gauthierdmn/Nominal Code Review
AI-powered code review for pull requests using Claude
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| anthropic_api_key | Anthropic API key (required when provider is 'anthropic') | no | "" |
| openai_api_key | OpenAI-compatible API key (required when provider is 'openai', 'deepseek', 'groq', 'together', or 'fireworks') | no | "" |
| google_api_key | Google API key (required when provider is 'google') | no | "" |
| github_token | GitHub token for posting review comments (defaults to github.token) | yes | — |
| provider | LLM provider for the reviewer (anthropic, openai, google, deepseek, groq, together, fireworks). Also the default for the explorer. | no | anthropic |
| image | Docker image to use. The default all-in-one image works with any provider; provider-specific variants are available for smaller footprint. | no | ghcr.io/gauthierdmn/nominal-code:latest |
| model | Reviewer model (e.g. claude-sonnet-4-20250514, gpt-4.1). Empty uses the provider's default. | no | "" |
| max_turns | Maximum agentic turns for the reviewer. Empty uses the bundled default (8), 0 means unlimited. | no | "" |
| explorer_provider | Explorer sub-agent provider. Empty inherits from `provider`. | no | "" |
| explorer_model | Explorer sub-agent model. Empty inherits from `model`. Useful for pairing a strong reviewer with a cheaper explorer. | no | "" |
| explorer_max_turns | Maximum turns for explorer sub-agents. Empty uses the bundled default (32), 0 means unlimited. | no | "" |
| prompt | Custom review instructions appended to the default reviewer prompt. | no | "" |
| coding_guidelines | Inline coding guidelines content. Appended to the reviewer system prompt. | no | "" |
| coding_guidelines_file | Path to a coding guidelines file (relative to repo root). Appended to the reviewer system prompt. Wins over `coding_guidelines` if both are set. | no | "" |
| reviewer_system_prompt | Inline reviewer system prompt that fully REPLACES the bundled prompt. | no | "" |
| reviewer_system_prompt_file | Path to a file whose contents fully REPLACE the bundled reviewer system prompt. | no | "" |
| explorer_system_prompt | Inline explorer sub-agent system prompt that fully REPLACES the bundled prompt. | no | "" |
| explorer_system_prompt_file | Path to a file whose contents fully REPLACE the bundled explorer system prompt. | no | "" |
| language_guidelines_dir | Directory containing per-language guideline files (e.g. python.md). Overrides the bundled language guidelines. | no | "" |
| inline_suggestions | Whether to render findings as inline code suggestions. Defaults to true; set 'false' to disable. | no | "" |
| ignore_existing_comments | When 'true', skip fetching prior PR comments so they don't bias the review (useful for re-runs). | no | "" |
| dry_run | When 'true', run the full review pipeline but skip posting comments to the PR. | no | "" |
Outputs
no outputs