AI Review

Deepreviewbeta

Angada's deepest review tier for high-impact pull requests, available in private beta.

Deepreview is Angada's deepest review. It runs a wider analysis on Angada's most capable models for the pull requests where a missed bug hurts the most — database migrations, auth, payments, concurrency, and large refactors. A standard review runs on every pull request; Deepreview is for the ones where the blast radius of a missed bug is severe.

private beta — Deepreview is available to workspaces with beta access. Contact Angada support to request it for your workspace.

How it differs from a standard review

A standard review runs on every eligible pull request and posts inline findings with a verdict and walkthrough summary. On the same pull request, Deepreview does more:

Standard reviewDeepreview
Credits13
ModelsDefault review modelsAngada's most capable models
Candidate budgetStandardWider
Blast radius in the summary—Yes

Credits

Angada meters reviews in credits. Deepreview costs more because it runs a wider analysis on more capable models.

ReviewCreditsWhen it runs
Standard1Every eligible pull request
Deepreview3On request, or automatically for high-risk changes

If your workspace is short on credits when a Deepreview starts, the run downgrades to a standard review and notes the downgrade on the pull request. See Limits and caps for review caps and credits.

Reserve Deepreview for high-impact changes — migrations, auth, payments, concurrency, and large refactors — so its extra credits go where a missed bug is most expensive.

Availability

During the private beta, the Angada team enables Deepreview for your workspace and sets how it runs. Two modes are available:

  • Manual — A Deepreview runs when someone requests it on a specific pull request, optionally with a short instruction to focus the run.
  • Automatic — Angada starts a Deepreview on its own for high-risk changes, using a risk gate and file-path rules. Most pull requests stay on the standard review.

Deepreview is not self-serve yet. Contact Angada support to turn it on and choose how it runs for your workspace.

What to expect

  • Deepreview reviews the current pull request head. Pushing new commits does not restart a Deepreview that is already running.
  • An automatic Deepreview notes on the pull request why it started, so the decision is never a mystery.
  • A Deepreview is scoped to one pull request, so a broad or architectural concern may still need follow-up.

Best practices

  • Use it where it counts — Turn on automatic Deepreview for sensitive paths, such as migrations or auth, and request it by hand for other high-risk pull requests.
  • Read the blast radius — Use the impact summary and control-flow diagram to check the change beyond the lines it touches.
  • Watch your credits — Deepreview costs 3 credits per run. See Limits and caps.

For how review results carry across pull request updates, see Re-runs and continuity. For the commands you can use on a pull request, see Interactive comments.

On this page