes from Connected apps can add the things they know about. Documents can add their own. Nothing is fixed in advance: the kinds of things and links come from your connectors and documents. ## Declare things in a Markdown document A Markdown document can start with a `graph:` block in its front matter. Each node has a `label`, a `key`, a `name`, optional `aliases` and optional `props`. Each edge has a `type`, a `from` and a `to`, written as `Label:key`. ```yaml --- graph: nodes: - label: Person key: ana@example.com name: Ana Petrova aliases: [Ana] props: team: Finance - label: Project key: apollo name: Project Apollo edges: - type: OWNS from: Person:ana@example.com to: Project:apollo --- ``` The block is removed before the document is read for search. The document gets a link to what it declares. Later documents that mention a declared name or alias are linked to it too. ## Tips - For a `Person`, use the email address as the key. Email senders and recipients then land on the same person. - `Document`, `Section` and `Folder` are reserved labels. Pick another label. - Name both ends of an edge as `Label:key`, the same way the nodes declare them. - After loading many documents at once, the deploying team can call the admin-only `POST /api/knowledge/documents/extract-graph` route so every document sees the full set of things.The knowledge graph links the things your documents talk about: people, projects, customers, tasks. Instead of reading documents one by one, the brain can follow a link from one thing to the next to answer questions about how things relate.
Where you see it
Section titled “Where you see it”The graph is the second view behind the brain on the skill tree home. Click the brain, then use the switch to choose Graph. If the graph is not turned on for your workspace, or has nothing in it yet, the switch is not there. You can search with “Find an entity”, filter by kind in the legend, click a node for details and use Fit to View. Very large graphs show the most connected nodes only and say so.
Where the graph comes from
Section titled “Where the graph comes from”Connected apps can add the things they know about. Documents can add their own. Nothing is fixed in advance: the kinds of things and links come from your connectors and documents.
Declare things in a Markdown document
Section titled “Declare things in a Markdown document”A Markdown document can start with a graph: block in its front matter. Each node has a label, a key, a name, optional aliases and optional props. Each edge has a type, a from and a to, written as Label:key.
---graph: nodes: - label: Person key: ana@example.com name: Ana Petrova aliases: [Ana] props: team: Finance - label: Project key: apollo name: Project Apollo edges: - type: OWNS from: Person:ana@example.com to: Project:apollo---The block is removed before the document is read for search. The document gets a link to what it declares. Later documents that mention a declared name or alias are linked to it too.
- For a
Person, use the email address as the key. Email senders and recipients then land on the same person. Document,SectionandFolderare reserved labels. Pick another label.- Name both ends of an edge as
Label:key, the same way the nodes declare them. - After loading many documents at once, the deploying team can call the admin-only
POST /api/knowledge/documents/extract-graphroute so every document sees the full set of things.