Bots settings

The Bots tab decides which accounts Limn treats as bots, what kind of bot each one is, and what it costs. It also sets up bot comment scoring. These choices feed Bots Monitoring, the Feed and the Timeline. Bot comment scoring applies to the whole computer; the classification applies to the current workspace only.

#Bot comment scoring

A small model reads each review comment a bot leaves and labels how serious it is, from nit to critical. It runs on your computer: nothing is sent anywhere to score a comment. Setting it up downloads the model and the program that runs it, about 240 MB, once.

• Select Set up (one-time ~240 MB download), or run limn ml install in a terminal.
• When it is ready the section reads Set up. Bot review comments are scored on this machine.

Scoring is kept gentle on your computer: one comment at a time, at low priority, and it pauses on battery. See Severity scoring.

#Bot classification

Bot classification lists every account that has reviewed or commented in this workspace: who counts as a bot, what kind of bot it is, and what it costs here. Bots are found automatically. Bots are hidden on the Feed and the Timeline by default. A change here reaches every repository in the workspace; the repository chips on each card show where that bot is active.

Bot or person

Select Bot or Person to correct Limn’s guess for this workspace. Your choice wins. Reset classification forgets it and lets detection decide again. Your other workspaces are not affected.

Role

Each bot has one role in the workspace. The role decides which figures count it:

• Review bot: An AI code reviewer. The only role counted in the review-bot figures: ROI, behaviour, duplicates and the benchmark.
• Quality check: Static analysis, coverage, scanners, CI. Posts pass or fail verdicts rather than findings. Shown in the Feed, left out of the review-bot figures.
• Dependency updates: Version bumps (Dependabot, Renovate). Opens pull requests and never reviews; its merges are reported as overhead, not team output.
• Code agent: Writes code that is not a version bump: coding agents, autofix, generated-content sync. Its merged pull requests are reported as work no person typed.
• Release automation: Merge queues, release trains, changelogs, backports. Moves code without writing or judging it; its approvals do not count as review.
• Housekeeping: CLA and DCO checks, triage, labels, stale-closers, size and preview reports. Left out of every review figure.

Name

You can pick the vendor and give the bot a label for this workspace. It does not change whether it counts as a bot.

Price for this Workspace

Type a monthly price in US dollars, choose flat or per seat, and select Save price. Type 0 for a free bot. A seat is a person who opened a pull request in this workspace’s repositories in the last 30 days, so a per-seat monthly figure moves with the team. Clear the box to remove the price.