Skip to content
← Blog
4 min read

Ship Verified AI: Add VEROQ to Your CI/CD Pipeline

Every PR fact-checked. Contradicted claims block the deploy. Verification receipts for compliance. One GitHub Action, zero hallucinations in production.

Your AI generates text. Your users see it. But between generation and display, who checks if the claims are actually true? Most teams don't — and that's how hallucinations ship to production.

VEROQ's CI/CD Shield adds verification to your deployment pipeline. Every PR that touches AI-generated content gets its claims fact-checked before merge. Contradicted claims block the deploy.

The GitHub Action

Add this to your workflow. Every push runs veroq test against your AI outputs. Claims below the trust threshold fail the build.

yaml
# .github/workflows/veroq-shield.yml
name: VEROQ Shield
on: [push, pull_request]

jobs:
  verify:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: veroq-ai/shield-action@v1
        with:
          api-key: ${{ secrets.VEROQ_API_KEY }}
          threshold: 0.7
          fail-on-contradiction: true
          source: "ci-pipeline"

What It Checks

The action scans your codebase for AI-generated content — LLM outputs stored in fixtures, generated documentation, automated reports, chatbot responses. Each factual claim is extracted and verified against VEROQ's intelligence corpus + live web sources.

  • Supported claims — pass silently
  • Contradicted claims — fail the build with corrections
  • Unverifiable claims — warn but don't block (configurable)
  • Low trust score — fail if overall confidence drops below threshold

CLI Integration

Don't use GitHub? Run veroq test anywhere — local dev, GitLab CI, Jenkins, CircleCI. Same verification, any pipeline.

bash
# Install
npm install -g @veroq/cli

# Test a file
veroq test --file output.json --threshold 0.7

# Test inline
veroq shield "NVIDIA reported $22B in Q4 revenue" --fail-on-contradiction

# In any CI pipeline
veroq test --glob "fixtures/**/*.json" --source ci

Verification Receipts

Every CI run generates verification receipts — hashable proof that your AI outputs were fact-checked at deploy time. Compliance teams can audit any release and see exactly which claims were verified, by which sources, at what confidence level.

Receipts are stored permanently and accessible via API:

bash
# JSON receipt
curl https://api.veroq.ai/api/v1/verify/receipt/vr_abc123

# Shareable HTML page
https://api.veroq.ai/api/v1/verify/receipt/vr_abc123?format=html

# Embeddable badge
https://api.veroq.ai/api/v1/verify/badge/vr_abc123

Who This Is For

  • Teams shipping AI-powered products — catch hallucinations before users see them
  • Compliance-regulated industries — audit trail for every AI output in your system
  • Content teams — fact-check AI-generated articles, reports, and summaries
  • Trading desks — verify market claims before they influence decisions
Get API KeyDocsTry Shield