alogram-payrisk 0.3.2

The official Alogram Payrisk 'Smart' SDK for Python. Features built-in resiliency, ergonomic risk intelligence, and automated identity management.

Alogram PayRisk SDK for Python

The official Python client for the Alogram PayRisk Engine.

Alogram PayRisk is a decision management and risk orchestration engine for global commerce. It fuses machine learning, behavioral analytics, and deterministic business rules into a high-fidelity scoring pipeline designed for enterprise scale and auditability.

🧠 The Three-Expert Architecture

The SDK provides unified access to three specialized risk experts:

• Risk Scoring: Real-time assessment and decision orchestration for purchases.
• Signal Intelligence: Ingestion of behavioral telemetry and payment lifecycle events.
• Forensic Data: Deep visibility into historical assessments and decision transparency.

---

🔐 Security: Trust Boundaries

Alogram enforces a strict separation between client-side telemetry and server-side decisioning.

Client Type	Key Prefix	Environment	Capabilities
AlogramPublicClient	pk_...	Browser / Mobile	Ingestion only. Cannot perform risk checks or view scores.
AlogramRiskClient	sk_...	Secure Backend	Full access. Authorized for risk decisions and forensic retrieval.

[!WARNING] Never use a Secret Key (sk_...) in a client-side environment. This will expose your tenant's sensitive forensic data.

---

🔄 Full Lifecycle Integration

A best-in-class Alogram integration follows the three-step lifecycle:

from alogram_payrisk import AlogramRiskClient
from payrisk_v1 import CheckRequest, PaymentEvent

# 1. Initialize the Secret Client (Backend Only)
with AlogramRiskClient(api_key="sk_live_...", tenant_id="tid_mycorp") as client:

    # 2. Assessment: Call before charging the customer
    decision = client.check_risk(CheckRequest(...))

    if decision.decision == "approve":
        # Process payment via your gateway...

        # 3. Lifecycle: Send the outcome back to Alogram
        client.ingest_event(PaymentEvent(
            payment_intent_id=decision.payment_intent_id,
            event_type="authorization",
            outcome={"approved": True}
        ))

---

🚀 High-Performance Integration

• 🏢 Smart Client Architecture: Specialized clients for server-side (AlogramRiskClient) and public-facing (AlogramPublicClient).
• 🛡️ Automated Identity: Injects x-api-key, Authorization, and tenant headers automatically.
• 🔄 Built-in Resiliency: Automatic exponential backoff and jittered retries (powered by tenacity).
• 🕵️ Native Observability: Built-in OpenTelemetry tracing for monitoring risk decision latency and outcomes.
• 🧩 Type Safe: Built with Pydantic v2 and full PEP 561 compliance (py.typed).

📦 Installation

pip install alogram-payrisk

🛠️ Quick Start

1. Evaluate Risk (Risk Scoring Expert)

Assess a purchase in real-time. This invokes the authoritative scoring pipeline.

from alogram_payrisk import AlogramRiskClient, CheckRequest, Purchase, Identity

client = AlogramRiskClient(api_key="sk_live_...", tenant_id="tenant_123")

# Perform the check via the Risk Scoring expert
decision = client.check_risk(CheckRequest(
    purchase=Purchase(amount=99.99, currency="USD"),
    identity=Identity(email="[e-mail]")
))

print(f"Decision: {decision.decision} | Score: {decision.decision_score}")

2. Ingest Lifecycle Events (Signal Intelligence Expert)

Stream payment lifecycle updates to the Engine for continuous model training.

from alogram_payrisk import PaymentEvent, PaymentOutcome, PaymentAuthorizationOutcome

client.ingest_event(PaymentEvent(
    event_type="authorization",
    payment_intent_id="pi_123...",
    amount=99.99,
    currency="USD",
    outcome=PaymentOutcome(
        authorization=PaymentAuthorizationOutcome(approved=True, responseCode="00")
    )
))

🔄 3. Full Lifecycle Workflow

For a complete end-to-end example showing how to correlate client-side signals, risk scoring, and fraud labeling, see: 👉 examples/full_lifecycle_workflow.py

This workflow demonstrates how to:

1. Anchor pre-order signals using sessionId and deviceId.
2. Correlate those signals during the check_risk call.
3. Handoff to the server-minted paymentIntentId for post-order events (Auth, Capture, Chargeback).

---

🚀 High-Performance Integration

To ensure sub-second risk assessment latencies and handle high-volume signal telemetry efficiently, please adhere to these network best practices:

• Persistent Client (Mandatory): Maintain a single, global instance of the AlogramRiskClient or use it as a Context Manager.
  • Anti-pattern: Creating a new client for every request forces a fresh TCP/TLS handshake.
  • Best Practice: Reuse the client to keep the underlying httpx session "hot."
• HTTP/2 Multiplexing: The SDK natively supports HTTP/2. By reusing the client, multiple requests (e.g., several signals) are automatically multiplexed over a single persistent pipe, eliminating connection management overhead.

🛡️ Error Handling & Resiliency

The SDK distinguishes between transient network issues and validation errors.

from alogram_payrisk.exceptions import ValidationError, AlogramError

try:
    decision = client.check_risk(request)
except ValidationError as e:
    # ❌ Handle invalid input (e.g., invalid BIN or malformed email)
    print(f"Validation Failed: {e.body}") 
except AlogramError as e:
    # 🚨 Handle generic API or Authentication errors
    print(f"API Error [{e.status}]: {e.message}")

🕵️ Observability (OpenTelemetry)

The SDK automatically detects OpenTelemetry. If present, it creates spans for all API calls.

from opentelemetry import trace

# Spans will automatically include 'alogram.decision' and 'alogram.trace_id'
with trace.get_tracer(__name__).start_as_current_span("checkout"):
    decision = client.check_risk(request)

🏗️ Environments

Environment	Base URL	Key Type
Production	https://api.alogram.ai	sk_live_...
Sandbox	https://api-sandbox.alogram.ai	sk_test_...
Local	http://localhost:8080	test

---

🤖 For AI Agents

If using an AI agent (ChatGPT, Claude, Gemini) for integration, provide this context:

"Use the Alogram PayRisk Python SDK. Always prefer the AlogramRiskClient for backend operations. Access the authoritative blended score via the decision_score attribute of the response. Use the idempotency_key parameter for all write operations."

---

⚖️ License

Apache License 2.0. See LICENSE for details.