← Back to home

Documentation

Overview

EDI Sentinel is a Snowflake Native App that parses, monitors, and manages EDI (Electronic Data Interchange) documents — X12 and EDIFACT supply-chain transaction sets — entirely within your Snowflake account. It is listed on the Snowflake Marketplace as a paid product.

You can ingest raw EDI files, parse them into structured data, detect when trading partners change their message formats (schema drift), and use AI to map EDI fields to your own schema — all without leaving Snowflake.

Prerequisites

  • A Snowflake account with access to the Marketplace
  • An existing database (or willingness to create one) for the app
  • For SPCS (container-based parsing): Snowpark Container Services enabled in your account
  • For AI features (drift detection, field mapping): Snowflake Cortex enabled

Installation

  1. Log in to Snowflake and open the Marketplace.
  2. Search for “EDI Sentinel” and open the listing.
  3. Click “Get” and choose the account/database/warehouse where you want to install the app.
  4. Accept the provider terms and complete the installation. The app will create the necessary objects and roles (app_admin, app_user).
  5. Run the provider bootstrap script if provided (e.g. for initial setup of stages and tasks).

Quick Start

After installation:

  1. Create or use an existing stage (e.g. edi_raw_stage) and upload sample EDI files (X12 850, 856, 810, 997 or EDIFACT ORDERS, DESADV, INVOIC, CONTRL).
  2. From the Streamlit UI (Home or Dashboard), use “Process staged files” to run the parser.
  3. Inspect parse_history for structured output and check the Trading Partners and Drift Alerts pages.
  4. In Field Mappings, use AI-suggested mappings and approve or adjust as needed.

Configuration

From the Streamlit Settings page you can:

  • SPCS pool — Configure the Snowpark Container Services pool used by the FastAPI-based parser for high-volume or API-style parsing.
  • Cortex — Enable/disable Cortex-based drift detection and field mapping; select the model used for AI suggestions.
  • Drift task — Enable or disable the hourly drift detection task.

Supported Transaction Sets

X12: 850 (Purchase Order), 856 (ASN/Shipment Notice), 810 (Invoice), 997 (Functional Acknowledgment).

EDIFACT: ORDERS, DESADV (Despatch Advice), INVOIC, CONTRL (Syntax Acknowledgment).

Architecture

Data flow:

  • EDI files land in edi_raw_stage.
  • Parsing runs via Snowpark UDFs (warehouse) or the SPCS FastAPI container (edi-processor).
  • Parsed data is written to parse_history.
  • A scheduled task (drift_detection_task) runs hourly to detect format anomalies; results go to drift_alerts.
  • Cortex powers drift analysis and zero-shot field mapping; mappings are stored for review in field_mappings.

FAQ

Do I need SPCS to use EDI Sentinel? No. You can use Snowpark UDF–based parsing without SPCS. SPCS is for higher volume or API-style use.

Is Cortex required? Cortex is used for AI-powered drift detection and field mapping. You can disable these features in Settings if needed.

How do I add a new trading partner? Partners are auto-discovered from ISA headers when you parse. You can also add or edit them in the Trading Partners page.