The screenshot displays a user profile icon featuring a stylized red panda wearing glasses, set against a dark circular background. This icon is likely used for user identification or account access within the HostedAI platform, providing a personalized touch to the user interface. The icon is positioned within a curved, grey interface element, suggesting it might be part of a sidebar or a navigation menu.

Overview

hosted·ai Documentation

Generating Article

Understanding GPU accelerated VM and GPUaaS Instances

hosted·ai

Admin & User Panel Installation

Installation

The screen is the login page for the HostedAI platform, designed to allow users to access their accounts. At the top, the HostedAI logo is displayed. Below the logo, the page prompts users to 'Login to your account' by entering their credentials. There are three input fields: 'Email ID' for entering the user's email address, 'Password' for entering the account password, and 'OTP' for entering a one-time password if multi-factor authentication is enabled. The password field includes an eye icon to toggle password visibility. Below the OTP field, there is a 'Forgot password?' link for users who need to reset their password. At the bottom, there are two buttons: a blue 'Login' button for submitting the entered credentials, and a 'Login through recovery keys' button for alternative access using recovery keys. The 'Login' button is centrally located, while the 'Login through recovery keys' button is positioned below it.

Signing In

The screenshot displays the HostedAI user interface, specifically the AI Models section within the My Teams panel. On the left, there is a navigation menu with icons for different sections, including a user icon, folder icon, and settings icon. Below the navigation menu, the My Teams section lists teams such as AI Models, ML Ops Team, and Agentic AI Team, with the AI Models team currently selected. The main panel shows details for the AI Models team, including the creation date and options to view members and workspaces. Tabs for Overview, Billing, Members, Workspaces, Instances, Accelerators, and Shared Storage are available, with the Overview tab currently active. Resource allocation details are displayed, including CPU, GPU, RAM, and Storage, all showing unlimited availability. The top right corner features a user profile section with the default admin user information, including an email address and a dropdown menu for account settings and signing out. The 'Sign out' button is located at the bottom of this dropdown menu.

Signing Out

Access

The screenshot displays the 'Instances' section of the HostedAI User Panel, where users can view and manage cloud instances. At the top, there is a search bar labeled 'Search for instance' for finding specific instances. Below, a table lists instances with columns for 'Instance ID & Name', 'Type', 'Public IP(s)', 'Region', 'Created On', 'Workspace/Team', 'OS', and 'Status'. Each row represents an instance with details such as name, type, and status, which can be 'Running' or 'Deploying'. On the right, there is a 'Create instance' button for adding new instances. Above the table, dropdown menus allow filtering by 'Teams/subteams' and 'All status'. The top right corner shows a user profile section with 'Default admin user' and a 'Create' button. A summary box indicates the total number of instances, with counts for 'Running' and 'Stopped'.

Create Instance (KVM)

The 'View all Instances' screen in HostedAI displays a list of all instances available to the user. At the top, there is a search bar labeled 'Search for instance' for filtering instances by name. To the right, dropdown menus labeled 'Teams/subteams' and 'All status' allow further filtering by team and instance status. A 'Create instance' button is prominently located on the right side. Below, a table lists instances with columns for 'Instance ID & Name', 'Type', 'Public IP(s)', 'Region', 'Created On', 'Workspace/Team', 'OS', and 'Status'. Each row provides details for an instance, including its name, type (e.g., Small, Medium), IP address, region, creation date, associated workspace or team, operating system, and current status, which is indicated as 'Running' with a green label. A sidebar on the left contains navigation icons for different sections of the platform. The top right corner shows the user profile with the name 'Default admin user' and role 'Administrator'. A summary box displays the total number of instances and their statuses: 4 Running and 0 Stopped.

Manage Instances (KVM)

Instance Details (KVM)

Instances (KVM)

Manage Teams

The screenshot displays the Teams/User Panel within the HostedAI platform, specifically focusing on the billing section for GPU-as-a-Service (GPUaaS). On the left side, there is a navigation panel listing teams such as AI Team Alpha, ML Ops Team Delta, Juggernaut Omega, and John Doe, with a search bar above to filter teams by name. Below the team list, there is a button labeled 'Create new team'. The main section of the screen is divided into tabs labeled Workspace, GPUaaS Pool, and Shared storage, with the GPUaaS Pool tab currently active. The central area shows the GPUaaS billing usage in a daily format, with a toggle to switch between HeatMap and Graph views. The overall GPUaaS cost is prominently displayed as $18,199.76. Below this, a graphical representation of daily costs from October 25 to November 15 is shown, with specific daily costs highlighted in purple for November 10 to November 15. On the right side of the graph, there is a checkbox labeled 'ci_pool1'. At the bottom, there is a section titled 'GPUaaS pool overview' for the period from October 15, 2025, to November 15, 2025, showing the total cost of $18,199.76 and a link labeled 'Pool usage'. The top right corner features a user profile section with the label 'Default admin user' and a 'Create' button next to it.

Billing

Members

The screenshot displays the 'Teams / User Panel' feature of the HostedAI platform, specifically focusing on the 'Juggernaut Omega' team. On the left side, there is a navigation panel listing teams such as 'AI Team Alpha', 'ML Ops Team Delta', 'Juggernaut Omega', and 'John Doe', with an option to create a new team at the bottom. The main section shows details for 'Juggernaut Omega', including the creation date, number of members, and workspaces. Tabs for 'Overview', 'Billing', 'Members', 'Workspaces', 'Accelerators', and 'Shared Storage' are available, with the 'Accelerators' tab currently active. Below, there is a search bar for accelerators and a list showing 'ci_pool1' with details like vRAM usage, TFLOPS, and cost. The 'GPUaaS Instance 1' is detailed with its image, CPU, RAM, and vGPU specifications. The 'Subscribe to Pool' button is located on the right side, and the status of the instance is shown as 'Running'. The top right corner features a 'Create' button and a user profile dropdown for the 'Default admin user'.

Accelerators

The screenshot displays the 'Shared Storage' section of the HostedAI platform within the Teams/User Panel. The left sidebar shows a list of teams, including 'AI Team Alpha', 'ML Ops Team Delta', 'Juggernaut Omega', and 'John Doe', with an option to 'Create new team' at the bottom. The main panel is focused on 'AI Team Alpha', created on November 13, 2025, with options to view '1 member' and '1 workspace'. The top navigation tabs include 'Overview', 'Billing', 'Members', 'Workspaces', 'Accelerators', and the currently selected 'Shared Storage'. Below, there is a search bar labeled 'Search for Shared Storage' and a message indicating 'No shared storage found'. On the right, there are dropdown menus for selecting 'All Regions' and another unspecified filter, along with a 'Create shared volume' button. The top right corner features a 'Create' button and a user profile section for the 'Default admin user', who is an Administrator.

Shared Storage

Teams

The screenshot displays the 'Manage Workspaces' section of the HostedAI platform, focusing on the 'NeuroTech Industries' team. The left sidebar shows a list of teams with 'Team 1: NeuroTech Industries' selected, displaying its workspaces: 'Production Workspace' and 'Default workspace'. Below, 'Team 2: NextAI Technologies' and 'Team 3: NeuroSphere Systems' are listed with dropdown arrows. A search bar allows filtering by team name, and a button at the bottom left corner labeled 'Create new team' enables adding new teams. The main panel provides details about 'NeuroTech Industries', including its creation date, a brief description, and buttons indicating '4 members' and '2 workspaces'. Tabs for 'Overview', 'Billing', 'Members', 'Workspaces', 'Instances', and 'GPUaaS' are available, with 'Workspaces' currently active. Below, two workspace cards are shown: 'Production Workspace' for production services and 'Default workspace' as the default. A 'Create workspace' button is located at the top right of the workspace section. The top right corner features a 'Create' button and a user profile dropdown for the 'Default admin user'.

Creating Workspaces (KVM)

Delete Workspace (KVM)

The screenshot displays the 'Workspaces' section of the HostedAI platform's User Panel. On the left, there is a sidebar with icons for navigation, including a user icon, a team icon, and a settings icon. Below the sidebar, there is a 'Create new team' button. The main area shows details for 'NeuroTech Industries,' including the creation date and a description of 'Measuring brain activity.' There are buttons for '4 members' and '2 workspaces.' Tabs for 'Overview,' 'Billing,' 'Members,' 'Workspaces,' 'Instances,' and 'GPUaaS' are present, with 'Workspaces' currently selected. Resource allocation is displayed with cards for CPU, GPU, RAM, and Storage, each showing usage and limits. The top right features a 'Create' button and a user profile dropdown labeled 'Default admin user, Administrator.'

Manage Workspaces (KVM)

Policies

Objects

Policies and Objects

The SMTP server settings screen in the HostedAI platform allows users to configure the outgoing email server and customize email settings. The page is divided into two main sections: 'Server settings' and 'Email identity information'. The 'Server settings' section includes fields for 'Server address' set to smtp.maileroo.com, 'Port' set to 465,587, 'Authentication' which is enabled, and 'Encryption method' which is SSL Enabled. An 'Edit' button is located to the right of this section for making changes. The 'Email identity information' section contains fields for 'From address' set to vip@web.hosted.ai, 'From display name' as Hosted AI, 'Reply to address' also set to vip@web.hosted.ai, and 'Footer information' which includes a physical address, support contact information, and a link to the Privacy policy. Another 'Edit' button is available to the right of this section. At the top right, there is a 'Test connection' button to verify the settings. The user interface also includes a navigation panel on the left with options like 'Settings' and 'Audit logs'.

SMTP Settings

The screenshot displays the Global Teams Management section of the HostedAI platform, which allows users to manage configuration settings that affect all teams. It includes a heading 'Global teams management' with a subheading advising caution when changing settings. Below, there is a 'Settings' button on the left side. The 'Team creation settings' section is visible, explaining the ability for team members to create additional teams. To the right, there are two buttons: 'Disabled' and 'Manage', which allow users to toggle and manage team creation settings.

Teams Management

Global Configuration

My Profile

Security

Account Settings

GPUaaS FAQs

Billing FAQs

API FAQs

Instance FAQs

Pricing and Resource Policies FAQs

Update CUDA Version from Inside a Pod

Useful GPUaaS Articles

FAQs

Integrations

User Panel

Regions

GPUaaS Node Preparation

The screenshot displays the HostedAI Admin Panel interface, specifically the section for managing GPUaaS Nodes. On the left side, there is a vertical navigation menu under the 'Sections' heading, which includes options such as Dashboard, Infrastructure (expanded to show Regions, Clusters, GPUaaS Nodes, and Resource Groups), GPUaaS, App Library, Health Status, Settings, and Logout. The main area of the screen shows a message 'GPUaaS Node not added' with an illustration of a GPU card. Below this message, there is a prominent blue button labeled 'Add GPUaaS Node' located towards the bottom right of the main content area. The top right corner of the screen displays the user profile icon with the username 'Hosted_ai Admin' and a dropdown for selecting the country and language, currently set to Germany and DE.

Adding a GPUaaS Node

The screenshot displays the 'GPUaaS Nodes' section of the HostedAI Admin Panel, which is part of the Infrastructure menu. The main area lists nodes with columns for Name, Initialized status, GPUaaS Status, Role, IP Address, Port, and Username. Each node entry includes options to 'Test connection' and 'Rescan GPUs' on the right. The top right corner features a button to 'Add GPUaaS Node' and a status indicator showing 'GPUaaS Enabled'. The left sidebar contains navigation links for Dashboard, Infrastructure (expanded to show Regions, Clusters, GPUaaS Nodes, Resource Groups, GPUs), Backups, Jobs, GPUaaS, and App Library. The user profile is shown in the top right corner with a language selector set to 'United States'.

Testing Connection

The screenshot displays the GPUaaS Nodes section within the HostedAI Admin Panel, which is part of the Infrastructure menu. The screen is used for managing nodes that provide GPU-as-a-Service. At the top right, there are buttons labeled 'GPUaaS Enabled' and 'Add GPUaaS Node' for enabling the service and adding new nodes, respectively. The node listing table includes columns for node name, GPUaaS status, role, IP address, SSH port, and username. The node 'node_1' is shown as initialized with a checkmark, and its role is listed as Controller, Worker, Storage Service. The IP address is 194.103.198.230, the SSH port is 22, and the username is 'administrator'. To the right of the node details, there are buttons for 'Test connection' and 'Rescan Resources'. The 'Rescan Resources' button has a dropdown menu with options such as Enable GPUaaS, Add as Redundant Controller, Reset Node, Display SSH Keys, Logs, Initialize Node, Edit Node, Delete Node, and Reboot Node, allowing for various node management actions.

Initializing the Node and Enabling GPUaaS

The screenshot displays the 'GPU as a Service' section of the hosted.ai Admin Panel, specifically focusing on creating a GPUaaS Pool. On the left, a navigation menu includes options such as Dashboard, Infrastructure, Backups, Jobs, GPUaaS, App Library, Health Status, Settings, Updates, and Logout. The main area shows existing pools in the United States, listing details like pool names, GPU types, and time quantum. On the right, a panel titled 'Add Pool' allows users to create a new GPU pool by selecting a region, entering a pool name, and choosing a GPU accelerator. The 'Add Pool' panel includes fields for 'Selected region' with a dropdown, 'Pool name' as a text input, and 'Select accelerator (GPU)' as a dropdown menu. The 'Add Pool' panel is located on the right side of the screen.

Creating a GPUaaS Pool

GPUaaS Infrastructure API

GPUaaS

Admin Panel

Cluster Panel Support

Data Protection

Cluster Panel (KVM)

Instances VS GPUaaS

Node Types

GPUaaS Access Methods

Compatibility Matrix

GPUaaS Nodes Roles

The screenshot illustrates the architecture of the HostedAI platform, focusing on deployment models for managing GPU resources. It shows two users, User A in Namespace A and User B in Namespace B, each with their own Resource Access interfaces. These interfaces provide connection details for accessing resources via Ray and SSH, with specific IP addresses and ports. Below the interfaces, there are Task Schedulers for each user, indicating the management of tasks with a priority-aware scheduler. The lower section displays the HAIShare Core and Memory Manager, which includes multiple CPUs and GPUs organized into two pools, Pool 1 and Pool 2, demonstrating the allocation and management of GPU resources.

Deployment Models

Terminology

Platform Architecture

Networking

Access by persona

Architecture components

The screenshot displays the connection information for a GPUaaS instance labeled 'vanilla-worker-2-5399-pool10'. It includes an SSH connection string 'ssh ubuntu@101.136.12.113...' with a copy icon next to it for easy copying of the SSH command. Below the SSH information, there is a password field represented by asterisks with a copy icon and an eye icon, likely for copying and revealing the password, respectively. On the right side, there is a service labeled 'ollama' with a URL 'http://101.136.12.113:200...' and a copy icon for copying the URL. Below this, a button labeled 'Discovered service' is present, indicating a service that has been identified. Two additional sections labeled 'No service' are present, indicating areas where services could be listed but are currently not available.

Service Exposure (2.0.1)

Billing and Metering

Analyze Billing Details

ollama

Service Exposure Sample Apps

AI Tools

Further information - Architecture and compatibility

The image depicts an anthropomorphic red panda character wearing glasses and a headset, sitting at a desk and typing on a keyboard. The character is wearing a hoodie with the text 'hosted.ai' on it, suggesting a connection to the HostedAI platform. In the background, a computer monitor displays a chat interface, possibly related to support or communication features of the platform. The desk is cluttered with sticky notes and an open notebook, indicating a busy work environment. A mug with a logo and a desk lamp are also present, adding to the office setting.

Model	T4	V100	L4	L40S	A2	A100 PCIe	A100 SXM4	H100 PCIe	H200 SXM5	B100 SXM5
Bare Metal	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
KVM	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
NexVisor	✅	✅	✅	✅	✅	✅	❌	✅	❌	❌

	GPUaaS Passthrough	GPUaaS Pools
Bare Metal	❌	✅
KVM	✅	✅
NexVisor	✅	✅

Model	Virtual Machines	Containers
Bare Metal	❌	✅
KVM	✅	✅
NexVisor	✅	✅

Compatibility Matrix

Nvidia GPU model support per node type

GPUaaS type per node type

Types of Instances per node type