Team memory & precisionIn testing
It learns what your team means by noise.
Dismiss a finding as “not useful here” and Forseti remembers, for that repo. Team rules live in a file on your base branch. And the Quality page shows how often its findings are actually right.
Quality · last 30 days
Example
Precision trendaccepted ÷ (accepted + incorrect)
LearningsExport .forseti/rules.md
- Style nits in generated/**dismissal · @mayaOn
- Prefer Result<T> over throw in services/.forseti/rules.mdOn
- Missing JSDoc on internal helpersdismissal · @samOff
Noise budget for a 400-line PR6 inline · rest in summary
How it works
Step by step.
It learns what your team dismisses, and shows its hit rate.
- 01Dismiss with a reason“Not useful here” becomes a learning: a category or a specific pattern, suppressed for that repo.
- 02Rules as code.forseti/rules.md on the base branch is read as team rules for every review.
- 03Noise budgetThe number of inline comments scales with PR size. Anything over the budget goes to the summary.
- 04MeasurePrecision = accepted ÷ (accepted + dismissed as incorrect), by repo, category and author, with a 30-day trend.
In detail
What it does, exactly.
- Your precision is shown to you, not hidden behind a marketing number.
- Learnings are plain text you own and can take with you.
- On GitHub
- Dismiss reasons, fewer repeat comments
- In the dashboard
- Quality page: precision trend, learnings, noise budget
Learnings you can see
Every learning is listed with where it came from. Toggle it, delete it, or export the lot.
Export as a file
Download your learnings as a ready .forseti/rules.md, so your team's memory is portable and reviewable.
Configurable budget
Set the noise budget per organisation on the Quality page.
Linked from Insights
Precision sits next to spend, so quality and cost are read together.
Put Forseti on your next pull request.
Connect GitHub in four clicks. Your first review starts by itself on your latest small open PR, priced before it runs.