Admin Guide

This guide covers tenant administration for SUPERWISE Chat: managing users and roles, configuring projects and channels, governing knowledge bases and policies, and monitoring usage.

---

Roles and Permissions

SUPERWISE Chat uses a hierarchical role model. Each role inherits all permissions of the roles below it.

Role	Description
guest	Read-only access to assigned channels
viewer	Can read conversations and search knowledge
member	Can start and participate in conversations
project_manager	Can manage channel membership and settings
project_owner	Project ownership with elevated member rights
project_admin	Full project and user management within the tenant
tenant_admin	Full tenant administration, including plan and billing
tenant_super_admin	Super-level tenant administration

In addition to the hierarchy, these roles can be granted independently (they inherit from member):

Role	Description
governance_manager	Can view traces, policy alerts, and export audit data
channel_moderator	Can manage membership and settings for a specific channel

Assigning Roles

1. Navigate to Settings → Users.
2. Click a user to open their profile.
3. Under Roles, select a role from the dropdown and click Assign.
4. To grant channel-specific access, open Settings → Channels → [Channel] → Members and assign a role scoped to that channel.

Channel-level role assignments override the user’s tenant-level role within that channel only.

---

User Management

Inviting Users

1. Go to Settings → Users → Invite.
2. Enter the user’s email address and select their initial role.
3. Click Send Invite. The user receives an email with a sign-in link.

Invited users appear as Pending until they accept the invitation and complete sign-in.

Deactivating Users

Deactivated users cannot sign in but their history and contributions are retained.

1. Open Settings → Users.
2. Click the user, then Deactivate.

To re-activate, click Activate on the same screen.

Removing Users from Projects

1. Open Settings → Projects → [Project] → Members.
2. Find the user and click Remove.

Removing a user from a project removes their access to all channels within that project. Their tenant-level role is unchanged.

---

Projects and Channels

Creating a Project

Projects are workspaces that contain channels. A tenant can have multiple projects.

1. Navigate to Settings → Projects → New Project.
2. Enter a name and optional description.
3. Configure default member access (who can see the project).
4. Click Create.

Creating Channels

Channels within a project scope conversations to a topic or team.

1. Open a project and click Channels → New Channel.
2. Enter a name and choose a channel type:
   • Standard — open to all project members
   • Private — invite-only
3. Optionally attach a knowledge collection to pre-load context.
4. Click Create.

---

Knowledge Base Management

Collections

Knowledge is organized into collections. Each collection can be attached to one or more channels.

To create a collection:

1. Navigate to Settings → Knowledge → New Collection.
2. Enter a name and optional description.
3. Set access: Tenant-wide or Project-scoped.

Adding Documents

Supported formats: PDF, DOCX, TXT, Markdown, HTML, CSV.

1. Open a collection and click Upload.
2. Select one or more files (max 50 MB per file).
3. Documents are chunked and indexed automatically. Status shows Processing → Ready.

Web Crawl

To crawl a website into a collection:

1. Open a collection and click Add Source → Web Crawl.
2. Enter the starting URL and set crawl depth (1–3 levels recommended).
3. Click Start Crawl.

Crawl jobs run in the background. Estimated time is shown in the job status panel.

Removing Documents

1. Open the collection.
2. Click the document, then Remove from Collection.

Removal takes effect immediately for new conversations. Existing conversations that cited the document are not retroactively updated.

---

Policies and Guardrails

Policies govern what content the AI may produce or relay. Guardrails enforce policies at specific points in the pipeline.

Guardrail Gates

Gate	When it runs	What it does
G0 — Input check	Every inbound message, before search or reasoning	Blocks jailbreaks and policy violations
G1 — Evidence filter	After knowledge search, before prompt assembly	Drops or redacts retrieved content matching guardrail rules
G2 — Prompt check	After prompt assembly (roadmap — not yet active)	Final check on assembled context before the LLM call
G3 — Output check	After the LLM response, before delivery to user	Blocks or redacts guardrail violations in the response

Configuring Policies

Requires the tenant_admin role (or higher).

1. Navigate to Settings → Policies.
2. Click New Policy or edit an existing one.
3. Define conditions (keywords, topic categories, content types) and the action (block, warn, redact, log-only).
4. Assign the policy to a gate and optionally scope it to a project or channel.
5. Click Save and Activate.

Policy changes take effect for new conversations immediately. Use Test Policy to validate a rule against sample inputs before activating.

Reviewing Policy Blocks

When a message is blocked by policy, it appears in Settings → Governance → Policy Blocks. Each entry shows the message (redacted if configured), the matching rule, and the user.

---

Usage Governance

Token Quotas

Tokens are the unit of AI consumption (swTokens). Quotas can be set per user or per project.

1. Go to Settings → Usage → Quotas.
2. Set a monthly limit per user, per project, or tenant-wide.
3. When a user hits their quota, new AI requests are blocked until the quota resets or an admin raises it.

Rate Limits

Rate limits cap the number of requests per minute from a single user.

Defaults are set by your plan. To override for specific users or projects, go to Settings → Usage → Rate Limits.

Usage Dashboard

Settings → Usage → Dashboard shows:

• Token consumption by user, project, and channel
• Request volume over time
• Top users by consumption
• Blocked requests (policy and quota)

Export data as CSV from the dashboard for external billing or reporting.

---

Plan and Entitlements

Every plan includes the full feature set; plans differ by capacity. Your plan determines the limits available across your tenant.

Capacity	Solo	Professional	Business	Business Plus	Enterprise
Users	1	Unlimited	Unlimited	Unlimited	Custom
Agents	1	5	15	25	Custom
Knowledge collections	3	10	20	25	Custom
Guardrails	1	5	15	25	Custom
Monthly tokens	1,000,000	1,750,000	2,250,000	2,750,000	Custom

White-labeling and a custom domain are available on Business Plus and Enterprise. See Plans & Pricing for prices and the full comparison.

To view your current plan and usage against limits, go to Settings → Plan.

To upgrade, click Upgrade Plan and follow the checkout flow, or contact your account manager for Enterprise pricing.

---

Audit Log

The audit log records all administrative actions across the tenant.

1. Navigate to Settings → Audit Log.
2. Filter by date range, action type, or user.
3. Each entry shows: timestamp, actor, action, target, and outcome.

Audit log entries cannot be deleted by tenant administrators. Retention is governed by your contract — contact your account team for the duration applicable to your plan.

Exportable actions include: role assignments, user invitations/deactivations, policy changes, knowledge uploads/removals, and settings changes.

To export: set your filter, then click Export CSV.

---

Settings Reference

Setting	Location	Description
Tenant name and logo	Settings → General	Brand settings shown to users
Single Sign-On (SSO)	Settings → Authentication	SAML/OIDC configuration for your IdP
Data retention	Settings → Data	Conversation retention period (default 90 days)
Session timeout	Settings → Security	Idle session expiration
Allowed email domains	Settings → Users	Restrict invitations to specific domains
Webhook endpoints	Settings → Integrations	Receive event notifications via webhook
API keys	Settings → Integrations	Issue keys for the Agent and Policy APIs

---

Common Admin Tasks

GDPR Erasure Request

To permanently delete all data for a specific user, use the erasure endpoint:

DELETE /v1/admin/users/:userId/data

Requires tenant_admin role. Returns a 202 Accepted with a per-table deletion summary. Note: Frontegg identity records must be removed separately via the Frontegg admin console — see the GDPR runbook for the full procedure.

Investigate a Policy Block

1. Open Settings → Governance → Policy Blocks.
2. Find the event by user or timestamp.
3. Click the entry to see the matched rule, gate, and (if not redacted) the flagged content.
4. If the block was a false positive, edit the policy rule and click Re-evaluate to re-process the blocked message.

---

Getting Help

• In-app: click the ? icon in the lower left corner.
• Documentation: you are here.
• Support: contact your account manager or file a ticket at support.superwise.ai.

Enterprise customers have access to a dedicated Customer Success Manager and SLA-backed support response times.