Insights dashboard
See what users ask, which questions got no answer, and which pages are cited most.
Why it matters: Turns the assistant into a report on where your docs fall short.
What we plan to build, in the order we expect to build it. There are no dates, and the order can change.
Planned for the coming releases.
See what users ask, which questions got no answer, and which pages are cited most.
Why it matters: Turns the assistant into a report on where your docs fall short.
A thumbs up or down on each answer, with an optional reason.
Why it matters: Your team finds weak answers and fixes the page behind them.
When the docs do not cover a question, offer a contact step that sends the conversation to email or your helpdesk.
Why it matters: No user ends at a dead end, and your team sees the full context.
Planned after the items above.
A monthly estimate of tickets avoided, based on answered conversations that did not continue to support.
Why it matters: A number to bring to the budget review.
Let the assistant read fields you approve, such as a user's plan or enabled features, to tailor answers.
Why it matters: More useful answers, while the assistant still cannot change anything.
Create tickets with the conversation attached in the helpdesk tools your team already uses.
Why it matters: Handoff without copy and paste.
Under consideration.
Re-fetch help-centre pages on a schedule.
Why it matters: Answers follow your docs as they change.
In the open-source release today.
Each answer lists the pages it came from, linked to your own URLs.
Questions your docs do not cover are declined instead of guessed.
Threads bound to the signed-in user, with history the user can return to.
A record of each turn for debugging and compliance.
Try questions against your docs, with the passages shown, before the agent goes live.
Run the server on Node, or at the edge on Workers with Workers AI.
Missing something your team needs? Open an issue on GitHub.
Open an issue