Roles and Permissions
Understanding the two-layer permission system -- access roles and content roles.
Two types of roles
Planara has two distinct role systems that serve different purposes. This is a common source of confusion, so it is worth understanding clearly.
1. Access roles (User management)
These control who can access the backoffice and what they can do. Managed on the Team page.
| Role | Backoffice access | Frontend access |
|---|---|---|
| Super Admin | Full access to all namespaces | Full access |
| Tenant Admin | Full access to their namespace | Full access |
| User | No backoffice access | Query access via frontend |
2. Content roles (Response configuration)
These control what information appears in Planara's AI-generated responses. Managed on the Roles page.
| Role | Typical configuration |
|---|---|
| Technician | All doc types visible, part numbers shown, torque specs shown, CKL enabled |
| Operator | Procedures and narrative only, no part numbers, escalation message for low-confidence answers |
| Inspector | Safety warnings prioritized, CKL enabled, retrieval relevance scores visible |
How they interact
A user in the system has:
- An access role (super_admin, tenant_admin, or user) -- determines what they can do
- A content role (technician, operator, etc.) — determines what they see in AI responses
These are independent. A tenant_admin can query the frontend as an "operator" role to test the operator experience, while still having full backoffice access.
Content role settings explained
doc_type_visibility
An array of document content types that are included in retrieval for this role. Options:
spec_table-- technical specification tablesprocedure-- step-by-step proceduressafety_warning-- DANGER/WARNING/CAUTION/NOTICE labelsnarrative-- descriptive text, theory of operationparts_list-- part numbers and quantitiesdiagram-- technical diagrams and schematics
confidence_threshold
A number from 0 to 1. When the pipeline's confidence score falls below this threshold, the response is replaced with the escalation_message.
- 0 -- show all responses (default for technician roles)
- 0.5 -- moderate filter
- 0.7 — strict filter (common for operator roles)
escalation_message
The text shown when confidence is below the threshold. Example: "We could not find a reliable answer in the documentation. Please contact your authorised dealer for assistance." Tenants typically substitute their own support number or contact URL here.
This is critical for operator-facing roles. It prevents low-confidence AI answers from being shown to users who cannot evaluate their accuracy.
Boolean toggles
| Setting | Description | Typical: Technician | Typical: Operator |
|---|---|---|---|
ckl_visible | Show compliance citations | Yes | No |
show_part_numbers | Include part numbers | Yes | No |
show_torque_specs | Include torque values | Yes | No |
show_rerank_scores | Show retrieval relevance scores | No (debug) | No |
show_skill_indicator | Show active skill name | No (debug) | No |
show_recovery_indicator | Show recovery status | No (debug) | No |
Vertical templates
When you click "Load Templates" on the Roles page, Planara seeds pre-configured roles based on your workspace's vertical:
- Marine vertical: Technician + Operator roles with marine-appropriate defaults
- Factory vertical: Technician + Supervisor roles
- Vehicle vertical: Technician + Customer roles
Templates can be customized after loading. They provide a sensible starting point so you do not need to configure every setting from scratch.