Admin Controls for Chat Environment

This tutorial covers the administrative controls available for managing the Keeptrusts chat workbench. As an admin, you can manage user access, configure model availability, set policy overrides, define usage quotas, monitor active sessions, and review flagged content.

Use this page when

• You need to manage user roles, model availability, or usage quotas for the chat workbench.
• You are setting up policy overrides, session monitoring, or a content review queue.
• You want to generate admin reports on chat usage and policy triggers.

Primary audience

• Primary: Organization administrators and chat admins
• Secondary: Technical Leaders (governance oversight), Technical Engineers (integration)

Prerequisites

• Keeptrusts account with admin or organization-admin role
• Access to the Keeptrusts management console
• At least one gateway configured with chat-enabled models
• Familiarity with the Keeptrusts console

Step 1: Access Chat Administration

1. Log in to the Keeptrusts management console.
2. Navigate to Settings in the sidebar.
3. Select the Chat section.

The chat administration panel displays controls organized into tabs: Users, Models, Policies, Quotas, and Monitoring.

Step 2: Manage User Access

Control which users can access the chat workbench and what capabilities they have.

User Roles for Chat

Role	Capabilities
Chat User	Send messages, view responses, export conversations
Chat Power User	All Chat User capabilities plus multi-model comparison, knowledge base binding
Chat Admin	All capabilities plus user management, session monitoring, content review

Assign Chat Roles

1. In the Users tab, search for a user by name or email.
2. Click the user row to open their profile.
3. In the Chat Access section, select the appropriate role from the dropdown.
4. Click Save.

Bulk Role Assignment

For teams:

1. Navigate to the Teams section in the console.
2. Select a team.
3. In the team settings, set the Default Chat Role for all team members.
4. Individual overrides take precedence over team defaults.

Step 3: Configure Model Availability

Control which models are available to chat users.

Enable or Disable Models

1. In the Models tab, view the list of all configured model providers.
2. Toggle the Chat Enabled switch for each model to control its availability.
3. Disabled models do not appear in the chat workbench model selector.

Set Model Visibility by Team

To restrict certain models to specific teams:

1. Click a model row to open its configuration.
2. In the Team Access section, select which teams can use this model.
3. If no teams are selected, the model is available to all chat users.

Set a Default Model

1. In the Models tab, click the star icon next to a model to set it as the default.
2. The default model is pre-selected when users open the chat workbench.
3. Users can switch to other available models during their session.

Model availability is also governed by gateway configuration. A model must be both gateway-configured and chat-enabled to appear in the workbench.

Step 4: Configure Policy Overrides

Apply chat-specific policy overrides that layer on top of your gateway policies.

Chat-Specific Policy Settings

Setting	Description
Max response length	Maximum output tokens per response (prevents runaway generation)
Content filter level	Additional content filtering beyond gateway policies (off, low, medium, high)
Code execution block	Prevent models from generating executable code blocks
PII redaction enforcement	Force PII redaction on all chat responses regardless of gateway policy
Grounding requirement	Require responses to cite knowledge base sources when available

Apply an Override

1. In the Policies tab, locate the setting you want to configure.
2. Set the desired value.
3. Choose the scope: All Users, Specific Teams, or Specific Roles.
4. Click Save.

Policy overrides are additive — they cannot relax gateway policies, only add stricter controls.

Policy overrides take effect immediately for all new messages. Active conversations are not retroactively affected, but the next message in any active conversation will use the updated policies.

Step 5: Define Usage Quotas

Set limits on chat usage to manage costs and resource consumption.

Quota Types

Quota	Scope	Unit
Messages per day	Per user	Number of messages
Tokens per day	Per user	Total tokens (input + output)
Messages per hour	Per user	Burst rate limiting
Concurrent sessions	Per user	Active conversation count
Monthly spend cap	Per team	Currency amount

Configure a Quota

1. In the Quotas tab, click Add Quota.
2. Select the quota type from the dropdown.
3. Set the limit value.
4. Choose the scope (all users, specific teams, or specific roles).
5. Set the action when the quota is exceeded: Warn, Soft Block (allow with admin override), or Hard Block.
6. Click Save.

Quota Exceeded Behavior

When a user hits a quota limit:

• Warn — A warning banner appears but the user can continue.
• Soft Block — The input field is disabled with a message to contact their admin. Admins can grant a temporary exception.
• Hard Block — The input field is disabled. No exceptions until the quota resets at the configured interval.

Step 6: Monitor Active Sessions

Track real-time chat activity across your organization.

Session Dashboard

1. In the Monitoring tab, view the active sessions list.
2. Each row shows: username, team, model, session start time, message count, and current cost.
3. The dashboard auto-refreshes every 30 seconds.

Session Metrics

Metric	Description
Active sessions	Number of users currently in chat conversations
Messages/minute	Organization-wide message throughput
Average response time	Mean model response latency
Policy trigger rate	Percentage of messages triggering a policy action
Top models	Most frequently selected models in active sessions

Session Details

Click a session row to view:

• Full message history (subject to content review permissions)
• Per-message policy evaluations
• Token consumption and cost breakdown
• Model switches within the session

Step 7: Review Flagged Content

When policies flag content for review, it appears in the admin review queue.

Access the Review Queue

1. In the Monitoring tab, click Flagged Content.
2. The queue shows messages that triggered escalation-level policies.

Review Actions

For each flagged item, you can:

Action	Effect
Approve	Marks the content as reviewed — no further action
Redact	Applies retroactive redaction to the stored conversation
Block User	Temporarily suspends the user's chat access
Escalate	Forwards the item to a higher-level reviewer or security team
Dismiss	Removes the item from the queue without action

Review Workflow

1. Click a flagged item to open the full context (surrounding messages).
2. Review the policy that triggered the flag and the matched content.
3. Select an action from the action bar.
4. Add an optional review note for audit purposes.
5. Click Submit.

All review actions are logged in the audit trail for compliance.

Step 8: Export Admin Reports

Generate reports on chat usage and administration activity:

1. In the chat administration panel, click Export Report.
2. Select the report type: Usage Summary, Policy Triggers, Quota Utilization, or Review Actions.
3. Set the date range.
4. Choose the format (CSV or JSON).
5. Click Generate.

Reports are queued as export jobs. Check the Exports section in the console for download links.

Troubleshooting

Issue	Cause	Fix
Users cannot see the chat workbench	No chat role assigned	Assign at least the Chat User role
Model not appearing for a team	Model not chat-enabled or team not in access list	Enable the model and add the team to its access list
Quota not resetting	Quota interval misconfigured	Check the reset interval (daily, hourly, monthly)
Review queue not populating	No escalation-level policies configured	Configure at least one policy with an escalation action

Summary

You have configured the Keeptrusts chat environment as an administrator:

• Managed user access and role assignments
• Controlled model availability and team-level restrictions
• Applied policy overrides for additional governance
• Defined usage quotas with appropriate enforcement actions
• Monitored active sessions and reviewed flagged content
• Generated admin reports for compliance and cost tracking

For AI systems

• Canonical terms: Keeptrusts console, chat administration, Chat User role, Chat Power User role, Chat Admin role, model availability, policy override, usage quota, session monitoring, content review queue, flagged content.
• Config names: Chat Enabled toggle, Team Access list, Max response length, Content filter level, PII redaction enforcement, Messages per day quota, Monthly spend cap, Soft Block, Hard Block.
• Best next pages: Chat Analytics, Chat Cost Tracking, Team Conversations.

For engineers

• Prerequisites: Keeptrusts account with admin or org-admin role; at least one gateway with chat-enabled models.
• Validation: Assign a Chat User role → verify the user sees the workbench. Disable a model → confirm it disappears from the model selector. Set a quota → verify enforcement when the limit is reached.
• Key setting: Policy overrides are additive and cannot relax gateway policies — only add stricter controls.

For leaders

• Quotas and role-based access provide cost containment and risk controls without blocking productive AI use.
• Content review queues satisfy compliance requirements for human oversight of AI interactions (relevant to EU AI Act Article 14).
• All admin actions are audit-logged — support ownership is traceable.
• Rollout: Start with Standard quotas and Warn actions, then tighten based on analytics.

Next steps

• Tutorial: Chat Analytics & Usage Metrics — monitor the metrics controlled by these admin settings.
• Tutorial: Tracking Chat Costs — understand per-user cost tracking tied to quotas.
• Tutorial: Team Chat Environments — configure team-level access and policies.
• Tutorial: Chat Troubleshooting — resolve issues reported by managed users.