tusharshah21/PRowl Code Review

Multi-agent AI PR reviewer — a fast model triages, a smarter one explains and fixes. Works with any LLM (BYOK).

Trust Signals

name	description	required	default
GITHUB_TOKEN	GitHub token to interact with the repository.	yes	—
LLM_API_KEY	API key for your chosen LLM provider (OpenAI, Anthropic, Google, etc.). Bring Your Own Key (BYOK).	yes	—
LLM_MODEL	Model name (e.g., gpt-4o, claude-3-5-sonnet-20241022, gemini/gemini-pro, azure/gpt-4).	no	gpt-4o
LLM_BASE_URL	Optional: Custom LiteLLM proxy endpoint URL. Leave empty for direct provider calls.	no	""
LLM_REVIEWER_MODEL	Fast/cheap model for issue detection (falls back to LLM_MODEL if not set).	no	""
LLM_FIXER_MODEL	Smarter model for explanation and fix generation (falls back to LLM_MODEL if not set).	no	""
exclude	Glob patterns to exclude files from the diff analysis	no	""
CONTEXT_LINES	Unchanged lines kept around each +/- in the TOON diff (lower = cheaper, higher = more context). Set 0 to drop all context.	no	2
ENABLE_CACHE	Cache LLM responses by chunk hash across runs in the same job. Set 'false' to disable.	no	true
SEMGREP_RULES	Optional: comma-separated Semgrep rulesets (e.g. 'p/security-audit,p/owasp-top-ten'). If set and 'semgrep' is on PATH, findings are passed to Agent 1 as priors.	no	""
DISCORD_WEBHOOK_URL	Optional: Discord webhook URL for commit/review notifications	no	""
SLACK_BOT_TOKEN	Optional: Slack bot token (xoxb-...) for threaded notifications	no	""
SLACK_CHANNEL_ID	Optional: Slack channel ID used with SLACK_BOT_TOKEN	no	""
SLACK_WEBHOOK_URL	Optional: Slack incoming webhook URL (fallback, non-threaded)	no	""

no outputs