Review tiers (how much review a change gets)
When an agent asks for an AI code review it says how much it wants — light, standard or heavy — and Omniscio runs the reviewers you configured for that size. A change touching a database migration, signing in or money, or one simply too large to read closely, is always reviewed at the heavy line-up whatever was asked for. Every review records what was asked, what actually ran, and how big the change was.
On this page
What it is
When a session asks Omniscio for an AI code review, it says how much review the change needs: light, standard or heavy. Each of those names a reviewer line-up you configure, and Omniscio runs that line-up — except at a review step that names its own reviewers, which keeps them.
Before this, every review ran whoever the pipeline step happened to name, whether the change was a one-line log fix or a database migration. You either paid for a full review of a stylesheet edit or got one cheap reader on something that could drop a table.
Two things make the choice safe:
- A floor. Omniscio reads the change itself. Anything touching a database migration, signing in or accounts, or money and spend — or simply too large to read closely — is always reviewed at the heavy line-up, no matter what was asked for. The floor only ever raises a request, never lowers it.
- It is recorded. Each review keeps what was asked for, the one-line reason the asking agent gave, what actually ran, and what the change measured. So you can look back and ask "was light the right call for that?" and have the facts to answer it.
Where to find it
Dev Pipeline → Setup → AI code review. There is a section with the three sizes, and each one is a list of AIs:
- Light — one quick reviewer once you list one. For small, low-risk changes.
- Standard — two reviewers once you list them. The size an agent gets when it does not say which one it wants.
- Heavy — the full line-up once you list one. Big or risky changes.
These lists choose for the review steps that name nobody. A review step that lists its own reviewers — Dev Pipeline keeps that list in its phase manifest — is the line-up at that point in the run: those AIs review, whatever size the agent asks for.
Left blank, a tier does not run the size its title advertises. None of the three ships filled in: a tier that names nobody runs the line-up a review would have run without tiers at all — the pipeline step's own reviewers.
In each list you can name a speed level (basic, worker, smart, genius — those run a
Claude model inside the session) or another company (codex, gemini, deepseek, glm,
kimi), which sends the work to them and costs money.
Below each list is how many reviewers run — leave it at 0 to run one per name you listed — and an optional trial reviewer:
- A trial AI runs beside your usual reviewers for that size — never instead of one — so you can compare how it does on the same work.
- You choose what share of that size's reviews it joins (a percentage), so trying something new costs a fraction of the traffic.
Changing any of this applies to the next review. No restart, no code change.
How it behaves
When a reviewer cannot start
A reviewer the app cannot start is not simply lost, and knowing which case you are in tells you what the asker was told:
- The reviewer queue was full (ten reviewer launches already queued or starting). The round waits up to three minutes for a slot and retries on its own, so a busy moment costs a short wait rather than a reviewer. Past that window the round settles and says what happened.
- The seat was on a cloud machine that never came up. The app runs that same reviewer once on this computer instead, and the seat says it stayed home and why.
The card in Dev Pipeline → Setup → AI code review always shows the last figures it actually read. If the database is momentarily too busy to re-read, it shows those earlier figures rather than an error, and it re-reads on the next open.
What the asking agent sees
An agent asks for a review with a size and a one-line reason. It cannot pick or swap an individual
reviewer — the line-up is always yours. At a step that lists its own reviewers, those are who
review, whatever size the agent names. If it asks for light on something that turns out to be a
migration, the app raises it to heavy and the heavy line-up reviews instead, and the record says
which rule raised it (and names any of the step's reviewers that raise left out).
An agent that names no size at all gets the reviewers configured on the pipeline step it is at, exactly as you set them up — a missing size never swaps your step's line-up for a smaller one. That is why the skill still tells an agent to name a size even at such a step: it is the record of what was asked, and naming the real size is what keeps a risky change from being raised.
A request that names its own reviewers is IGNORED: a plan or code review always runs your configured
line-up, and the 202 carries a note saying the list was dropped and which size picked instead.
(Only the app's own crew mission round keeps a list of its own — it is the one caller that names
reviewers.)
If the line-up that actually runs leaves out a reviewer you configured — the size floor raised the
review past your step's list, or a count is shorter than your list — the 202 says so: unseatedReviewers names each
one with its reason, and the note repeats it in a sentence. A round can run fewer reviewers than
you configured, but it can never again do so silently.
Seeing what was chosen
- The scorecard (in the same Setup area) reports each AI's record split by size — the same AI at light and at heavy are different things to compare, and averaging them would answer neither question.
- From the command line, the local control server reports, for the session asking: how many reviews it started, how many named a size, how often the floor raised one, the counts by size asked for and by size actually run, which floor rule fired, and the largest change measured.
Where it goes wrong
- A size is not permission to skip review. Light is still a real review. Anything that lets a change through unreviewed is the exact failure this feature exists to prevent — so every size always names real reviewers, and a list left blank falls back to the step's line-up rather than to nobody or to fewer.
- A trial reviewer is never a replacement. If turning one on quietly dropped a reviewer, a comparison built from two different line-ups would look identical to a real one and be worthless.
Related
- dev-pipeline.md — the pipeline these tiers are a step of.
- review-session-titles.md — how a review session is named once it is running.
Last verified 2026-11-15