PowerDecision User Guide

Decision Wizard

Overview

The Decision Wizard is a visual rule editor that allows business users to create decision models without writing code. It provides a step-by-step interface for defining inputs, outputs, rules, and conditions that are automatically converted to DMN format for execution.

Wizard Workflow

The wizard guides you through a structured process to create decision models:

Features

• Template loading — Start from any of the 7 pre-built templates (loan approval, credit limit, etc.)
• Condition builder — Visual dropdowns for operators (equals, greater than, between, in list, etc.)
• Multi-output support — Define multiple output columns per rule
• Default rules — Add a catch-all rule that matches when no other rules do
• JSON export — Download the model definition as JSON for version control
• Direct upload — Submit the model directly to the PowerDecision API for immediate use
• Validation — Real-time validation of inputs, outputs, and rule completeness

API Reference

All API endpoints are served on port 8031 by default. The base path for decision endpoints is /api/powerdecision/v1.

Submit Decision

Submit a decision request for asynchronous evaluation. Returns a job ID for polling.

Request Body (DecisionSpec):

{
  "model_id": "loan_approval",
  "version": "1.0",
  "context": {
    "purpose": "Evaluate loan application"
  },
  "inputs": {
    "credit_score": 720,
    "annual_income": 85000,
    "loan_amount": 250000,
    "employment_status": "employed",
    "debt_to_income": 0.28,
    "loan_purpose": "home_purchase"
  }
}

Request Fields:

Response (202 Accepted):

{
  "jobId": "dec-a1b2c3d4",
  "status": "PENDING"
}

Error Response (400 Bad Request):

{
  "error": "Model not found: unknown_model"
}

Get Decision Result

Poll for the result of a submitted decision job.

Response (200 OK — completed):

{
  "model_id": "loan_approval",
  "version": "1.0",
  "status": "SUCCESS",
  "decision": {
    "approval_status": "approved",
    "max_loan_amount": 300000,
    "interest_rate": 4.5,
    "reason": "Good credit score with stable employment"
  },
  "explanation": "Rule 2 matched: credit_score >= 700 AND employment_status = employed",
  "metadata": {
    "evaluation_time_ms": 45,
    "hit_policy": "FIRST",
    "rules_evaluated": 5,
    "rules_matched": 1
  }
}

Response Fields:

Upload Model

Upload a decision model from wizard JSON. The JSON is validated, converted to DMN XML, and stored for evaluation.

Request Body:

{
  "model_id": "customer_risk",
  "name": "Customer Risk Assessment",
  "description": "Evaluates customer risk level based on profile data",
  "version": "1.0",
  "inputs_schema": [
    { "name": "account_age_months", "type": "number" },
    { "name": "transaction_count", "type": "number" },
    { "name": "country", "type": "string" }
  ],
  "outputs_schema": [
    { "name": "risk_level", "type": "string", "allowed_values": ["low", "medium", "high"] },
    { "name": "review_required", "type": "boolean" }
  ],
  "policy": "FIRST",
  "rules": [
    {
      "conditions": [
        { "input": "account_age_months", "operator": "lt", "value": 3 },
        { "input": "country", "operator": "in", "value": ["NG", "PH", "RO"] }
      ],
      "outputs": { "risk_level": "high", "review_required": true },
      "priority": 1,
      "reason": "New account from high-risk country"
    }
  ]
}

Response (200 OK):

{
  "model_id": "customer_risk",
  "version": "1.0",
  "status": "uploaded",
  "message": "Model uploaded and DMN generated successfully"
}

List Models

Returns all registered decision models (both built-in and uploaded).

Response (200 OK):

{
  "models": [
    {
      "model_id": "capacity_adjustment",
      "name": "Capacity Adjustment",
      "version": "1.0",
      "type": "builtin",
      "description": "Built-in capacity adjustment model"
    },
    {
      "model_id": "loan_approval",
      "name": "Loan Approval",
      "version": "1.0",
      "type": "runtime",
      "description": "Uploaded via wizard"
    }
  ],
  "count": 2
}

Health Endpoints

PowerDecision provides multiple health check endpoints for different orchestration needs:

Additional Development Endpoints:

Model Management

Built-in Models

PowerDecision ships with pre-compiled DMN models that are available immediately without upload:

Built-in models are compiled into the application at build time via Kogito. They are evaluated through the Kogito runtime directly, which provides the fastest execution path.

Uploading Models

Models uploaded via the wizard or API go through this pipeline:

1. Validation — The model_id is checked for uniqueness and format
2. DMN Generation — DmnGenerator converts the JSON to DMN XML with FEEL expressions
3. File Storage — The DMN XML is written to models/{model_id}.dmn
4. Registry Update — Model metadata is stored in Redis for quick lookup
5. Runtime Ready — The model is immediately available for evaluation via RuntimeDmnEvaluator

Model Registry

The model registry uses Redis as the metadata store with filesystem as the source of truth for DMN files:

• Redis — Stores model metadata (ID, name, version, type, timestamps)
• Filesystem — Stores the actual DMN XML files in the models/ directory
• Startup sync — On startup, the registry loads from Redis and scans the filesystem for orphaned DMN files not yet registered
• Fallback — If Redis is unavailable, models are still loaded from the filesystem

Configuration

Environment Variables

PowerDecision is configured via Quarkus application properties and environment variables:

CORS Configuration

CORS is enabled by default with permissive settings for development. For production, restrict origins:

# application.properties
quarkus.http.cors=true
quarkus.http.cors.origins=https://yourdomain.com,https://wizard.yourdomain.com
quarkus.http.cors.methods=GET,POST,PUT,DELETE,OPTIONS
quarkus.http.cors.headers=Content-Type,Authorization

Redis Configuration

Redis is used for the model registry metadata. If Redis is unavailable, the system falls back to filesystem-only mode.

# Default Redis configuration
quarkus.redis.hosts=redis://localhost:6379

# Docker Compose maps container port 6379 to host port 6381
# to avoid conflicts with other Redis instances
quarkus.redis.hosts=redis://redis:6379  # inside Docker network

Redis stores:

• Model metadata (JSON hash per model_id)
• Model version history
• Registry index for fast model listing

Deployment

Docker

PowerDecision uses a multi-stage Docker build (Maven build → JRE 17 Alpine runtime):

# Build the image
docker build -t powerdecision-api ./decision-api

# Run standalone
docker run -d \
  --name powerdecision \
  -p 8031:8031 \
  -v $(pwd)/models:/app/models \
  -e QUARKUS_REDIS_HOSTS=redis://host.docker.internal:6379 \
  powerdecision-api

The Dockerfile uses a custom entrypoint that starts the Quarkus application with the -Dquarkus.http.host=0.0.0.0 flag to listen on all interfaces.

Docker Compose

The recommended way to run PowerDecision with its Redis dependency:

# Start all services
docker compose up -d

# View logs
docker compose logs -f decision-api

# Stop all services
docker compose down

The Docker Compose configuration includes:

Persistent volumes:

• ./models:/app/models — DMN model files (filesystem storage)
• redis-data:/data — Redis persistence

Hetzner Cloud Deployment

For production deployment on Hetzner Cloud, refer to the Hetzner Deployment Guide in the project root. Key considerations for PowerDecision:

• Memory — JVM-based application; allocate at least 512 MB for the container
• Redis — Can share a Redis instance with other PowerTools products, but use a different database number
• Models volume — Mount a persistent volume for the models/ directory to survive container restarts
• Health checks — Use /health/liveness for Docker HEALTHCHECK and orchestrator monitoring
• Reverse proxy — Expose through Traefik or nginx with TLS termination

Troubleshooting