ecodrix/Erix Buddy AI Code Reviewer

Bring-your-own-key AI code reviewer supporting Gemini, OpenAI, Anthropic, Ollama, Groq, and more.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Jun 26, 2026
License
MIT

Pinned Snippet

workflow.ymlSHA-pinned
uses: ecodrix/erix_buddy@90024453ab1020121e77c83b97e77f905203e616 # code_reviewer

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

namedescriptionrequireddefault
providerLLM provider id: gemini | vertex | openai | anthropic | ollama | openrouter | groq | azure-openai | deepseek | mistral | xai | ecodrixnogemini
api-keyAPI key for the chosen provider. Use a repository secret.no
modelModel id (provider-specific). Falls back to the provider's recommended default.no
base-urlOverride the provider's API base URL (Azure / Ollama / custom relays)no
vertex-projectGoogle Cloud project id (only for provider=vertex)no
vertex-locationVertex AI region (only for provider=vertex)nous-central1
azure-deploymentAzure OpenAI deployment name (only for provider=azure-openai)no
modeReview mode: branch | staged | last | unstaged | full | auditnobranch
branchBase branch ref to diff against in branch modenoorigin/main
full-rootSubdirectory root for full-codebase audit modeno.
pathsComma- or newline-separated path filters (regex/substring) for audit modeno
max-filesCap files reviewed in audit modeno500
severityMinimum severity to report: critical | major | minor | nitpick | infonominor
categoriesComma-separated list of categories to surface (security,correctness,performance,maintainability,style,testing,documentation,architecture,accessibility,design-system). Defaults to all.no
concurrencyParallel file reviewsno4
report-pathMarkdown report output pathnocode-review-report.md
fail-on-blockingExit non-zero when critical/major issues are foundnotrue
no-walkthroughSkip the PR walkthrough generation phase (saves one LLM call)nofalse
no-mermaidSkip the mermaid sequence diagram in the walkthroughnofalse
no-architectureSkip the cross-cutting architecture pass (saves one LLM call)nofalse
no-testsSkip generating test-coverage gap suggestionsnofalse
custom-instructionsFree-form project-specific guidelines injected into the system promptno
custom-instructions-filePath to a file in the repo whose contents become the customInstructions. Useful for multi-line prompts that don't fit cleanly in YAML.no
post-pr-commentPost (or update) a sticky PR comment with the markdown reportnotrue
github-tokenOverride GITHUB_TOKEN for PR comment posting (e.g. when using a GitHub App token with elevated permissions)no
verboseEnable verbose logging in the underlying CLI (--verbose). Useful for debugging schema-validation retries.nofalse
namedescription
report-pathPath to the generated markdown report