Train your knowledge
How the Content Library cleans and rewrites your content
The Content Library is the structured knowledge base that emerges after Living Knowledge processes your sources. It is not a simple copy of your documents; it is the Engine’s clean, rewritten version of everything the agent needs to know. This article explains how that transformation happens, what you can do with the result, and how the library keeps itself current.
From raw sources to structured units
The process starts with your raw sources. These are often messy: different writing styles, repeated information, contradicting answers, and content scattered across systems. The Engine reads every source, breaks it into individual ideas, deduplicates across sources, and rewrites everything into one consistent voice. The result is a set of structured content units, where each unit covers one idea, written in a consistent style and stripped of duplicates.
This is the key difference between a simple vector index over raw documents and a maintained library. Ambiguity gets removed at processing time, not at retrieval time. When the agent answers a question, it works from this clean, non-ambiguous library rather than from the raw source material.
Continuous restructuring and source traceability
The library is not a static snapshot. It rebuilds itself whenever your sources change. Add a new source, and the library updates. Update an existing source, and the library updates again. There is no schedule to manage and no manual maintenance step.
Every content unit in the library traces back to the source it came from. When the agent answers a question, you can follow the chain from the answer back to the original content. This source traceability is essential for auditing and for fixing wrong answers.
Conflicts, gaps, and the Inbox
The Engine actively flags two types of problems. First, it identifies conflicting content across sources. Second, it spots topics where Living Knowledge is thin. Both appear in the Inbox as Messages waiting for human judgment. You open the conflict, decide which source is correct, archive or update the other, and the library reprocesses automatically.
What you can do in the library
The Content library gives you a set of practical tools:
- Review the structured content units the Engine has derived from your sources
- Trace a content unit back to its original source
- See which units the agent references most often
- Inspect flagged conflicts and gaps
- Approve, edit, or discard suggestions from Team Feedback
You cannot edit content directly in the library. Edits happen at the source level. The library is derived, not authored. If a content unit is wrong, open it to see its source chain. Usually one of the upstream sources is wrong, outdated, or contradicts another source. Fix the source and the library updates.
When to use the library
Use the Content library when you want to understand what the agent actually knows, not just what is in your sources. It is also the right place when the agent gave a wrong answer and you want to find the offending content unit, or when you are auditing Living Knowledge for an internal review.
Do not use it to add new content. Add new content as a source, and the library will pick it up. Do not use it to publish customer-facing articles either; that is what the Help Center is for.
The Read more button
When the AI answers a question in a conversation, a Read more button appears if your workspace has the content library enabled. Clicking this button opens a full article view on the subject, allowing users to explore the topic in more detail.
Recent changes: automatic updates
A recent release removed the manual accept step from content library updates. Changes from Living Knowledge now land in your content library automatically by default. A review task appears only when a change conflicts with existing content or when confidence is low. A full activity log lets you review recent changes and roll one back whenever you need to.
Conclusion
The Content library is the engine’s clean version of your raw sources, continuously restructured and rewritten into one non-ambiguous library. It removes ambiguity at processing time, keeps itself current as sources change, and gives you full traceability back to the original content. When something goes wrong, you fix the source, not the library.