Settings Reference¶

Overview¶

This page lists every CloudAEye setting, grouped as in the console, and says whether it is set in the console settings, in .cloudaeye.config, or both. Each console page is described in the Settings UI section on the left. The Configuration File page explains the file's format and which file applies to which change.

In the tables, Line is the setting as you write it in .cloudaeye.config, under the section and group shown above the table. A setting you leave out takes its Default.

LLM¶

Console settings: Settings > LLM, and Settings > LLM > Bring your own LLM. See LLM in the Settings UI section.

File: ## Review > ### LLM, read from the repository's top-level .cloudaeye.config only. Used by /cloudaeye:inspect, /cloudaeye:security, /cloudaeye:review and ultrareview's review. Other commands use the models set in the console settings.

Line	Options	Default
`- LLM:`	`CloudAEye` or `your own`	The console setting
`- Provider:`	With `your own` only: `openai`, `anthropic`, `azure`, `aws_bedrock`, `gcp_vertex`, `groq`, `deepseek`	The console setting
`- Model:`	With `CloudAEye`, a model the console offers. With `your own`, the provider's exact model name	The console setting
`- Model inversion:`	`true` or `false`. When AI authorship is detected, a model from a different family reviews the code. A `Model` or `Provider` set here wins over it	The console setting
`- API key:`	With `your own` and Provider `openai`, `anthropic` or `deepseek`: an environment variable that holds the key, like `$OPENAI_API_KEY`, or the key itself	The key saved in the console settings

## Review

### LLM
- LLM: your own                      <!-- CloudAEye | your own -->
- Provider: openai                   <!-- your own only -->
- Model: gpt-6.1-terra
- Model inversion: true              <!-- true | false -->
- API key: $OPENAI_API_KEY           <!-- an environment variable that holds the key, or the key -->

Prefer an environment variable for the API key. $OPENAI_API_KEY (or ${OPENAI_API_KEY}) is read on the machine running the command, so the key stays out of the repository. A key written in the file can be read by anyone who can read the repository.
The key goes to CloudAEye only. The commands send it straight to CloudAEye and never show it in your coding agent's output; CloudAEye keeps it encrypted only for the reviews it was sent for. A key in the file wins over the one saved in the console settings.
Name the model beside the key. With your own key, Model is the model the review runs on.
Azure, AWS Bedrock, GCP Vertex and Groq need more than one key, so their credentials stay in the console settings, under Settings > LLM > Bring your own LLM. An API key line with one of them is left out, and the command says so.
When the variable is not set, or the key cannot be sent, the review uses the key saved in the console settings, and the command says so.
A model option you pass to a command wins over the file for that run.
See Model Inversion for how inversion works.

Ultrareview¶

File: ## Ultrareview, used by /cloudaeye:ultrareview. The console settings hold the same settings, and a new file can start from them.

Checks¶

## Ultrareview > ### Checks. A check set to false is not run, and its row in the report reads turned off in the repository's settings. With Bugs, Security, AI slop score or Risk score set to false, no change is approved automatically.

Line	Options	Default
`- PR checklist:`	`true` or `false`. Pull requests only; the checklist items are set in the console settings	`true`
`- Bugs:`	`true` or `false`	`true`
`- Security:`	`true` or `false`. Includes leaked secrets	`true`
`- AI slop score:`	`true` or `false`	`true`
`- Risk score:`	`true` or `false`	`true`
`- Architecture:`	`true` or `false`	`true`
`- Control flow:`	`true` or `false`	`true`
`- Task check:`	`true` or `false`. Runs when a task is given	`true`

Bugs and Security are the two halves of one review. With one of them off, the review runs the other half alone.

Limits¶

## Ultrareview > ### Limits. Hard limits: a change above any of them is never approved automatically, and the report says which limit it crossed.

Line	Options	Default
`- Risk score:`	`0` to `100`, written like `40/100`	CloudAEye's own check
`- AI slop score:`	`0` to `100`, written like `50/100`	CloudAEye's own check
`- Review severity:`	`medium` or `low`: the worst finding a change may have. With `medium`, a High finding needs a human review; with `low`, a Medium one does too	CloudAEye's own check

A leaked secret, a Critical finding or an open bug in a changed file still needs a human review, whatever the limits.

Human review policy¶

## Ultrareview > ### Human review policy. Pull requests only. When CloudAEye does not approve a change automatically, it names the people who should review it in its report on the pull request.

Line	Options	Default
`- Auto-assign human reviewer:`	`true` or `false`. With `false`, nobody is named, and a human must still review the change	`true`
`- Max human reviewers to assign:`	`1`, `2` or `3`. One is named unless the instructions ask for more, and never more than this. An instruction asking for two people needs `2` or `3` here	`1`
`- Instructions:`	A paragraph in plain words, up to 2,000 characters, on indented lines under the label. It can ask for more review, never less	None

### Human review policy
- Auto-assign human reviewer: true   <!-- true | false. Default true -->
- Max human reviewers to assign: 2   <!-- 1 | 2 | 3. Default 1 -->
- Instructions: Changes under `services/payments/` always need two humans. Pull requests
  from the Platform team need one reviewer from outside the team.

CloudAEye names reviewers in its report. It does not request their review on GitHub. See Rules and Instructions for writing instructions.

Code Review¶

Console settings: Settings > Code Review. No Secrets in Code is set in the console only. See Code Review in the Settings UI section. The coding-agent commands look for leaked secrets in /cloudaeye:security, /cloudaeye:review, and ultrareview's Security check.

File: ## Review, used by /cloudaeye:inspect, /cloudaeye:security and /cloudaeye:review.

Severity floor¶

## Review > ### Severity floor. Report only findings at or above this level. A level you pass to the command wins for that run. Ultrareview ignores this setting.

Line	Options	Default
`- Severity floor:`	`all`, `critical`, `high`, `medium` or `low`	`all`

Paths never reviewed¶

## Review > ### Paths never reviewed. Before a commit only. One path or pattern a line, relative to the directory of the file it is in. **/ matches any number of directories, including none.

Line	Options	Default
`- Skip:`	A path or pattern in backticks	None

### Paths never reviewed
- Skip: `**/generated/**`
- Skip: `vendor/`

Review rules¶

Review rules are kept in .cloudaeye.rules, beside .cloudaeye.config. See Rules and Instructions.

Linter¶

Console settings only. Linters are enabled and configured for each repository under Integrations > Repos > Linters. See Linters.

PR Checklist¶

Console settings only. The checklist items, and whether each one is a Warning or an Error, are set under Settings > PR Checklist and for each repository. See PR Checklist.

In the file, - PR checklist: under Checks only decides whether ultrareview runs the checklist.

Unit Tests¶

Console settings only. The test framework for each language, for your account or one repository, and automatic test updates on code change. See Unit Tests in the Settings UI section.

API Keys¶

Console settings only. These are keys for calling CloudAEye's own API, created and deleted under Settings > API Keys. See API Keys in the Settings UI section. Your LLM provider's key is a different key; it can go in the file, under LLM.

Data Retention¶

Console settings only. How long review history is kept, 5 to 30 days, 15 by default. See Data Retention in the Settings UI section.

Control Flow¶

File only: ## Control flow, used by /cloudaeye:control-flow and ultrareview.

Line	Options	Default
`- Effort:`	`low`, `medium` or `high`: how far the map follows calls	`medium`
`- Across repositories:`	`true` or `false`: whether it looks for callers in your other connected repositories	`true`

Triage PRs¶

File only: ## Triage PRs, read from the repository's top-level .cloudaeye.config only. Used by /cloudaeye:triage-prs.

Line	Options	Default
`- OKRs:`	Your team's key results in plain words, or a file that holds them, in backticks	None

A Complete File¶

## Ultrareview

### Checks
- PR checklist: true                 <!-- true | false. Pull requests only -->
- Bugs: true                         <!-- true | false -->
- Security: true                     <!-- true | false. Includes leaked secrets -->
- AI slop score: true                <!-- true | false -->
- Risk score: true                   <!-- true | false -->
- Architecture: true                 <!-- true | false -->
- Control flow: false                <!-- true | false -->
- Task check: true                   <!-- true | false. Runs when a task is given -->

### Limits
- Risk score: 40/100                 <!-- 0-100 -->
- AI slop score: 50/100              <!-- 0-100 -->
- Review severity: medium            <!-- medium | low. The worst finding a change may have -->

### Human review policy
- Auto-assign human reviewer: true   <!-- true | false. Default true -->
- Max human reviewers to assign: 2   <!-- 1 | 2 | 3. Default 1 -->
- Instructions: Changes under `services/payments/` always need two humans. Pull requests
  from the Platform team need one reviewer from outside the team.

## Review

### Severity floor
- Severity floor: medium             <!-- all | critical | high | medium | low -->

### LLM
- LLM: your own                      <!-- CloudAEye | your own -->
- Provider: openai                   <!-- your own only -->
- Model: gpt-6.1-terra
- Model inversion: true              <!-- true | false -->

### Paths never reviewed
- Skip: `**/generated/**`

## Control flow
- Effort: medium                     <!-- low | medium | high -->
- Across repositories: true          <!-- true | false -->

## Triage PRs
- OKRs: `docs/okrs-q4.md`