Fortza API Guide

1 Create an API key

In Fortza, open Connect a system and find Generate key. Give your key a recognizable name, such as salesforce-test, then generate it.

2 Find your model and fields

Open Models. Copy a live model’s full name and the field names listed on its card. Use the types and formats from the data you trained it with.

This guide uses fortza-acme-ops-demo. Replace the example model, fields, and values with your own.

3 Build and send your request

Replace YOUR_API_KEY, the model name, and the record fields. Use a key and model from the environment selected above.

curl -X POST "https://sandbox-api.fortza.ai/api/v1/inference" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "fortza-acme-ops-demo",
    "records": [
      {
        "recorded_at": "2026-10-22T09:15:00Z",
        "source_id": "SRC-104",
        "category": "standard",
        "amount": 1250,
        "quantity": 12,
        "duration_seconds": 48.5
      },
      {
        "recorded_at": "2026-10-22T09:16:00Z",
        "source_id": "SRC-104",
        "category": "standard",
        "amount": 98400,
        "quantity": 1,
        "duration_seconds": 2.1
      }
    ]
  }'

Run this in Bash (macOS, Linux, or Git Bash on Windows). For a visual option, import the cURL example into Postman and select Send. This multiline command uses Bash syntax, not PowerShell syntax.

1 Store the key in credentials

In Salesforce Setup, create an External Credential using custom authentication and a named principal. Store your Fortza key as an authentication parameter named ApiKey.

Configure the Authorization header to send Bearer followed by that parameter. Create a Named Credential, Fortza_API, pointing to your environment’s API host and linked to the External Credential. Grant the flow’s running user access to the principal through a permission set.

Keep the key in credentials, rather than a flow variable or formula. Follow Salesforce’s Named Credentials documentation ↗ for the settings available in your org.

2 Create fields for the result

3 Map your fields and create a callout

Map each trained Fortza field to a Salesforce field, preserving its training format. For example, your model’s amount might map to Opportunity Amount. This mapping depends on your own model.

In Flow Builder, create an HTTP Callout action using Fortza_API, method POST, and path /api/v1/inference. Use a one-record version of the request example and the response example to define the action. Read the returned results collection, rather than expecting a top-level verdict.

See Salesforce’s HTTP Callout guide ↗. Check that nested evidence and nullable values work with the generated types; a simplified response mapping or custom action may be needed.

4 Run after save and handle failures

Create a record-triggered flow optimized for Actions and Related Records. Put the HTTP callout on its Run Asynchronously path. After success, write the verdict, percentile, and current time back to the record, plus the reasons when the verdict is an anomaly. Branch on the HTTP status listed in Errors.

Only trigger when relevant input fields change. Updating Fortza’s output fields must not trigger another check. Add a fault path and handle unsuccessful HTTP responses so an admin can investigate. An owner review task for an anomaly is optional.

5 Test before activating

Check a successful response, missing-field error, invalid key, unavailable model, and empty evidence. Confirm that writing the results back does not cause repeated callouts. Plan batches and monthly row usage before a mass import; never automatically replay a failed batch without checking whether it was processed.