Understanding the Knowledge Index

What gets indexed

When you click Reindex Now, EnvaTicket reads from five sources and stores semantic vectors in the knowledge_chunks table:

• KB articles — every published Documentation entry.
• FAQs — every FAQ in your database.
• Pages — every published CMS page (About, How it works, Pricing, etc.).
• Product descriptions — the description field on each Product.
• Resolved tickets — closed tickets from the last 365 days. The buyer never sees other buyers' tickets directly, but the AI can learn from the staff resolutions.

Smart, incremental reindex

Every record's content is hashed. On reindex:

• Hash matches an existing chunk → reused. No API call, no cost.
• Hash differs (you edited the content) → re-embedded.
• Record is new → embedded normally.
• Record was deleted → its chunk is pruned.

So clicking Reindex Now after no changes is essentially free, and the result message tells you exactly what happened: "X new chunks, Y unchanged, Z removed."

When to reindex

• After adding or editing KB articles, FAQs, pages, or product descriptions.
• After closing a significant batch of tickets you want the AI to learn from.
• After switching the embedding provider or model — vectors from different models live in different spaces and aren't comparable.