VectorLint GitHub Action

A GitHub Action wrapper for VectorLint - A command-line tool that evaluates and scores content using LLMs. It uses LLM-as-a-Judge to catch content quality issues that typically require human judgement.

Usage

Basic Setup

name: VectorLint

on:
  pull_request:
    paths:
      - '**/*.md'
      - '**/*.mdx'

permissions:
  contents: read
  pull-requests: write
  checks: write

concurrency:
  group: vectorlint-${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:
  lint:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Run VectorLint
        uses: TRocket-Labs/vectorlint-action@v1
        with:
          llm_provider: ${{ secrets.LLM_PROVIDER }}
          # OpenAI
          openai_api_key: ${{ secrets.OPENAI_API_KEY }}
          openai_model: ${{ secrets.OPENAI_MODEL }}
          openai_temperature: ${{ secrets.OPENAI_TEMPERATURE }}
          # Perplexity
          perplexity_api_key: ${{ secrets.PERPLEXITY_API_KEY }}
          # Reviewdog options
          reporter: github-pr-check
          filter_mode: added
          fail_on_error: false
       

---

Configuration Options

Core Settings

Input	Description	Required	Default
llm_provider	LLM provider (openai, anthropic, gemini, azure-openai)	No	anthropic
vectorlint_version	VectorLint npm package version	No	latest
vectorlint_flags	Additional CLI flags for vectorlint	No	''

LLM Provider API Keys

Input	Description
openai_api_key	OpenAI API key
anthropic_api_key	Anthropic API key
gemini_api_key	Google Gemini API key
azure_openai_api_key	Azure OpenAI API key
perplexity_api_key	Perplexity API key (for technical accuracy verification)

Note: You must provide the API key matching your selected llm_provider.

Model Configuration

Input	Description	Example Values
openai_model	OpenAI model to use	gpt-4, gpt-4-turbo, gpt-4o
anthropic_model	Anthropic model to use	claude-sonnet-4-20250514, claude-3-sonnet-20240229
anthropic_max_tokens	Max tokens for Anthropic responses	4096
gemini_model	Gemini model to use	gemini-2.5-pro, gemini-3-pro
azure_openai_model	Azure OpenAI deployment name	Your deployment name

Temperature Settings

Temperature controls the randomness of LLM responses. Lower values (e.g., 0.1) produce more deterministic outputs, while higher values (e.g., 0.8) produce more creative outputs.

Input	Description	Range
openai_temperature	Temperature for OpenAI	0.0 - 2.0
anthropic_temperature	Temperature for Anthropic	0.0 - 1.0
gemini_temperature	Temperature for Gemini	0.0 - 2.0
azure_openai_temperature	Temperature for Azure OpenAI	0.0 - 2.0

Recommendation: Use 0.2 - 0.4 for consistent linting results.

Azure OpenAI Configuration

Input	Description
azure_openai_api_key	Azure OpenAI API key
azure_openai_endpoint	Azure OpenAI endpoint URL
azure_openai_deployment_name	Azure OpenAI deployment name
azure_openai_api_version	API version (e.g., 2024-02-15-preview)

Perplexity Integration

Perplexity is used for technical accuracy verification. When enabled, VectorLint can cross-reference claims and technical statements against real-world data.

Input	Description
perplexity_api_key	Perplexity API key for search-based fact verification

Reviewdog Settings

Input	Description	Default
reporter	Reporter type	github-pr-check
filter_mode	Filter mode	added
fail_on_error	Fail workflow if errors found	true
tool_name	Name shown in GitHub Checks	vectorlint
github_token	GitHub token for reviewdog	${{ github.token }}

Reporter Types

Type	Description
github-pr-check	Annotations in Files Changed tab
github-pr-review	Comments directly on PR lines
github-check	Check status only (no inline annotations)

Filter Modes

Mode	Description
added	Only new/changed lines (recommended)
diff_context	Lines in diff context
file	Entire changed files
nofilter	All lines in changed files

---

Examples

OpenAI with GPT-4

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: openai
    openai_api_key: ${{ secrets.OPENAI_API_KEY }}
    openai_model: gpt-4
    openai_temperature: '0.3'

Anthropic with Claude

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: anthropic
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
    anthropic_model: claude-3-sonnet-20240229
    anthropic_temperature: '0.2'
    anthropic_max_tokens: '4096'

Google Gemini

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: gemini
    gemini_api_key: ${{ secrets.GEMINI_API_KEY }}
    gemini_model: gemini-1.5-pro
    gemini_temperature: '0.3'

Azure OpenAI

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: azure-openai
    azure_openai_api_key: ${{ secrets.AZURE_OPENAI_API_KEY }}
    azure_openai_endpoint: ${{ secrets.AZURE_OPENAI_ENDPOINT }}
    azure_openai_deployment_name: my-gpt4-deployment
    azure_openai_api_version: '2024-02-15-preview'
    azure_openai_temperature: '0.3'

With Perplexity for Technical Accuracy

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: openai
    openai_api_key: ${{ secrets.OPENAI_API_KEY }}
    openai_model: gpt-4
    perplexity_api_key: ${{ secrets.PERPLEXITY_API_KEY }}

Strict Mode (Fail on Errors)

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: anthropic
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
    fail_on_error: true
    filter_mode: nofilter

Advisory Mode (Comments Only)

- uses: TRocket-Labs/vectorlint-action@v1
  with:
    llm_provider: gemini
    gemini_api_key: ${{ secrets.GEMINI_API_KEY }}
    fail_on_error: false
    reporter: github-pr-review

---

How It Works

1. Installs VectorLint from npm
2. Detects changed Markdown files in the PR
3. Runs AI-powered linting using your configured LLM provider
4. (Optional) Verifies technical accuracy via Perplexity
5. Pipes results to reviewdog for PR annotations

Requirements

• LLM Provider: At least one of openai_api_key, anthropic_api_key, gemini_api_key, or Azure OpenAI credentials
• Permissions: contents: read, pull-requests: write, checks: write

License

MIT