Skip to content

Asking Nodrik a follow-up

Nodrik posts a card and a thread. Mention Nodrik in that thread to get a reply.

@Nodrik did the connection count spike before or after the deploy?

The reply starts from the investigation that already ran — its transcript, every tool result gathered — and adds a short, bounded attempt at your question. It is the conversation about that incident, not a second investigation.

A thread under a card is where you and your colleagues work the problem out between yourselves, and a bot that answers every reply is a bot in the way. Nodrik reads nothing until it is mentioned by name.

That is not just etiquette, it is the permission the integration holds. Mentioning needs app_mentions:read, which delivers only the messages that name Nodrik. Reading a thread unprompted would need permission to read every message in your channel, which is not something the product should be asking for.

Same read-only tools as the investigation itself — logs, metrics, error groups, revisions, recent changes, your connected repositories — against the same project.

Your own connected tools are deliberately out of reach. An investigation may call them: it was started by your own alert, in your own project. A follow-up is started by anyone who can type in a Slack channel, and Slack cannot tell us who that is — a guest, a Slack Connect member from another company, and a read-only console user all post the same way. We do not inspect what your tools do, so we do not let that surface reach them. See your own tools.

One incident is one investigation, and the conversation about it is free. Charging for questions would teach people not to ask at exactly the moment asking is most valuable.

There is a cap of fifteen questions per investigation, which is there so a thread cannot run away rather than to ration you. Past that, re-run the investigation for a full pass.

The person reading the thread at 3am is frequently not the person who manages the Nodrik account, and requiring a console login would make the feature useless in the situation it exists for.

Who asked is recorded on the investigation, so the history says so. It grants nothing: Slack carries no notion of a Nodrik role.

Read a credential. If your question contains something that looks like a live key, token or private key, Nodrik says so and stops — without reading it — and names what to rotate, because by then it is already in the thread, in Slack’s search and in any export of that workspace. Stripping it quietly would clean our logs and leave your key exposed.

Read a full pasted log file. There is a length limit. Trim it to the part that matters.

Ordinary personal data — an email address, an IP — is removed on the way in and the question is answered normally. Being made to rewrite a question mid-incident would teach people not to ask.

  • Check Nodrik is in the thread’s channel. It replies only in threads under its own cards.
  • Check Settings. If Slack shows Reconnect to finish, this workspace installed Nodrik before it could receive mentions. Slack only grants that on a fresh install, so reconnecting is the fix; everything else keeps working in the meantime.
  • Check the investigation has finished. A question asked while the investigation is still running gets told to wait, rather than answered from half a transcript.