TrustModel Eval Action

AI trust evaluation in your CI/CD pipeline. Evaluate any AI model across 10 safety dimensions — safety, fairness, accuracy, privacy, transparency, robustness, accountability, explainability, compliance, reliability.

First scan free. No credit card required.

Quick Start

- name: Evaluate AI Trust Score
  uses: pdxlab/trustmodel-eval-action@v1
  with:
    api-key: ${{ secrets.TRUSTMODEL_API_KEY }}
    model-name: 'my-llm-v2'
    min-score: '0.80'

Usage

Basic — Free Scan (no credits needed)

name: AI Trust Check
on: [pull_request]

jobs:
  trust-eval:
    runs-on: ubuntu-latest
    steps:
      - name: TrustModel Evaluation
        uses: pdxlab/trustmodel-eval-action@v1
        id: trust
        with:
          api-key: ${{ secrets.TRUSTMODEL_API_KEY }}
          model-name: 'my-chat-model'
          model-type: 'chat'
          min-score: '0.75'

      - name: Comment on PR
        if: always()
        uses: actions/github-script@v7
        with:
          script: |
            github.rest.issues.createComment({
              owner: context.repo.owner,
              repo: context.repo.repo,
              issue_number: context.issue.number,
              body: `${{ steps.trust.outputs.summary }}`
            });

Gate deployment on trust score

- name: Evaluate Trust
  uses: pdxlab/trustmodel-eval-action@v1
  id: trust
  with:
    api-key: ${{ secrets.TRUSTMODEL_API_KEY }}
    model-name: 'production-model'
    min-score: '0.80'
    free-scan: 'false'  # Use credits for full evaluation

- name: Deploy
  if: steps.trust.outputs.passed == 'true'
  run: ./deploy.sh

Evaluate specific dimensions only

- uses: pdxlab/trustmodel-eval-action@v1
  with:
    api-key: ${{ secrets.TRUSTMODEL_API_KEY }}
    model-name: 'hr-screening-model'
    dimensions: 'fairness,privacy,compliance'
    min-score: '0.85'

Inputs

Input	Required	Default	Description
api-key	Yes	—	TrustModel API key (get one free)
model-name	Yes	—	Name of the AI model to evaluate
model-type	No	chat	Model type: chat, completion, embedding, classification, agent
dimensions	No	all	Comma-separated dimensions, or all for all 10
min-score	No	0.0	Minimum score to pass (0.0–1.0). Action fails if below.
endpoint-url	No	—	Model API endpoint (for custom model evaluation)
free-scan	No	true	Use FreeScan (first eval free per model)
fail-on-error	No	false	Fail action if API errors (vs warning)

Outputs

Output	Description
trust-score	Overall trust score (0.0–1.0)
passed	Whether score met min-score threshold (true/false)
dimensions-json	JSON with per-dimension scores
report-url	Link to full report on TrustModel console
summary	Markdown summary (also written to GitHub Step Summary)

10 Trust Dimensions

Dimension	What it measures
Safety	Harmful content, toxicity, dangerous instructions
Fairness	Bias across demographics, stereotyping, equal treatment
Accuracy	Factual correctness, hallucination rate, citation quality
Privacy	PII leakage, data retention, consent handling
Transparency	Decision explainability, confidence calibration
Robustness	Adversarial resistance, prompt injection, jailbreak resilience
Accountability	Audit trail, versioning, reproducibility
Explainability	Can the model explain its reasoning?
Compliance	EU AI Act, NIST AI RMF, ISO 42001 alignment
Reliability	Consistency, uptime, degradation handling

Pricing

Tier	Price	What you get
Free Scan	$0	First eval per model, all 10 dimensions
Credits	$100/dimension	Pay-as-you-go, JSON response
Full Eval	$1,000	All 10 dimensions, compliance report

Example Output

The action writes a rich summary to GitHub Step Summary:

✅ TrustModel AI Evaluation

Model: my-chat-model (chat)

Overall Trust Score: 🟢 0.87 / 1.00 (threshold: 0.75)

Dimension Scores

Dimension	Score	Status
Safety	0.92	✅
Fairness	0.85	✅
Accuracy	0.88	✅
Privacy	0.79	✅
Transparency	0.83	✅
...	...	...

Get Your API Key

1. Go to trustmodel.ai/developer
2. Sign up (free)
3. Get your API key
4. Add it as a GitHub secret: TRUSTMODEL_API_KEY

Links

• TrustModel.ai — AI Assurance platform
• Developer Docs
• Python SDK — pip install trustmodel
• MCP Server — For AI agents

License

MIT