What Are Knowledge Snippets?
Snippets are short, structured notes you manage under Knowledge alongside uploaded files. They are ideal for FAQs, policy exceptions, pricing notes, or anything you want the model to see as first-class text without maintaining a separate document.
In the app, open Knowledge, then Snippets (/knowledge/snippets). Create a snippet with a clear title and body, assign tags, and control whether it is included in the next training run.
How Snippets Get Into Training
Training pulls the same kinds of sources for files and snippets: only items that are marked Training Ready and whose tags overlap the agent’s document tags appear in the training preview and are embedded with that agent.
- Training Ready: Treat this as an explicit opt-in so drafts and scratch notes never train by accident.
- Tags: Use the same tag vocabulary you use on files. If a snippet’s tags do not intersect the agent’s document tags, that agent will not include it.
- Retrain after edits: When you change snippet text, tags, or Training Ready, run Train or Retrain on the agent so answers reflect the update (same as for documents).
Practical Tips
- One topic per snippet: Narrow snippets are easier to cite and less likely to dilute retrieval than one giant catch-all note.
- Stable titles: A descriptive title helps your team find the snippet in Knowledge and shows up clearly in training previews.
- Pair with files: Use snippets for deltas (e.g. “As of March 2026, PTO carryover is …”) and keep the authoritative handbook in Files (
/knowledge/files).
Related Guides
For the full onboarding path, see Getting Started. For tagging, previews, and retraining, see AI Agent Training and Document Best Practices.