> For the complete documentation index, see [llms.txt](https://witfy.gitbook.io/clowrk/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://witfy.gitbook.io/clowrk/knowledge-base/articles.md).

# Articles

## Types

Pick the type that matches what the article *is*, because it changes where the article surfaces.

| Type                  | Use it for                                                                          |
| --------------------- | ----------------------------------------------------------------------------------- |
| **How-to**            | A procedure someone follows on purpose — provisioning, onboarding, a routine change |
| **Troubleshooting**   | Diagnosing and fixing a symptom. This is the type that carries commands             |
| **FAQ**               | A short answer to a recurring question                                              |
| **Known error**       | A defect with no fix yet, and the workaround that gets people moving                |
| **Policy**            | A rule the team must follow                                                         |
| **Resolution record** | What was actually done on a specific case, kept as evidence                         |

## Lifecycle

```
Draft ──submit──▶ In review ──approve──▶ Published ──▶ Retired ──▶ Archived
  │                    │                      ▲            │
  │                    └──reject──────────────┘            │
  └──────────publish directly───────▶                       └──republish──▶ Published
```

| State         | What it means                                                          |
| ------------- | ---------------------------------------------------------------------- |
| **Draft**     | Being written. Invisible to search, suggestions and AI                 |
| **In review** | Waiting for someone to approve it. Still invisible                     |
| **Published** | Live. This is the only state that is suggested, searched or read by AI |
| **Retired**   | Superseded or no longer true. Removed from suggestions, kept readable  |
| **Archived**  | Final. Kept for history only                                           |

A retired article can be **republished** after being updated. An archived one cannot — archiving is the end of the line, so archive when you are sure.

**Publishing can skip review.** An article can go straight from draft to published — that is how AI-generated drafts and quick fixes get out. Use review when the content is a policy, or when being wrong is expensive.

### Who approved it is recorded permanently

When an article is published, the person who published it is stamped on it and **never removed** — not when it is retired, not when it is archived. That stamp is the audit answer to "who said this was correct?".

One consequence worth being deliberate about: if the same person writes and publishes an article, nobody else has checked it. That is allowed, and sometimes right. It is just not review.

## Versions

Every edit snapshots the previous version before overwriting. The history is never deleted, so you can always see what an article said at the time somebody followed it — which matters when a case went wrong and the question is what the analyst was actually told to do.

## Visibility

| Visibility   | Who sees it                                                             |
| ------------ | ----------------------------------------------------------------------- |
| **Internal** | Your team only, following the same access rules as the rest of the case |
| **External** | Can be offered to a requester by an AI assistant during a conversation  |

External does **not** mean a public library. Requesters do not browse the Knowledge Base; an article reaches them only when an assistant judges it answers what they asked.

## Tags and services

* **Tags** are free text, scoped to your team, and feed search.
* **A service** from the catalog can be attached to an article — an article can belong to several services at once, because a procedure like "answering a customer by e-mail" genuinely belongs to every service that answers customers.

When an analyst links an article to a case that has a service, the article picks up that service automatically, so the catalog fills itself in as the team works.

## Writing with AI assistance

From a case you can ask Clowrk to draft an article from what already happened there — the comments, the chat, the commands that were run. The result is a draft like any other: it is invisible until somebody publishes it, and you can keep talking to refine it before you do.

The point is to write the article while the case is fresh, not to replace the person who knows whether it is right.

## Related

* [Automated Remediation](/clowrk/knowledge-base/automated-remediation.md) — attaching commands to a troubleshooting article
* [Knowledge Base overview](/clowrk/knowledge-base.md)