API Documentation v1.1.0

Introduction

H-LLM Enterprise provides real-time content safety analysis that runs in parallel with your LLM. Analyze both user inputs and LLM outputs to detect harmful content before it reaches your users.

Key Features

• 14 Detection Categories - From grooming to jailbreaks
• Multi-stage Grooming Detection - 4-stage pattern recognition
• Fraud Detection - Romance scams, phishing, tech support fraud
• Session Analysis - Track escalating threats across conversations
• Low Latency - Sub-100ms response times

Authentication

All API requests require a Bearer token in the Authorization header:

Authorization: Bearer your_api_key_here

Contact api@hallucinations.cloud to obtain your API key.

Quick Start

Get started with a simple content analysis request:

curl -X POST https://api.h-llm.enterprise/v1/analyze \
  -H "Authorization: Bearer your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Hello, how can I help you today?",
    "type": "input"
  }'

import requests

response = requests.post(
    "https://api.h-llm.enterprise/v1/analyze",
    headers={
        "Authorization": "Bearer your_api_key",
        "Content-Type": "application/json"
    },
    json={
        "content": "Hello, how can I help you today?",
        "type": "input"
    }
)

result = response.json()
if result["verdict"] == "potentially_harmful":
    print(f"Blocked: {result['category']}")
    print(f"Suggested response: {result['suggested_response']}")
else:
    print("Content is safe")

const response = await fetch('https://api.h-llm.enterprise/v1/analyze', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer your_api_key',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    content: 'Hello, how can I help you today?',
    type: 'input'
  })
});

const result = await response.json();
if (result.verdict === 'potentially_harmful') {
  console.log(`Blocked: ${result.category}`);
  console.log(`Suggested: ${result.suggested_response}`);
} else {
  console.log('Content is safe');
}

API Reference

{
  "verdict": "potentially_harmful",
  "category": "fraud",
  "confidence": 0.92,
  "severity": "high",
  "suggested_response": "I can't help create content that could be used to deceive or defraud others.",
  "details": "Matched 3 fraud/scam patterns",
  "request_id": "hlm_abc123def456",
  "timing_ms": 45
}

{
  "verdict": "safe",
  "category": null,
  "confidence": 0.95,
  "severity": null,
  "suggested_response": null,
  "details": "No harmful content detected.",
  "request_id": "hlm_xyz789ghi012",
  "timing_ms": 32
}

Detection Categories

H-LLM Enterprise detects 14 categories of harmful content across 3 priority tiers:

Tier 1 - Critical Highest Priority

Tier 2 - High Secondary Priority

Grooming Detection

H-LLM Enterprise uses multi-stage pattern recognition to detect grooming behavior:

Fraud Detection

Detects multiple types of fraud and scam patterns:

• Romance Scams - Gift card requests, stranded overseas, frozen accounts
• Tech Support Scams - Fake virus alerts, remote access requests
• Phishing - Urgent verification, credential harvesting
• Investment Fraud - Guaranteed returns, insider tips
• Authority Impersonation - IRS/FBI threats, arrest warrants

Rate Limits

Pricing

Changelog