ViriSIM · Best Practices

Overview

Following these best practices will help you optimize your ViriSIM integration, reduce costs, improve security, and get the most value from AI observability and compliance.

Token Management

Cost Optimization

Tokens are consumed with each audit submission (max 2,000 tokens per request). Efficient token usage reduces costs and ensures continuous operation.

💰 Token Best Practices

Monitor token usage — Check Analytics → Token Usage dashboard regularly

Optimize payload size — Keep input/output under 1,000 tokens when possible

Set up alerts — Configure low token notifications in Settings → Notifications

Batch processing — Send multiple audits sequentially, not in parallel (one at a time)

Estimate costs — Each token costs less than $0.00001 (varies by plan)

Token Requirements:

Manual fine-tuning: 20,000+ tokens
Automation: 1,000,000+ tokens
Regular audits: As needed based on volume

API Key Security

Authentication

Never hardcode keys — Use environment variables or secret managers

Rotate keys every 90 days — Generate new keys and update your applications

Revoke compromised keys immediately — Delete from Integration → API Keys

Use separate keys per environment — Different keys for development, staging, production

Use descriptive names — "Production API", "Dev Testing" for easy identification

// ✅ DO: Use environment variables
const apiKey = process.env.VIRIDEED_API_KEY;

// ❌ DON'T: Hardcode keys in source code
// const apiKey = "sk-viri-abc123...";  // NEVER DO THIS

Data Minimization

Efficiency

Sending only necessary data reduces token consumption and improves processing speed.

Truncate long inputs — Keep user inputs under 1,500 tokens when possible

Redact PII before sending — Remove emails, phone numbers, addresses

Set country fields accurately — Helps with jurisdiction-specific regulations

Use correct format tags — Set format to "text" or "json" accurately

Token Limit: Maximum 2,000 tokens per submission (input + output combined). Larger requests will be rejected with error Tokens above the limit of 2k. Incorrect requests could incure 10 tokens per request.

Webhook Optimization

Integration

Don't await audit completion — The API returns immediately with a submission_id

Store submission IDs — Reference them later to find results in I/O Logs

Implement retry logic — Use exponential backoff for failed requests

Log errors — Capture and monitor error responses for debugging

Test before production — Use cURL or Postman to verify integration

// ✅ Best practice: Async audit submission
async function submitAudit(userInput, aiOutput) {
    const response = await fetch('https://https://analyzecompliance-vrbinbrbkq-uc.a.run.app', {
        method: 'POST',
        headers: { 'X-API-Key': apiKey, 'Content-Type': 'application/json' },
        body: JSON.stringify(payload)
    });
    
    const result = await response.json();
    const submissionId = result.submission_id;
    
    // Store ID for later reference — don't wait for audit completion
    await saveSubmissionId(submissionId);
    
    // Return immediately — audit runs asynchronously
    return { submissionId, status: 'processing' };
}

Test Webhook: Use the interactive demo on CodePen: ViriSIM Webhook Demo →

Fine-Tuning

Model Improvement

Start with sufficient examples — Minimum 10, recommended 50-500 examples

Use automation for continuous improvement — Schedule weekly or monthly retraining

Select quality training data — Focus on non-compliant audits with clear violations

Monitor job status — Check Fine-Tune → History for job progress

Download audit reports — Save compliance documentation for your records

Use recommended hyperparameters — 3 epochs, auto batch size for most cases

Automation Requirements:

Minimum 10 training examples (2+ audit logs)
1,000,000+ tokens required to activate automation
If disabled for 3+ cycles, automation will be terminated

Storage Management

Data Retention

ViriSIM provides 1 GB of free storage for audit logs. Proactive management prevents service interruption.

Download logs regularly — Export archives before storage fills up

Delete old logs — Remove logs older than 90 days to free space

Monitor usage — Check Settings → Logs Storage for current usage

Configure storage alerts — Set notifications at 70% (warning) and 90% (critical)

Plan for retention — Logs retained for 7 years (EU AI Act compliance)

Storage Breakdown:

AI I/O Logs: ~60% of storage (metadata & compliance audits)
Audit Trails: ~30% of storage (regulatory records)
API Request Logs: ~10% of storage (calls and analytics)

Integration Security

Safe Connections

Always use HTTPS — Never send API keys over HTTP

Monitor integration logs — Review webhook activity for anomalies

Use IP allowlisting — Restrict API access to trusted IPs (Enterprise feature)

Test with sandbox first — Validate integration before production

Keep dependencies updated — Regular security patches

// ✅ Secure integration example
const apiKey = process.env.VIRIDEED_API_KEY;
const headers = new Headers({
    'X-API-Key': apiKey,
    'Content-Type': 'application/json'
});

const response = await fetch('https://https://analyzecompliance-vrbinbrbkq-uc.a.run.app/', {
    method: 'POST',
    headers: headers,
    body: JSON.stringify(payload)
});

Quick Reference Checklist

Store API keys in environment variables

Rotate keys every 90 days

Monitor token balance regularly

Keep payloads under 2,000 tokens

Store submission IDs for reference

Implement retry logic with backoff

Download and delete logs to manage storage

Test webhooks before production

Enable MFA for admin accounts

Set up notifications for low tokens/storage

Best Practices