michaelfu1998-create/PR Code Review

LLM‑powered PR reviews with static analysis, multi‑language support, personas, and learning‑oriented scoring.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit May 2, 2026
License
MIT

Pinned Snippet

workflow.ymlSHA-pinned
uses: michaelfu1998-create/pr-review-assistant@ead2ee5bc2fcc9f5ebbd70e317639d7c1d7aceec # v1.1.2

tags can be moved; commit SHAs can't. why a SHA?

namedescriptionrequireddefault
openai_api_keyOpenAI API Key (or Anthropic key if using llm_provider: anthropic)yes
github_tokenGitHub API Token with PR read/write permissionsyes
github_pr_idGitHub PR number to reviewyes
openai_modelLLM model name (e.g., gpt-5.4-mini-2026-03-17, gpt-5-nano, claude-sonnet-4-6)nogpt-5.4-mini-2026-03-17
openai_temperatureSampling temperature [0, 1]. Higher values = more creative reviews.no1
openai_max_tokensMaximum tokens in the LLM response.no32000
llm_providerLLM provider: openai, anthropic, or openai (with api_base_url for compatible APIs)noopenai
api_base_urlCustom API base URL for OpenAI-compatible APIs (Ollama, vLLM, Azure, etc.)no""
anthropic_api_keyAnthropic API Key (required when llm_provider is 'anthropic')no""
filesComma-separated glob patterns for files to review (e.g., '*.py,src/*.ts')no*
max_filesMaximum number of files to review per PRno10
toolsTools to run: 'auto' (detect from tech stack), 'none' (LLM only), or comma-separated list (e.g., 'semgrep,ruff,eslint')noauto
severity_thresholdMinimum severity to report from tools: critical, high, medium, low, infonolow
review_focusReview focus areas: security, quality, performance, education, allnoall
review_personaReview style: normal (balanced), mentor (educational), security-auditor. All personas share the same standardized 8-category defect checklist; the persona only changes tone and emphasis.nonormal
custom_instructionsAdditional instructions for the LLM reviewer (e.g., 'We use Django REST Framework')no""
enable_scoringEnable educational scoring rubric (true/false). Adds a 0-25 score to each review.nofalse
loggingLog level: debug, info, warning, errornowarning

no outputs