Skip to main content
A source is where a workspace’s documents come from. Each source syncs its documents into the workspace, and a Document scan curates them. Sources live on Context.

​
Add a source

Press Add context on any Context page.
1

Source

The connected accounts come first: with an Atlassian site connected in Settings › Connections, a Jira row and a Confluence row, each naming the site. Then the two kinds that need no account: Repository and Documentation site.
2

Scope

For a Repository source, pick the repository and set its Include and Exclude patterns, one per line. The defaults include docs/** and **/*.md, and exclude changelog and license files. In hosted mode the repositories are the ones your connected GitHub accounts can see, whether or not Code has connected them. Pick the Account first when there is more than one. In local mode they are the folders connected in Code. A repository can have only one Repository source.For a Documentation site, enter its llms.txt URL.For Jira, enter the Project key and, if you want less than the whole project, a JQL filter. For Confluence, enter the Space key.
3

Check, then add

Press Check. It counts the files, pages or issues the scope would yield and lists the first few, and it stores nothing. Then press Add and sync. The source’s page opens and reads Syncing until the first sync lands.

​
Who reads a source

A repository reads only the sources it is linked to.
  • A Repository source is read by the repository it scopes, as long as Code has connected that repository.
  • Every other source is read by nobody until you link it.
Link a source in three places: the Read by switches on the source’s page, the switches on a repository’s Context tab, or the Context step when you connect a repository. Changing a repository’s links starts Flow setup or Flow generation for that repository.

​
The Sources list

Context opens on the Sources list, worst first. The columns are Source, Kind, Repositories, Status, Documents and Last sync. A source’s status is one of these:

​
A source’s page

Open a source to act on it. The header carries its status and its actions: Sync now, Pause or Resume, and for every kind but a repository, Remove. The page has three sections:
  • Scope shows what the source reads. For a repository that is the repository, its branch, and its include and exclude patterns. For a site it is the llms.txt URL. For Jira it is the project key and the JQL filter; for Confluence, the space key. Save stores a new scope and starts a sync.
  • Read by lists every connected repository with its switch. A repository source’s own repository is always on.
  • Syncs lists each sync with how many documents it added, changed, removed and left unchanged. The document count links to Documents, narrowed to this source.

​
When a source syncs

  • When it is added, and when its scope is saved.
  • When you press Sync now.
  • A Repository source syncs when its repository’s default branch receives a push. A local folder syncs when a file in it changes.
  • A documentation site, a Jira project and a Confluence space sync once a day. The server checks every hour and syncs each one whose last sync is more than a day old, unless it is paused.
  • A source that has never synced is picked up by the next check, and when the server starts.
A sync that added, changed or removed documents marks the corpus stale: the Scan button shows a dot until you run the next Document scan.

​
Documentation sites

A site must publish an llms.txt index. A sync reads the index and fetches, as markdown, the pages it lists on the site’s own origin. There is no crawling of sites without an llms.txt.

​
Next steps

Document scan

How the synced documents become one curated corpus.

Resolving conflicts

Settle the disagreements the scan finds between documents.