Chapter 05 · Knowledge Base
Versions, statuses, and source actions
Every ingestion is a numbered version. The version model, every status label, the drawer detail grid, and what each action button does.
Updated Sep 20, 2026 · 6 min read
On this page
Every time Serpio ingests a source, the result is a numbered version. Versions make the Knowledge Base safe to refresh: the version in use never changes until a newer one has fully succeeded. This guide covers the version model, every status label, the detail grid, and each action button.
The version model
A source keeps one identity. Under it sit versions, numbered from 1. Version history in the drawer lists them newest first, marks the one in use as Active, and shows each state:
queued: created, waiting for a worker.processing: being extracted and indexed.ready: indexed successfully. The newest ready version becomes active.failed: extraction or indexing failed. The error is shown under the entry.superseded: either replaced by a newer ready version, or found identical to the active one and never activated.
The last good version stays in use until a new one activates, so a refresh never leaves a gap. A failed refresh keeps the old version, and the drawer says so: Version 3 remains available. Latest processing issue: followed by the error. An unchanged check produces a superseded version rather than a duplicate, so the version number only moves when content really changed.
Superseded versions are cleaned up after a newer one activates. Serpio retries interrupted processing up to three times before marking the version failed.
Status labels
The list shows one status per source, chosen in this order of precedence.
| Status | Meaning | What to do |
|---|---|---|
| Removed | The URL returned 404 or 410 on a check. | If the page is back, press Check URL now. Otherwise Delete it. |
| Paused | You pressed Pause use. Excluded from writing. | Press Resume use when you want it back. |
| Processing | The first version is queued or running. Nothing is available yet. | Wait for the status strip. |
| Needs attention | The first version failed, so there is nothing to fall back on. | Fix the cause shown in the drawer, then Reprocess original, Replace document, or Check URL now. |
| Skipped | The source was recorded but not indexed in its last run. | Open it and run Check URL now or Reprocess original. |
| Review needed | Today is on or after Review by. | Review the content and update the date. See Effective from and Review by dates. |
| Not effective yet | Today is before Effective from. | Nothing, unless the date is wrong. |
| Legacy index · refresh needed | Only a saved excerpt of this page is indexed, from an older pipeline. | Press Check URL now to index the full page. |
| Legacy index | Indexed by an older pipeline with no version history. | Press Check URL now or Reprocess original to start version history. |
| Ready · refresh failed | A newer version failed; the last good version is still serving. | Fix the cause, then retry with the relevant action. |
| Ready · updating | A newer version is queued or processing; the current one still serves. | Wait; the row returns to Ready on activation. |
| Ready | Active, eligible, nothing pending. | Nothing. |
Before the status, each row shows the usage label: Reference evidence or Brand knowledge, then Facts allowed or Context only. Sources that predate these settings read Usage settings pending; see Usage settings and permissions.
The detail grid
The grey grid in the drawer summarises the source.
- Available version
- The active version number, for example Version 4. When the source is not currently active, the label changes to Last ready version. Sources from the older pipeline read Legacy index; a source that never succeeded reads None yet.
- Use
- Active when the source is eligible, otherwise the status that is blocking it.
- Last checked / Last indexed
- When Serpio last contacted the URL, or when it was indexed if never checked.
- Last changed
- When the content hash last differed from the previous version. Unchanged checks do not move it.
- Next check
- The scheduled time of the next automatic check. Reads Manual for documents and for URLs set to Manually.
- Review by
- The review date, if set.
Reprocess, replace, and check
| Button | Shown for | What it does |
|---|---|---|
| Reprocess original | Documents | Runs extraction and indexing again on the stored file as a new version. Use after a failure. The file itself is unchanged. |
| Check URL now | URLs | Fetches the page now, compares it with the active version, and indexes a new version only if the content changed. |
| Reprocess saved content | URLs with a verified snapshot | Re-indexes the saved full-page text without contacting the site. Hidden when no snapshot exists or the page is Removed. |
| Replace document | Documents | Uploads a new file as the next version, same formats and 20 MiB limit. The active version stays available until the replacement is ready. |
| Original | Documents | Opens a download link for the stored file. The link expires after five minutes. |
Pause versus Delete
Pause use keeps everything, versions included, but removes the source from article retrieval and internal-link candidates. The status reads Paused and the button becomes Resume use. Paused URLs are not checked on schedule, and paused site pages are skipped by the crawler.
Delete asks Remove this source from future article writing? Existing articles and idea collections will stay unchanged. On confirm, the source leaves the list and its versions, passages, vectors, and stored files are cleaned up. Articles already written keep their text and Knowledge used entries. There is no undo.
Legacy index warnings
Sources indexed before versioning show amber warnings until you refresh them. For a URL: This legacy index has no verified version history. Check this URL now to verify its current content. For a document: This legacy index has no version history. Reprocess the original document to verify it. A URL that holds only a short saved excerpt shows Only a previously saved excerpt is indexed. with the status Legacy index · refresh needed. Running the suggested action creates version 1 and clears the warning. Until then the legacy content still serves.
Common questions
Why does my version number not increase after Check URL now?
The page had not changed. Serpio compared the content hash with the active version and marked the new version superseded instead of re-indexing. Last checked still updates.
Can I roll back to an earlier version?
Not from the interface. For a document, use Replace document with the earlier file if you kept it. For a URL, the page's current content is what Serpio indexes.
Related guides
Adding documents to the Knowledge Base
Upload PDF, DOCX, Markdown, or TXT files up to 20 MiB each, follow processing, and know when a document is available for writing.
Read guideAdding URLs and setting refresh policies
Add a public page as a source, choose how often Serpio checks it for changes, and understand how change detection differs from the website crawler.
Read guideEffective from and Review by dates
Time-box a source so Serpio only writes from it while it is current. How the two dates work, the statuses they produce, and how to review and extend.
Read guide