v1.1.0 — Production Ready

Fraud detection
that scales to millions

Real-time risk scoring combining deterministic rules with AI-enhanced analysis. Optional biometric face verification. Deploy in minutes, scale without limits.

142
ms Response
99.9%
Uptime SLA
0
External State
POST /analyze
$ asguard transaction
{
"transaction_id": "txn_8f4d2a",
"amount": 284750.00,
"currency": "USD",
"location": "Lagos, NG"
}
⟳ analyzing...
{
"risk_score": 47
"risk_level": MEDIUM
"processing_time": "142ms"
}

Trusted by engineering teams

FINTECH_CORP PAYFLOW SECUREBANK MERCHANT_X

Built for scale
designed for precision

Every component optimized for high-throughput transaction processing with zero external state dependencies. Deterministic scoring meets AI augmentation.

01

Rule-Based Scoring

Deterministic evaluation of amount thresholds, currency risk, device fingerprints, and geolocation patterns with configurable weights.

Sub-50ms latency
100% explainable
02

AI Augmentation

Automatic escalation to Llama-3.3-70b via Groq API for complex edge cases. AI enhances but never overrides critical rules.

Triggered at score ≥40
~500ms latency
03

Zero-Trust Stateless

No database required. No sensitive data persistence. Process millions of transactions without storage overhead or compliance friction.

BYOK AI integration
GDPR/CCPA compliant
04

Biometric Verification

Optional face recognition microservice for user onboarding and transaction-level validation. Secure, stateless, embeddable.

128-dim embeddings
Configurable threshold
05

Full Auditability

Every decision includes detailed reasoning, confidence scores, and factor breakdown for compliance and forensic analysis.

JSON audit trails
Regulatory ready
06

Performance at Scale

Horizontal scaling with no bottlenecks. Load balanced across multiple instances with zero coordination overhead.

Millions of TPS
Stateless design

Hybrid intelligence
architecture

A unique combination of speed and determinism through rule-based systems, enhanced with nuanced AI for sophisticated edge cases.

Step 01 — Authenticate

API key validation via middleware. Secure request origin verification.

Step 02 — Parse

JSON binding and schema validation. Type safety and data integrity checks.

Step 03 — Score

Weighted rule engine evaluation. Risk calculation with granular factor breakdown.

Step 04 — Escalate

AI analysis triggered for medium-high risk. Contextual reasoning and pattern detection.

Step 05 — Decide

Structured JSON response. Actionable recommendations and risk classification.

Hybrid Decision Logic
func Assess(t Transaction) Result { score := CalculateBaseScore(t) if score >= 40 { aiAnalysis := QueryAI(t, score) if aiAnalysis.Risk > score { score = aiAnalysis.Risk // Upgrade only } } return Finalize(score) }

AI enhances but never downgrades critical safety flags

Modern engineering stack

Built for performance and developer experience

Go

1.25+

Gin

Framework

Groq

LLM API

Docker

Containers

K8s

Orchestration

AWS

Cloud

Deploy anywhere: Kubernetes, AWS, GCP, Azure, On-Premise

Drop-in HTTP integration

Single endpoint. JSON in, JSON out. No SDKs. No complexity. Integrate into your payment flow in under 30 minutes.

1

Send transaction data

POST to /analyze with amount, currency, location

2

Receive risk assessment

Score, level, and reasoning returned immediately

3

Act on the decision

Approve, review, or block based on risk level

curl example
$ curl -X POST http://api.asguard.local/analyze \
-H "x-api-key: sk_live_..." \
-H "Content-Type: application/json" \
-d '{
"amount": 50000,
"currency": "USD",
"ip_address": "203.0.113.1"
}'
$ Response: 200 OK
{
"risk_score": 23,
"risk_level": "LOW"
}

Ready to secure your transactions

Deploy Asguard in minutes. Scale to millions of requests. Pay only for what you use with bring-your-own-key AI integration.

Apache 2.0 License Self-Hosted Community Driven