Introduction
IsoFluent is a containerized intelligence layer for payment processing. It acts as a middleware service that ingests raw ISO 8583 hex dumps and returns structured, human-readable JSON enriched with fraud analysis.
Unlike traditional parsers, IsoFluent uses a localized Small Language Model (SLM) to understand context, making it capable of explaining why a transaction failed, not just what failed.
Installation v2.5.0
IsoFluent is distributed as a private Docker image. It is designed to run air-gapped within your VPC.
1. Pull the Image
# Login to the private registry (Creds provided upon license purchase)
docker login registry.isofluent.com
# Pull the latest engine
docker pull registry.isofluent.com/engine:latest
2. Run the Container
docker run -d \
-p 8080:8080 \
-v $(pwd)/config:/app/config \
--name isofluent-engine \
registry.isofluent.com/engine:latest
Configuration
The engine is controlled via a single YAML file mounted to /app/config/isofluent.yaml.
# /app/config/isofluent.yaml
server:
port: 8080
workers: 4
iso_spec:
standard: "VISA_BASE_I" # Options: VISA_BASE_I, MC_IPM, ISO_87
encoding: "EBCDIC" # Options: ASCII, EBCDIC
# AI Analysis Settings
analysis:
fraud_detection: true
explainability: verbose # options: concise, verbose
privacy:
mask_pan: true # Masks Field 2 (Primary Account Number)
mask_track2: true # Masks Field 35
REST API Usage
Once running, the engine exposes a RESTful endpoint for real-time parsing.
POST /v1/parse
Accepts raw hex strings (with or without headers).
Request (cURL)
curl -X POST http://localhost:8080/v1/parse \
-H "Content-Type: application/json" \
-d '{
"hex": "0200B23A800122E000000000000004000000303030303030..."
}'
Response (JSON)
{
"meta": {
"latency_ms": 12,
"spec": "VISA_BASE_I"
},
"parsed": {
"mti": "0200",
"mti_desc": "Financial Transaction Request",
"fields": {
"3": {
"value": "000000",
"desc": "Purchase / Savings / None"
},
"39": {
"value": "05",
"desc": "Do Not Honor"
}
}
},
"ai_analysis": {
"risk_level": "HIGH",
"summary": "Transaction declined due to 'Do Not Honor'. Correlated with 3 previous attempts on this PAN within 60 seconds suggesting a velocity attack."
}
}
Need enterprise support? Contact Engineering.