Documentation Index

Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

​

Insights & settings

Studio provides analytics dashboards for understanding agent performance and usage trends, guardrail configuration for content safety and compliance, and project-level settings for access control, model configuration, and runtime behavior.

​

Insights & analytics

Insights provide an executive-level dashboard for understanding your AI agent program’s performance, usage trends, and cost profile. Navigate to the Insights group in the project sidebar to access analytics pages.

​

Insights dashboard

The main Dashboard page displays key performance indicators, trend charts, and cost breakdowns. KPI metric cards — five metric cards across the top of the dashboard show headline numbers: Each metric card includes a trend indicator showing the percentage change compared to the previous period (up arrow for increase, down arrow for decrease). Date range selector — a dropdown in the top-right corner lets you select the analysis period: Last 7 days, Last 30 days (default), or Last 90 days. Changing the date range refreshes all dashboard data. Volume and containment trend chart — an area chart below the KPI cards shows daily conversation volume and containment rate over the selected period. Hover over any point on the chart to see exact values for that day. Cost breakdown table — below the trend chart, a table groups spending by agent showing agent name, sessions handled, tokens consumed, and estimated cost. This helps identify which agents drive the most resource consumption.

​

Additional insights pages

The Insights group includes several specialized analytics pages: Agent performance — per-agent diagnostic metrics including response times, completion rates, and error frequencies. Helps identify agents that may need optimization. Quality monitor — monitors agent response quality through automated scoring and anomaly detection. Tracks quality trends over time. Customer insights — analyzes user behavior patterns including intent distribution, sentiment trends, and voice-of-customer signals. Voice analytics — aggregated metrics for voice channel interactions: call volume and duration trends, speech-to-text accuracy, voice-specific quality indicators, and latency measurements for the voice processing pipeline.

Note: Agent Performance, Quality Monitor, and Customer Insights may show a “coming soon” placeholder if they are not yet enabled for your workspace.

​

Working with date ranges

All insights pages respect the selected date range:

• Use 7 days for operational monitoring and quick health checks.
• Use 30 days for monthly reviews and reporting.
• Use 90 days for trend analysis and strategic planning.

Tip: Compare 30-day periods before and after agent changes to measure the impact of your improvements.

​

Guardrails configuration

Guardrails protect your agents and users by applying content safety policies, input/output filtering, and PII protection rules. Navigate to Govern > Guardrails from the project sidebar to manage guardrail policies.

​

Guardrails page layout

The guardrails configuration page is organized into three tabs:

​

Policies

Policies are the core building blocks of guardrail configuration. Each policy defines a set of rules that are evaluated against agent inputs, outputs, or both. Creating a policy:

1. Select the Policies tab.
2. Click Create Policy.
3. Fill in the policy form:
   • Name — a descriptive label (e.g., “Content Safety,” “PII Filter,” “Competitor Mentions”).
   • Description — explains what the policy protects against.
   • Scope — whether the policy applies to user inputs, agent outputs, or both.
   • Rules — one or more evaluation rules with severity levels.

Each rule within a policy specifies a rule type (the category of content being evaluated), a severity level, and an action (block, warn, or log). Severity levels: From the policies list you can toggle policies active/inactive, edit rules, or delete policies. For advanced configurations, click the YAML editor toggle to define policies in YAML format — useful for complex rule combinations, importing policies, or bulk editing.

Tip: Start with policies in “Low” severity (log-only) mode to understand what they catch before enabling blocking behavior.

​

Content filtering

Content filtering policies evaluate message content against predefined categories:

• Harmful content — violence, self-harm, hate speech.
• Inappropriate content — adult content, profanity.
• Misinformation — verifiably false claims, misleading information.
• Off-topic content — responses that deviate from the agent’s intended purpose.

​

PII protection

PII (Personally Identifiable Information) protection policies detect and handle sensitive data:

• Detection patterns — identify email addresses, phone numbers, social security numbers, credit card numbers, and other PII types.
• Actions — mask, redact, or block PII in agent inputs and outputs.
• Custom patterns — define organization-specific PII patterns.

PII protection can also be configured at the project level through Settings > PII Protection (see Project settings below).

​

Policy templates

Studio provides built-in templates for common guardrail configurations:

• Basic safety — blocks harmful content and explicit material.
• Enterprise compliance — includes PII protection, competitor mention filtering, and financial advice disclaimers.
• Healthcare — adds medical advice disclaimers and HIPAA-related protections.

Use templates as a starting point and customize them for your specific requirements.

​

Providers

The Providers tab manages workspace-level guardrail provider configurations. Providers are external services that perform content evaluation (e.g., content moderation APIs). To add a provider, configure a name, provider type, credentials, and provider-specific settings.

​

Audit tab

The Audit tab provides a history of guardrail evaluations, showing when policies were triggered, what content was affected, and what action was taken. Use the audit log to verify that policies work as intended, identify patterns in violations, and fine-tune severity levels and rules.

​

Project settings

Project settings control access, configuration, and behavior for your project. Navigate to the Settings group in the project sidebar to access settings pages.

​

Members and permissions

Settings > Members — manage who has access to your project and what they can do. The members page displays a table of all project members with name, email, role, and added date. To add a new project member, select an existing workspace member, choose a role, and save. Project membership is sourced from workspace members; it does not send a separate invitation email. When a workspace member creates a project, they become that project’s owner and can add existing workspace members from this screen. Only workspace Owners and Admins can invite brand-new people into the workspace itself. Project roles: Click the remove action next to a member to revoke their access immediately.

​

API keys

Settings > API Keys — manage API keys for programmatic access to the project runtime. Create keys with descriptive names and optional expiration dates, view active keys with usage metadata, and revoke keys that are no longer needed. See Testing, deployment & operations for more details on API key usage.

​

Model configuration

Settings > Models — configure which LLM models are available for agents in this project. View the catalog of available models across all configured providers, enable or disable specific models, and set a default model for new agents. Each model entry shows the model name and provider, supported capabilities (chat, function calling, vision, streaming), token limits, and pricing information.

​

Configuration variables

Settings > Config Variables — define key-value pairs that are available to agents and tools at runtime. Config variables are useful for storing API endpoints used by multiple tools, feature flags, and business-specific values (support hours, company name, etc.) referenced in agent personas. Variables can be overridden per deployment environment (see Testing, deployment & operations).

​

Git integration

Settings > Git — connect your project to a Git repository for version-controlled agent definitions. To set up Git integration:

1. Configure the repository URL, branch, and authentication (SSH key or token).
2. Test the connection.
3. Enable synchronization.

When Git integration is enabled, agent definitions can be pushed to the repository, changes can be pulled into Studio, and conflict resolution tools help manage simultaneous edits.

​

Runtime configuration

Settings > Runtime Config — configure runtime behavior for the project’s agent execution environment, including session timeout, max concurrent sessions, rate limits, and logging level.

​

Trace dimensions

Settings > Trace Dimensions — configure custom dimensions for trace data that appear in session traces and analytics. Define custom metadata fields captured with every trace event (dimension name, source, and data type). Custom dimensions enrich trace data, enabling more granular filtering and analysis in the session browser and insights dashboard.

​

Agent transfer

Settings > Agent Transfer — configure how agent-to-human transfers work in your project, including the default contact center provider, queue configuration, context passing rules, and post-transfer behavior. See Testing, deployment & operations for monitoring active transfers.

​

PII protection

Settings > PII Protection — configure project-level PII detection and handling rules. Built-in patterns — pre-configured patterns for common PII types (email, phone, SSN, credit card). Custom patterns — define organization-specific patterns using regular expressions. For each pattern, configure a name, detection regex, action (mask, redact, or flag), and scope (user inputs, agent outputs, or both). Test patterns against sample text before saving.

​

Advanced settings

Settings > Advanced — additional project-level configuration including project deletion, feature flags, webhook configuration, and data retention policies.

Warning: Changes to advanced settings can significantly affect project behavior. Review changes carefully before saving.

​

Related pages

• Studio overview — navigation and project management
• Agent development — building agents with guardrails and configuration
• Tools, knowledge & connections — managing agent resources
• Testing, deployment & operations — evaluations, deployment, and session monitoring