CoreAI Portal User Guide

Version: 0.15.5

Introduction

The CoreAI Portal is a web-based platform that provides access to artificial intelligence capabilities through a user-friendly interface.

The portal serves as a centralized access point for various AI functionalities provided by our AI Platform, designed to be accessible to users regardless of their technical background.

This documentation will guide you through the portal's features and capabilities and how to use them.

Core Functionality

The CoreAI Portal web UI provides the following key capabilities:

• GenAI Chatbot: Interact with Generative AI models through familiar conversational interface to ask questions, request analysis, or obtain assistance with various tasks
• Library: Upload and manage documents, images, and other files that can be processed and analyzed by the AI Platform components
• Knowledge Base: Organize and maintain the vectorized collections that the AI Platform uses to provide accurate and relevant responses in the Gen AI Chatbot interface
• Models Management: Access different AI models optimized for specific tasks such as text generation, data analysis, or document processing
• Model Repository: Own and control the models you want to use within your organization
• Applications and Services: Monitor the operational status of platform applications and services.
• Audit: Review logs of all actions performed within the portal by its users for accountability and compliance purposes
• Settings: Customize portal settings and preferences to meet organizational requirements and user needs, modify identity and access management settings.

---

Dashboard

By using your credentials, after a successful login, you'll be directed to the main dashboard.

The dashboard serves as the central hub where you can access various features and functionalities of the web application. Here's an overview of what you can expect:

• Navigation Menu: Located on the left-hand side, this menu provides quick access to different sections of the application. The navigation menu will be present on all the other portal pages. You can use the Icon at the top-left side to hide or show this sidebar anytime during using the portal.
• Statistics Overview: On the landing page some metrics will be displayed at the top of the page about the platform and models usage
• Quick Actions: Navigate to the most common used actions within the portal with one click

---

Chatbot

The Chatbot provides access to Generative AI capabilities.

Initial Page

To begin a new conversation, follow these steps in the order presented:

1. Model Selection: Choose from the available AI models based on your specific requirements. If not models are available in the list, please ask your Platform administrator to deploy a model.
2. Knowledge Base Integration: Enable "Use your Knowledge Base to answer questions" if you wish to incorporate data from your organization's knowledge base collections
3. Agent Mode Configuration: Activate advanced agentic capabilities by selecting the agent icon in the chat input area to access specialized tools such as: web search, web scrape or text to SQL
4. Message: Enter your message in the text field and press Enter to submit

Alternatively, you can press "View Previous Conversations" to access your conversation history and continue an already existing discussions.

Chatbot Layout

The main chat interface is organized into the following functional areas:

• Left Panel - Conversation History: Access and select from your previous chat sessions
• Top Section - Model Selector: Choose the AI model for new conversations or view the currently active model for existing sessions. You can only select a model if you pressed the "+ New Chat" button and did not enter any text in the message field at the bottom.
• Configuration Area: Located below the model selector, this section allows you to enable Retrieval-Augmented Generation (RAG) and select knowledge collections, or view current settings for existing conversations
• Bottom Section - Message Input: The primary text input area where you compose and send messages for both new and existing conversations
• Advanced Features: The agent icon within the message input area provides access to specialized tools and enhanced capabilities for your conversations

---

Library

The Library provides file management capabilities for organizing, storing, your documents and media files.

Upon accessing the Library, you will be presented with a file management interface that allows you to upload, organize, and manage your data.

Interface Layout

The main Library interface is organized into the following functional areas:

• Top Section - Statistics Dashboard: Displays key metrics including total files, storage used, shared files, and recent downloads
• Main Content Area - File Management: Contains two primary tabs for browsing existing files and uploading new ones
• Left Panel - Folder Tree: Hierarchical navigation structure for organizing files into folders and subfolders
• Right Panel - Content Area: Dynamic content area that changes based on the selected tab (Browse or Upload)
• Toolbar - Search and Filters: Advanced search, file type filtering, sorting options, and view mode controls
• Action Controls: Multi-select operations, file operations (download, delete, view details), and folder management

File Management Features

Browse Files Tab

The Browse Files tab provides all file management functionalities:

• Folder Navigation: Use the left sidebar tree structure to navigate between folders and subfolders
• Search Functionality: Search files by name, tags, or description using the search bar
• File Type Filtering: Filter files by type (Images, Videos, Audio, PDFs, Text files, or All Types)
• Sorting Options: Sort files by creation date (newest/oldest first), name (A-Z/Z-A), or file size (largest/smallest first)
• View Modes: Switch between grid view (card layout) and list view (detailed table layout)
• Multi-Selection: Select multiple files for batch operations (download, delete)
• File Operations: Individual file actions including download, view details, and delete
• Folder Operations: Create, rename, delete, and organize folders with drag-and-drop functionality

Upload Files Tab

The Upload Files tab allows user to upload more files in library with the following functionalities:

• Drag and Drop Interface: Drag files from your computer into the upload area
• Folder Selection: Choose the destination folder from the folder tree before uploading, otherwise the default user "/root" path will be used
• Batch Upload: Upload multiple files simultaneously with progress tracking
• File Metadata: Add tags and descriptions to files during upload for better organization
• Folder Creation: Create new folders during the upload process
• File Validation: Automatic validation of file types and size limits
• Upload Progress: Real-time progress tracking for each file being uploaded

Advanced Features

File Organization

• Hierarchical Folders: Create unlimited folder levels for complex organization structures
• Drag and Drop: Move files and folders between locations using intuitive drag-and-drop
• Favorites System: Mark frequently used folders as favorites for quick access
• Tagging System: Add descriptive tags to files for enhanced searchability
• File Descriptions: Add detailed descriptions to provide context for your files

File Type Support

The Library supports a wide range of file types including:

• Images: All common image formats (JPEG, PNG, GIF, WebP, etc.)
• Videos: Video files in various formats (MP4, AVI, MOV, etc.)
• Audio: Audio files (MP3, WAV, FLAC, etc.)
• Documents: PDF files and Microsoft Office documents
• Text Files: Plain text and formatted text documents
• Archives: Compressed files (ZIP, RAR, etc.)
• Data Files: JSON and other structured data formats

Storage and Limits

• File Size Limits: Individual files up to 1GB (configurable by the Platform IT Administrator)
• Storage Monitoring: Real-time tracking of total storage used by the logged in user
• File Count Tracking: Monitor total number of files in your library
• Download Tracking: Track file access and download statistics

---

Knowledge Base Management

The Knowledge Base module provides access to your organization's vectorized data collections stored in the system.

These collections contain processed information that can be utilized during AI chat conversations to provide contextually relevant responses based on your specific data.

Using collections and vectorized data requires AI Engineering expertise for proper usage and configuration.

Creating New Collections

To create a new vector collection, select "Create Collection" and provide the following required information:

• Collection Name: A unique identifier for your data collection
• Embedding Model: The AI model used to process and vectorize your data. The system will automatically configure the appropriate vector dimensions based on the selected model
• Description: An optional field to document the purpose and contents of the collection

Collection Management

You can access detailed information about any collection by selecting it from the overview table. This action will open a management interface that provides the following capabilities:

• Data Review: Examine the information currently stored within the collection
• Collection Removal: Delete the entire collection when it is no longer needed
• Document Upload: Add new PDF or TXT files to expand the collection's knowledge base
• Content Search: Query the existing data within the collection to locate specific information

---

Models Management

Models Overview

This page displays all available AI models, including both locally deployed models and configured cloud-based models, along with relevant information about each. The interface provides several management tools:

• Manual Refresh: Use the refresh function at the top of the table to immediately check for newly available models
• Search Functionality: Enter a specific model name or pattern in the search field to filter and locate models of interest
• Provider Filter: Use the "All Providers" dropdown to switch between viewing local or cloud-based models
• Model Deployment: Select the green "Model Deployment" button to access the model installation interface for deploying new models

The table below the control panel displays the complete list of models based on your selected filters.

Selecting any model from the list will redirect you to a dedicated model page where you can access detailed information about that specific deployed model.

Model Details

The model details page provides two primary views: Model Info and Model Logs. Both views share a common header displaying essential information including the model name, availability status, deployment location (on-premises or cloud), and a delete function.

Model Information

The "Model Info" tab displays specific technical information about the model deployment, including:

• Model deployment status and configuration
• Timeout and retry strategy settings
• Token processing limits
• Model capabilities and features
• Custom tags and metadata
• Supported parameters and configurations

Model Logs

The "Model Logs" tab provides access to detailed system logs for the specific model, sourced directly from the underlying Kubernetes infrastructure. This section is particularly valuable for technical personnel who need to troubleshoot model deployments without requiring direct platform infrastructure access.

This page is designed to provide a comprehensive overview of model health status, visualize all associated pods and deployment states, and present relevant logs for system monitoring and troubleshooting.

Model Repository

The Model Repository page provides access to locally available models that are ready for deployment. This feature enables organizations to utilize models from their own secure repositories rather than downloading potentially untrusted models from external sources. This functionality is primarily designed for AI Engineers and Platform Engineers who are responsible for managing the complete model lifecycle.

The page header includes access to the Model Downloader interface via the green button, which allows you to add new models to the repository by completing the required configuration form.

Selecting any model from the repository table will display detailed information about that specific model.

Model Repository Details

The model repository details page allows you to review comprehensive information about a model prior to deployment. This includes:

• Basic Information: Model name and version details
• Version History: Complete versioning records maintained in the MLFlow system
• Custom Tags: Metadata and classification tags associated with the model
• Storage Location: Direct S3 storage URL for accessing the latest model version

The page header features a "Deploy Model" button that automatically populates the deployment form with the selected model's configuration, streamlining the deployment process.

Model Installer

The Model Installer provides two deployment approaches: Easy Setup and Advanced Setup. Users can select the appropriate option based on their experience level with model deployment. Both approaches offer full customization capabilities at later stages of the deployment process.

Easy Setup

The Easy Setup interface enables users to deploy pre-configured models using a simplified one-click deployment strategy. This option utilizes predefined model configurations to streamline the deployment process.

If no model presets are available, contact your portal administrator to configure them through the "Settings - Model Presets" page.

The model presets interface includes two filtering options:

• Text Search: Filter models by name using the text input field
• Type Filter: Select models by category using the dropdown menu

Even when using presets, users can customize the deployment configuration by selecting the "Customize Deployment" button at the top of the page to modify default values as needed.

Advanced Setup

The Advanced Setup provides a comprehensive multi-step configuration process that guides users through all required settings and parameters for complete model deployment and installation.

Model Configuration

The model information step ensures proper configuration of the target model deployment, including:

• Source Location: Specify the download URL from public repositories (Ollama, Hugging Face) or internal storage systems
• Hardware Allocation: Configure CPU and GPU resource allocation
• Model Type: Define the model's purpose (text completion, embedding, etc.)

For models stored in the internal repository, switch to the "From Repository" tab and select the desired model from the dropdown menu instead of manually entering a download URL.

Deployment Configuration

The deployment step focuses on infrastructure-level scaling configuration and hardware resource prioritization for the model deployment.

Advanced Configuration

The advanced configuration section contains engine-specific parameters and values that require specialized knowledge in AI Engineering. This step allows configuration of specific arguments and environment variables based on the selected deployment engine.

Consult the documentation for supported engines (Ollama, VLLM, FastWhisper, or Infinity) to understand the required arguments and environment variables for your specific model. The deployment can be initiated from this step using the Deploy button.

Model Downloader

The Model Downloader enables users to add new models to the local model repository through the MLFlow system. This functionality can be accessed from the Model Repository page as previously described.

The Model Downloader interface provides a configuration form with the following required fields:

• Model Name: Specify the model identifier in the format "organization/model" from Hugging Face
• MLflow Experiment ID: The system generates a random identifier automatically, though a custom ID can be specified if required
• MLflow Artifact Path: Defaults to "models" directory - modify only if a different storage path is necessary
• Revision: Specify the Git revision or version tag to be downloaded

Once the download process is initiated, users can monitor the progress through the Download Logs interface, which displays real-time information as the model is retrieved from Hugging Face and transferred to the local storage system.

---

Applications

The Applications page provides monitoring and management capabilities for all applications deployed in your AI Platform environment.

Upon accessing the Applications page, you will be presented with a centralized dashboard that displays the health status, availability, and operational metrics of all platform components organized by functional categories.

This page is intended to be used by System or Platform administrators.

Overview

1. Dashboard Overview: Review the comprehensive metrics dashboard showing total applications, online/offline status, health status, and synchronization status
2. Application Categories: Navigate through organized categories including Security, Integration, Storage, and Inference applications
3. Health Monitoring: Monitor real-time health status, sync status, and service availability for each application
4. Access: Click on application cards to access running services through their ingress links

Application Categories

Security, Monitoring and Logging

This category includes essential security and monitoring applications:

• Keycloak: Identity and access management system for user authentication and authorization
• Grafana: Monitoring and observability platform for system metrics and dashboards

Integration and Delivery (CICD)

The CICD category includes workflow and deployment management tools:

• Argo Workflows: Workflow engine for running compute-intensive jobs and machine learning pipelines
• Argo CD: Continuous deployment tool for managing application deployments and configurations

Storage and Database

The Storage and Database category includes data storage and database management systems:

• MinIO: Object storage solution for storing files, datasets, and model artifacts
• PGAdmin: PostgreSQL database administration and management interface

Inference

The Inference category includes AI/ML model serving and relevant systems:

• CoreAI LLM Backend: Large language model backend service for AI inference developed by AI Platform
• KubeAI: Kubernetes-native AI platform for model deployment and management
• Langfuse: LLM observability and analytics platform for monitoring AI applications
• LiteLLM: Unified API for accessing various LLM providers and models
• MLFlow: Machine learning lifecycle management and experiment tracking
• Model Installer: Tool for installing and managing AI/ML models developed by AI Platform

---

Services

The Services page provides monitoring and management capabilities for all Kubernetes services and applications deployed in your AI Platform environment.

Upon accessing the Services page, you will be presented with a centralized dashboard that displays cluster-wide metrics, service discovery information, and ArgoCD application management tools organized into two main functional areas.

Overview

1. Cluster Monitoring: Review the metrics dashboard showing resource status, namespace health, and service availability across the entire cluster
2. Service Discovery: Navigate through discovered Kubernetes services organized by namespace with detailed service information and health status
3. ArgoCD Management: Monitor and manage ArgoCD applications with sync and refresh operations for deployment management
4. Real-time Monitoring: Enable auto-refresh for continuous monitoring of service health and application status

Services Interface Layout

The main Services interface is organized into the following functional areas:

• Top Section - Cluster Monitoring Dashboard: Displays cluster-wide metrics including resource status, resource usage, namespace health, and service health statistics
• Main Content Area - Service Management: Contains two primary tabs for service discovery and ArgoCD application management
• Service Discovery Tab: Grid-based view of Kubernetes services organized by namespace with filtering and search capabilities
• ArgoCD Tab: Table-based view of ArgoCD applications with sorting, health monitoring, and operational controls
• Search and Filter Controls: Advanced search functionality and status-based filtering for both services and applications
• Auto-refresh Controls: Configurable automatic refresh intervals for real-time monitoring

Cluster Monitoring Dashboard

Resource Status Monitoring

The cluster monitoring dashboard provides real-time insights into your Kubernetes cluster:

• Total Resources: Complete count of all pods across all namespaces
• Running : Number of pods currently running successfully (green indicator)
• Pending : Number of pods waiting to be scheduled or started (yellow indicator)
• Failed : Number of pods that have failed or are in error state (red indicator)

Resource Usage Analytics

• CPU Usage: Total CPU consumption across the cluster (when available)
• Memory Usage: Total memory consumption across the cluster (when available)
• Total Restarts: Monitoring of application stability through restart counts

Namespace Health Overview

• Total Namespaces: Count of all namespaces in the cluster
• Active Namespaces: Number of namespaces with all services in healthy state
• Total Services: Services per namespace and health distribution

Service Health Summary

• Healthy Services: Count of services responding normally (green indicator)
• Unhealthy Services: Count of services experiencing issues (red indicator)
• Services with Ingress: Number of services accessible via external ingress

Service Discovery Tab

Service Organization

The Service Discovery tab provides Kubernetes service management:

• Namespace-based Grouping: Services organized by Kubernetes namespaces for logical separation
• Service Grid Layout: Visual grid display of services with status indicators and metadata
• Service Cards: Individual service cards showing detailed information including:
  • Service name and namespace
  • Service type and cluster IP
  • Port mappings and protocols
  • Ingress host information
  • Health status indicators

Search and Filtering

• Namespace Search: Search services by namespace name with real-time filtering
• Status Filtering: Filter services by health status (All, Up, Down)
• Service Count Display: Real-time count of total, healthy, and unhealthy services
• Interactive Badges: Clickable status badges for quick filtering

Service Details

Each service card displays comprehensive information:

• Basic Information: Service name, namespace, type, and cluster IP
• Port Configuration: Detailed port mappings with target ports and protocols
• Ingress Access: External access URLs and ingress host information
• Health Status: Real-time health monitoring with visual indicators
• Service Metrics: Pod counts, restart information, and resource usage

Namespace Management

• Namespace Cards: Visual representation of each namespace with service counts
• Status Indicators: Color-coded status indicators (green for healthy, red for unhealthy, yellow for mixed)
• Service Counts: Up/down service counts per namespace
• Namespace Details: Age, status, and operational information for each namespace

ArgoCD Management Tab

Application Monitoring

The ArgoCD tab provides application lifecycle management:

• Application Table: Sortable table view of all ArgoCD applications
• Health Status: Real-time health monitoring of deployed applications
• Sync Status: Deployment synchronization status and revision tracking
• Application Age: Time since application deployment and last update

Sorting and Organization

• Name Sorting: Alphabetical sorting of applications by name (ascending/descending)
• Health Sorting: Sort by application health status for quick issue identification
• Sync Status Sorting: Sort by synchronization status to identify out-of-sync applications
• Age Sorting: Sort by application age for lifecycle management

Application Operations

• Individual Sync: Synchronize individual applications to their desired state
• Individual Refresh: Refresh application configuration and status
• Bulk Operations: Sync or refresh all applications within a namespace
• Operation Progress: Real-time feedback on operation status and completion

Application Details

Each application entry displays:

• Application Name: ArgoCD application identifier and display name
• Destination Information: Target server and namespace for deployment
• Health Status: Current health state (Healthy, Degraded, Unknown)
• Sync Status: Synchronization state (Synced, OutOfSync, Unknown)
• Application Age: Time since deployment in hours
• Operation Controls: Sync and refresh buttons with progress indicators

---

Audit

The Audit page provides tracking and analysis of all user activities and system events within the CoreAI Portal.

Upon accessing the Audit page, you will be presented with visual analytics and detailed activity logs that help you monitor portal user activities by using the tab navigator at the top of the page: "Charts" and "Activity".

Overview

1. Activity Analytics: Review visual charts showing login patterns, user activity trends, and page access statistics
2. User Activity Tracking: Monitor detailed logs of all user actions including logins, page visits, and system interactions

Charts

Login Statistics Chart

The login statistics chart provides insights into user authentication patterns:

• Time Period Selection: Choose between Today, Last 7 Days, or Last 30 Days for analysis
• Login Trends: Visual bar chart showing login frequency over the selected time period
• Total Logins: Aggregate count of all login attempts in the selected period
• Unique Users: Number of distinct users who logged in during the selected period

Page Access Statistics Chart

The page access chart shows which parts of the system are most frequently used:

• Page Popularity: Pie chart displaying the most accessed pages and features
• Access Counts: Numerical values showing how many times each page was visited
• Visual Legend: Color-coded legend with access counts for each page
• Interactive Elements: Hover over chart segments to see detailed information
• Usage Patterns: Understand which features are most popular among users

Activity Log Management

Activity Tracking

The activity log captures all significant user actions and system events:

• Timestamp Information: Exact date and time when each action occurred
• User Identification: User ID and display name for each logged action
• Group Membership: User groups and roles associated with each action
• Action Types: Categorized actions including logins, page access, and system operations
• Detailed Information: Additional context and details about each action

Action Categories

The system tracks various types of user activities:

• Login/Logout Events: User authentication and session management
• Page Access: Tracking which pages users visit and when
• Access Management: Granting and revoking user permissions
• Chat Interactions: User engagement with the chatbot system
• System Operations: Administrative actions and configuration changes

Advanced Filtering and Search

Multi-Dimensional Filtering

The audit system provides powerful filtering capabilities:

• User Filtering: Filter activities by specific users or user groups
• Action Filtering: Focus on specific types of activities (logins, page access, etc.)
• Time-Based Filtering: Search for activities within specific time ranges
• Group Filtering: Filter by user groups or organizational units
• Combined Filters: Use multiple filters simultaneously for precise analysis

Search Functionality

• Field-Specific Search: Search within user names, groups, or action types
• Timestamp Search: Find activities before, after, or within specific time intervals
• Real-Time Results: Search results update immediately as you type
• Reset Options: Clear all filters and searches to return to full view

Time-Based Analysis

• Before/After Search: Find activities that occurred before or after a specific time
• Interval Search: Search for activities within a specific time range
• Date Picker Integration: Easy-to-use calendar interface for time selection
• Flexible Time Formats: Support for various date and time input formats

Data Analysis

• Comprehensive Logging: All activities are permanently recorded for compliance
• Historical Analysis: Access historical data for trend analysis
• Compliance Support: Detailed audit trails for regulatory requirements
• Performance Monitoring: Track system usage patterns and user behavior

Real-Time Monitoring

• Live Updates: Audit logs are updated in real-time as activities occur
• Immediate Visibility: See user activities as they happen
• System Integration: Automatic logging of all system events
• Performance Tracking: Monitor system performance through user activity patterns

Customizable Views

• Flexible Filtering: Create custom views based on your specific needs
• Reporting Integration: Use audit data for compliance and security reporting

---

Settings

The Settings section is organized into multiple sub-sectons that focus on various portal and platform configuration options available to the administrators.

Identity and Access Management (IAM)

The Identity and Access Management subsection provides portal administrators with comprehensive user, role, and group management capabilities. The sidebar navigation within this section provides access to the following management areas:

• Platform User Management
• Platform Roles
• Portal Access Control

Platform User Management

The Platform User Management section enables administrators to perform comprehensive user administration tasks, including:

• Adding new local users to the Keycloak identity system
• Editing existing user account details and configurations
• Managing user group memberships and assignments within Keycloak

Platform Roles

This section provides management capabilities for Keycloak Realm Roles, allowing administrators to create, modify, or remove role definitions. To establish mappings between groups and realm roles specifically for portal access, use the "Portal Access Control" section. For additional group and user mappings beyond portal scope, direct Keycloak administration is required.

Portal Access Control

The Portal Access Control section enables administrators to configure role-based access permissions for portal groups, implementing the principle of least privilege across all portal pages and functionality.

Resource Profiles

Resource Profiles define standardized hardware configurations that streamline model deployment processes on the platform. Administrators can enable one or more resource profiles from the available options based on the specific hardware capabilities of their platform infrastructure.

Model Presets

The Model Presets section allows administrators and AI Engineers to configure default deployment settings for the "Easy Setup" model deployment method. This functionality reduces deployment time for repetitive configurations and provides simplified access for non-technical portal users.

Preset Management

The main page displays all existing model presets stored in the portal database. Administrators can view, modify, or remove existing presets as needed. To create a new preset, select "Create Model Preset" to access the preset configuration interface.

Preset Creation

The preset creation interface provides comprehensive configuration options similar to the Advanced Deployment mode. This page includes all required fields necessary to ensure users have complete deployment data available when utilizing the "Easy Setup" deployment interface.

Connectors

The Connectors section provides a comprehensive view of all active integrations between the portal and other platform components. This interface currently offers read-only access to monitor integration status and connectivity.