Skip to main content
Use this playbook for regular content checks. It covers failed documents, schedules, new pages, connected apps, the text snippet, custom responses, and old documents. The CLI calls content “knowledge”, so the commands start with sitegpt knowledge. For how sync and scan work, see Keep your content up to date.

​
Before you start

  • Plan: the Growth plan or above for CLI access. See Plans and limits.
  • Access: a CLI login with these scopes. sitegpt login with no --scope includes them all.
    conversations:write lets you send the test message at the end.
  • Input: the chatbot ID. Save it for the commands below:

​
Steps

1

Check content health

Look for failed documents, old documents, duplicate documents, a long queue of pending documents, a sitemap with no sync or scan schedule, and custom responses with old answers.
2

Resync failed or old documents

Pick one selector:
3

Remove page noise, if needed

If a page picked up menus or footers, change how SiteGPT reads it. Then resync it.
4

Set sync and scan schedules

To scan a sitemap now, run:
To stop a schedule, disable the job. The trained pages stay.
Keep these rules in mind:
  • Sync works only for website, sitemap, and GitHub jobs. It does not delete pages.
  • Scan works only for sitemap jobs. It adds new URLs and deletes pages whose URLs left the sitemap.
  • If sync and scan use the same frequency, only the sync runs on that schedule.
  • If the plan does not include a frequency, SiteGPT sets it to NEVER and returns a warning.
5

Add new website content

Use a sitemap if the site has one:
Otherwise, crawl the site:
To add a few exact pages, add links:
6

Update connected apps

List the source connections:
To pick more files from an app that uses a file picker, authorize again, list the selected files, and ingest them:
To ingest only some files, pass their IDs:
For GitHub:
For Confluence:
Only GitHub sources accept a --sync schedule. For other apps, refresh the documents with knowledge documents resync.
7

Replace the text snippet

Each chatbot has one text snippet. The command replaces the whole snippet. Read the current text first:
Then write the new text:
8

Update custom responses

List the custom responses that have no approved answer yet:
Add, update, or delete answers:
9

Delete old documents

Find the documents first:
Delete them by exact ID, or delete only failed documents:
Do not use --all --yes unless the user asked for a broad delete.

​
Check that it works

The failed list should be empty, or hold only documents that you report to the user. Read the test answer. It should use the content you added.

​
Hand back to the human

Report:
  • The content you added or updated.
  • The documents that still fail.
  • The sync jobs you changed, and any frequency that SiteGPT set to NEVER.
  • The documents and custom responses you deleted.
  • Follow-up work that needs the user’s approval.