Dashboard

Bug intelligence

Auto-cluster the things users complain about, score by impact, push to your tracker. Engineering stops guessing what's actually broken.

What it does

Every conversation is analyzed and grouped by the underlying issue — not the surface words. "login broken," "can't sign in," and "OAuth redirect loop" all collapse to one cluster if they're about the same root cause. You get a leaderboard of what's actually breaking.

How clusters are built

Each conversation is embedded into a vector and compared to existing clusters. New conversations either join an existing cluster or seed a new one. Clusters split or merge as evidence comes in.

Re-clustering runs continuously — there's no manual button to press. New issues surface within minutes of being reported.

How impact is scored

Each cluster gets a severity score (0–100) computed from:

Frequency — how many users are hitting it
Recency — how recent the reports are (a regression spikes faster)
User tier — Pro / Scale users weight higher than Free
Trajectory — is it growing, flat, or fading?

Engineering teams typically attack the top 5 each sprint and ignore everything below about 30. Tweak the weights from Bug intelligence → Settings if your priorities differ.

Cluster detail view

Open a cluster to see:

Every conversation that landed in it.
The shared root cause as the agent describes it.
Time series of new reports.
Affected users (with their plans, if you're using identify).
The pages and stack contexts most associated with the cluster.

Pushing to your tracker

See Linear / GitHub sync. One click pushes a cluster to your tracker as an issue, with the cluster summary, top 5 user quotes, and a back-link to the DuggAI cluster.

Auto-resolution

When you ship a fix and the cluster stops getting new reports for 7 days, DuggAI marks the cluster auto-resolved and links the originating issue as "closed." New reports after that auto-reopen the cluster — useful for catching regressions.

The right way to use this

Don't fire-fight every cluster. Use bug intelligence as a weekly review: top 5 items, decide what's on the next sprint, ignore the rest. The point is to spend engineering hours on what users actually feel — not what stands out in your inbox that morning.

Privacy

Cluster summaries quote conversation excerpts. If you have user identifiers in those quotes, they'll appear in the cluster view (and in any synced issue). Configure redaction under Bug intelligence → Settings → Redact PII if you need to scrub before exporting.