No. Past ticket bodies feed the model in your own account, with your own API key, and the bot summarizes the resolution rather than quoting another customer's personal details. Email addresses, account IDs and internal agent notes can be excluded from indexing entirely. The retrieval layer also respects user role at query time, so a public visitor never sees private ticket content even if it is in the corpus.

Yes. When the visitor needs an agent, the bot calls Awesome Support's REST endpoints to open a ticket pre-filled with the chat transcript, the product taxonomy term, and any custom fields surfaced in the conversation. The agent picks up a fully contextualized ticket instead of one that says 'help, this is broken'. Confirmation goes back to the visitor in the same chat session.

Yes. Awesome Support custom fields and post meta are exposed to the prompt context, so the bot can answer questions like 'is this fixed in 3.2?' by quoting the version field from a resolved ticket. Custom fields are also passed back through escalation, so when a new ticket gets opened the agent inherits the same structured data the bot already collected from the visitor.

Yes. Display conditions filter the bot by Awesome Support product taxonomy, so the License Manager product runs a separate bot from the Theme Builder product. Each scoped bot only sees tickets tagged with its product, which keeps cross-product noise out of retrieval and lets each product line evolve its own system prompt independently.

Yes. Agent-side bots can summarize long ticket threads on demand and suggest draft replies based on similar past resolutions. The agent reviews and edits before sending, so the bot acts as a research and drafting assistant rather than an autopilot. This is especially useful for new hires who do not yet know the product's quirks.

Yes. Pair with WPML, Polylang or TranslatePress for per-language tickets. The visitor's language is detected and the bot replies in it, even when the source ticket was written in another language. The escalated ticket is created in the visitor's language too, so the agent sees the conversation in context.

Attachments themselves are not embedded into the vector store, but their filenames and any descriptive metadata are. So the bot can say 'past ticket #3201 has a screenshot of the working settings, see the attachment on the ticket' and link the visitor to that ticket. Image and document understanding is on the roadmap, dependent on which model provider you use.

Status changes trigger a re-index on the affected ticket. A ticket that gets reopened loses its resolved-status ranking boost; a ticket that gets resolved gains one. So a question asked yesterday and answered with a then-open ticket will get the now-resolved version's higher confidence answer when asked again today.