New Tenant Setup

Overview

Getting a new tenant from zero to a working AI assistant takes four steps. The fastest path (self-serve onboarding from a single PDF) takes under 30 minutes. A production deployment with integrations takes 2-4 weeks — see the Customer Onboarding Guide for the full playbook.

This page covers the quick start path.

Step 1: Provision the tenant

From the backoffice

Go to the Provisioning page and fill in the form:

1. Enter a namespace (URL-safe identifier, e.g., acme_marine)
2. Enter the display name (e.g., "Acme Marine Services")
3. Enter the manufacturer name
4. Enter the equipment models the tenant services
5. Toggle demo data on if you want a pre-populated dashboard
6. Click Provision Tenant

The system creates the namespace, database records, prompt config, domain ontology, and optional demo data in under 10 seconds.

From the CLI

.venv/bin/python scripts/provision_tenant.py \
    --namespace acme_marine \
    --display-name "Acme Marine Services" \
    --manufacturer "Acme" \
    --equipment-models "Model A, Model B" \
    --demo-data

From the API

curl -X POST https://your-api/v1/admin/provision \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "namespace": "acme_marine",
    "display_name": "Acme Marine Services",
    "manufacturer": "Acme",
    "equipment_models": ["Model A", "Model B"],
    "demo_data": true
  }'

Step 2: Upload a manual

Go to Upload and Ingestion in the backoffice. Upload the manufacturer's service manual as a PDF. Tag it with the applicable equipment models.

The ingestion worker picks up the job automatically. You can monitor progress on the upload page — each pipeline stage shows as a progress step.

For a typical 500-page service manual, expect:

• Parsing and chunking: 3-5 minutes
• Entity extraction and linking: 1-2 minutes
• Indexing: 1-2 minutes
• QA gates: under 30 seconds
• Total: 5-10 minutes

Step 3: Run auto-onboarding (optional but recommended)

The self-serve onboarding pipeline generates a complete configuration from your ingested documents:

.venv/bin/python scripts/onboard.py \
    --namespace acme_marine \
    --equipment-models "Model A, Model B"

This runs the full onboarding chain automatically: prompt config, maintenance ontology, part number patterns, domain ontology, CKL triggers, eval queries, deterministic paths, diagnostic templates, failure-mode ontology, whole-asset overview, suggestion chips, a validation eval, and a readiness verdict. Takes about 30 seconds. See Customer Onboarding for the stage-by-stage detail.

After onboarding, the tenant has a tuned pipeline — not just raw documents, but manufacturer-specific terminology, fault code mappings, and test queries ready for review.

Step 4: Access the technician dashboard

Open the technician frontend URL with the tenant's namespace:

https://your-frontend-domain/acme_marine

If demo data was enabled, the dashboard shows a populated queue with realistic work orders. Select a job to start the full workflow: assessment, diagnostics, repair procedures, verification, and close.

If no demo data was enabled, the dashboard shows an empty queue. Create a job manually or connect a DMS integration to populate it.

What to do next

• Run an evaluation — Go to Eval Management and run the auto-generated eval queries to establish a quality baseline
• Invite technicians — Go to Team to send magic link invitations
• Configure CKL — Enable the Compliance Knowledge Layer if safety standard citations are needed
• Connect integrations — See the Customer Onboarding Guide for DMS, telemetry, and ERP setup

Related

• Tenant Provisioning — detailed provisioning options and configuration
• Demo Mode — how auto-generated mock data works
• Customer Onboarding Guide — the full production onboarding playbook
• Onboarding Wizard — the in-app checklist for new workspaces