Configuration¶

All settings are configured via environment variables.

Migration: AI_REVIEWER_ prefix

Since v1.0.0a7, all environment variables support the AI_REVIEWER_ prefix (e.g., AI_REVIEWER_GOOGLE_API_KEY). Old names (e.g., GOOGLE_API_KEY) still work as fallback. We recommend migrating to the new names to avoid conflicts with other tools in org-level CI/CD configurations.

Required Variables¶

Variable	Description	Example	How to get
`AI_REVIEWER_GOOGLE_API_KEY`	Google Gemini API key (comma-separated for multi-key rotation)	`AIza...`	Google AI Studio
`AI_REVIEWER_MISTRAL_API_KEY`	Mistral API key (comma-separated for multi-key rotation)	`sk-...`	Mistral Console
`AI_REVIEWER_GITHUB_TOKEN`	GitHub token (for GitHub)	`ghp_...`	Instructions
`AI_REVIEWER_GITLAB_TOKEN`	GitLab token (for GitLab)	`glpat-...`	Instructions

At least one LLM API key required

You need at least one LLM API key: AI_REVIEWER_GOOGLE_API_KEY or AI_REVIEWER_MISTRAL_API_KEY (or both). The key for the primary provider (set by AI_REVIEWER_LLM_PROVIDER) is required.

At least one platform token required

You need AI_REVIEWER_GITHUB_TOKEN or AI_REVIEWER_GITLAB_TOKEN depending on the platform.

GitLab token types

For GitLab, you can use a Personal Access Token (works on all plans, including Free) or a Project Access Token (requires GitLab Premium/Ultimate).

Optional Variables¶

General¶

Variable	Description	Default	Range
`AI_REVIEWER_LOG_LEVEL`	Logging level	`INFO`	DEBUG, INFO, WARNING, ERROR, CRITICAL
`AI_REVIEWER_API_TIMEOUT`	Request timeout (sec)	`60`	1-300

Language¶

Variable	Description	Default	Examples
`AI_REVIEWER_LANGUAGE`	Response language	`en`	`uk`, `de`, `es`, `it`, `me`
`AI_REVIEWER_LANGUAGE_MODE`	Detection mode	`adaptive`	`adaptive`, `fixed`

Language modes:

adaptive (default) — automatically detects language from PR/MR context (description, comments, linked task)
fixed — always uses the language from AI_REVIEWER_LANGUAGE

ISO 639

AI_REVIEWER_LANGUAGE accepts any valid ISO 639 code:

2-letter: en, uk, de, es, it
3-letter: ukr, deu, spa
Names: English, Ukrainian, German

LLM¶

Provider Selection¶

Variable	Description	Default
`AI_REVIEWER_LLM_PROVIDER`	Primary LLM provider	`google`
`AI_REVIEWER_LLM_FALLBACK_PROVIDER`	Fallback provider (used when primary is exhausted)	(none)

Supported providers: google, mistral.

When both primary and fallback providers are configured, the system tries all primary provider models first, then all fallback provider models. Example with LLM_PROVIDER=mistral and LLM_FALLBACK_PROVIDER=google:

mistral-large → mistral-small → gemini-2.5-flash → gemini-2.5-flash-lite

Google Gemini Models¶

Variable	Description	Default
`AI_REVIEWER_GEMINI_MODEL`	Primary Gemini model	`gemini-2.5-flash`
`AI_REVIEWER_GEMINI_MODEL_FALLBACK`	Fallback model chain (comma-separated)	`gemini-3-flash-preview,...`

Model	Description	Cost
`gemini-2.5-flash`	Fast, stable, reasoning (default)	$0.075 / 1M input
`gemini-3-flash-preview`	Frontier-class flash (preview)	$0.075 / 1M input
`gemini-2.5-flash-lite`	Fastest and cheapest in 2.5	$0.01875 / 1M input
`gemini-2.5-pro`	Most powerful, deep reasoning	$1.25 / 1M input

Gemini API Pricing

Mistral Models¶

Variable	Description	Default
`AI_REVIEWER_MISTRAL_MODEL`	Primary Mistral model	`mistral-large-latest`
`AI_REVIEWER_MISTRAL_MODEL_FALLBACK`	Fallback model chain (comma-separated)	(none)
`AI_REVIEWER_MISTRAL_API_URL`	Custom API base URL	(none)

Model	Description	Context	Cost (input / output)
`mistral-large-latest`	Most capable, MoE 41B/675B (default)	256k	$0.50 / $1.50
`mistral-medium-latest`	Balanced performance/cost	128k	$0.40 / $2.00
`mistral-small-latest`	Fast and cheap, hybrid 6.5B/119B	256k	$0.15 / $0.60
`codestral-latest`	Optimized for code, FIM support	128k	$0.30 / $0.90
`devstral-latest`	Coding agent, cheapest Mistral	256k	$0.10 / $0.30

Codestral free tier

Codestral has a separate free tier at https://codestral.mistral.ai with its own API key. Set AI_REVIEWER_MISTRAL_API_URL=https://codestral.mistral.ai and use a key from codestral.mistral.ai. The paid Codestral via api.mistral.ai does not require a custom URL.

Mistral Pricing

Other Settings¶

Variable	Description	Default
`AI_REVIEWER_REVIEW_SPLIT_THRESHOLD`	Char threshold for code+test split review	`30000`

Pricing accuracy

Prices are listed as of the release date and may change. Check the provider's pricing page for current information.

Free Tier

Both Google Gemini and Mistral offer free tiers sufficient for code review of a team of 4-8 developers. Combine two providers for maximum reliability.

Review¶

Variable	Description	Default	Range
`AI_REVIEWER_REVIEW_MAX_FILES`	Max files in context	`20`	1-100
`AI_REVIEWER_REVIEW_MAX_DIFF_LINES`	Max diff lines per file	`500`	1-5000
`AI_REVIEWER_REVIEW_MAX_COMMENT_CHARS`	Max MR comment chars in AI prompt	`3000`	0-20000
`AI_REVIEWER_REVIEW_INCLUDE_BOT_COMMENTS`	Include bot comments in prompt	`true`	true/false
`AI_REVIEWER_REVIEW_POST_INLINE_COMMENTS`	Post inline comments on lines	`true`	true/false
`AI_REVIEWER_REVIEW_ENABLE_DIALOGUE`	Group comments into dialogue threads	`true`	true/false

Discussion context

The AI reviewer reads existing MR/PR comments to avoid repeating suggestions that were already discussed. Set AI_REVIEWER_REVIEW_MAX_COMMENT_CHARS=0 to disable.

Inline comments

When AI_REVIEWER_REVIEW_POST_INLINE_COMMENTS=true (default), issues with file/line info are posted as inline comments on the code, with a short summary as the review body. Set to false for a single summary comment.

Dialogue threads

When AI_REVIEWER_REVIEW_ENABLE_DIALOGUE=true (default), comments are grouped into conversation threads so the AI understands reply chains. Set to false for flat rendering.

Discovery¶

Variable	Description	Default	Range
`AI_REVIEWER_DISCOVERY_ENABLED`	Enable project discovery before review	`true`	true/false
`AI_REVIEWER_DISCOVERY_VERBOSE`	Always post discovery comment (default: only on gaps)	`false`	true/false
`AI_REVIEWER_DISCOVERY_TIMEOUT`	Discovery pipeline timeout in seconds	`30`	1-300

Project Discovery

When enabled, AI ReviewBot automatically analyzes your repository (languages, CI pipeline, config files) before each review to provide smarter feedback. Set to false to disable. See Discovery → for details.

Verbose mode

When AI_REVIEWER_DISCOVERY_VERBOSE=true, the discovery comment is always posted and includes all Attention Zones (well-covered, weakly-covered, not-covered). Default mode only posts when there are gaps or uncovered zones.

GitLab¶

Variable	Description	Default
`AI_REVIEWER_GITLAB_URL`	GitLab server URL	`https://gitlab.com`

Self-hosted GitLab

For self-hosted GitLab, set AI_REVIEWER_GITLAB_URL:

export AI_REVIEWER_GITLAB_URL=https://gitlab.mycompany.com

.env File¶

It's convenient to store configuration in .env:

# .env

# LLM provider (pick one or both)
AI_REVIEWER_GOOGLE_API_KEY=AIza...
# AI_REVIEWER_MISTRAL_API_KEY=sk-...
# AI_REVIEWER_LLM_PROVIDER=google
# AI_REVIEWER_LLM_FALLBACK_PROVIDER=mistral

# Platform token
AI_REVIEWER_GITHUB_TOKEN=ghp_...

# Optional
AI_REVIEWER_LANGUAGE=uk
AI_REVIEWER_LANGUAGE_MODE=adaptive
AI_REVIEWER_GEMINI_MODEL=gemini-2.5-flash
AI_REVIEWER_LOG_LEVEL=INFO

Security

Never commit .env to git!

Add to .gitignore:

.env
.env.*

CI/CD Configuration¶

GitHub Actions¶

env:
  AI_REVIEWER_GOOGLE_API_KEY: ${{ secrets.AI_REVIEWER_GOOGLE_API_KEY }}
  AI_REVIEWER_GITHUB_TOKEN: ${{ github.token }}  # Automatic
  AI_REVIEWER_LANGUAGE: uk
  AI_REVIEWER_LANGUAGE_MODE: adaptive

GitLab CI¶

variables:
  # AI_REVIEWER_GOOGLE_API_KEY and AI_REVIEWER_GITLAB_TOKEN
  # are inherited from CI/CD Variables automatically
  AI_REVIEWER_LANGUAGE: uk
  AI_REVIEWER_LANGUAGE_MODE: adaptive

Validation¶

AI Code Reviewer validates configuration at startup:

Validation Errors¶

ValidationError: AI_REVIEWER_GOOGLE_API_KEY is too short (minimum 10 characters)

Solution: Check that the variable is set correctly.

ValidationError: Invalid language code 'xyz'

Solution: Use a valid ISO 639 code.

ValidationError: LOG_LEVEL must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL

Solution: Use one of the allowed levels.

Configuration Examples¶

Minimal (GitHub)¶

export AI_REVIEWER_GOOGLE_API_KEY=AIza...
export AI_REVIEWER_GITHUB_TOKEN=ghp_...

Minimal (GitLab)¶

export AI_REVIEWER_GOOGLE_API_KEY=AIza...
export AI_REVIEWER_GITLAB_TOKEN=glpat-...

Ukrainian language, fixed¶

export AI_REVIEWER_GOOGLE_API_KEY=AIza...
export AI_REVIEWER_GITHUB_TOKEN=ghp_...
export AI_REVIEWER_LANGUAGE=uk
export AI_REVIEWER_LANGUAGE_MODE=fixed

Self-hosted GitLab¶

export AI_REVIEWER_GOOGLE_API_KEY=AIza...
export AI_REVIEWER_GITLAB_TOKEN=glpat-...
export AI_REVIEWER_GITLAB_URL=https://gitlab.mycompany.com

Debug mode¶

export AI_REVIEWER_GOOGLE_API_KEY=AIza...
export AI_REVIEWER_GITHUB_TOKEN=ghp_...
export AI_REVIEWER_LOG_LEVEL=DEBUG

Configuration Priority¶

Environment variables (highest)
.env file in the current directory