Sandboxes
The OLLM Sandboxes dashboard gives you visibility into compute sandboxes running under your organization. Sandboxes are created and managed in ORGN, OLLM provides a read-only view for monitoring and tracking.
Sandboxes are created and used exclusively in ORGN. The OLLM Sandboxes dashboard is a read-only monitoring view, you cannot create or control sandboxes from here.
What Are Sandboxes
Sandboxes are isolated Linux compute environments. Each sandbox is a small, self-contained system with its own CPU, memory, and disk allocation, designed for running code securely without interfering with other workloads or the host system.
Sandboxes run on TDX-backed node pools, meaning the underlying infrastructure inherits the same confidential compute foundation as the rest of the OLLM platform.
Sandbox creation and usage happens in ORGN. The Sandboxes section in OLLM exists so you can monitor all sandboxes running under your organization from a single place, without switching contexts.
Dashboard Overview
The top of the Sandboxes page shows a live summary of your organization's sandbox activity.
| Metric | Description |
|---|---|
| Total Sandboxes | All sandboxes ever created under your organization |
| Total Compute Min | Cumulative compute minutes consumed across all sandboxes |
| Today's Sandboxes | Number of sandboxes created in the current day |
| Today's Compute Min | Compute minutes consumed today |
| Running Now | Sandboxes currently active |
| Active CPU Cores | CPU cores in use across all running sandboxes |
| Active RAM | Memory in use across all running sandboxes |
| Active Storage | Disk in use across all running sandboxes |
These metrics update in real time and give you an instant read on how much of your organization's compute capacity is being used.
Capacity vs. Consumption
Below the summary metrics, a set of ring gauges shows current utilization as a percentage of total available capacity across three dimensions:
- vCPU: CPU cores in use versus total vCPUs available in the pool
- RAM: Memory in use versus total memory in the pool
- Storage: Disk in use versus total storage capacity
These gauges reflect the aggregate capacity of the node pool backing sandboxes (the sandbox-c3-44 pool). Low utilization means new sandboxes can be provisioned immediately. As utilization climbs, the pool scales up automatically within its configured limits.
The Daily Compute Minutes chart beneath the gauges shows a bar-per-day breakdown of compute consumed over the past two weeks, giving you a quick view of usage trends.
Node Pools
The Node Pools panel shows the underlying infrastructure that sandboxes and the control plane run on. There are three pools, all tagged as TDX (Intel Trust Domain Extensions), meaning they operate as confidential virtual machines.
| Pool | Role | Machine Type |
|---|---|---|
tdx-us-central1 | General TDX workloads | c3-standard-4 |
control-tdx | Control plane | c3-standard-4 |
sandbox-c3-44 | Sandbox execution (Ubuntu 24.04 LTS) | c3-standard-44 |
Each pool has a defined scale range. The sandbox pool (sandbox-c3-44) scales between 1 and 110 nodes and provisions each sandbox with 44 vCPUs and 176 GiB of RAM per node, with 750 GB of balanced persistent disk. This is the pool your sandboxes run on.
The control-tdx pool is fixed at 3 nodes and handles orchestration. The tdx-us-central1 pool handles general-purpose TDX workloads and scales 1–5.
Sandbox List
The sandbox table lists every sandbox your organization has created, sorted by creation time with the most recent at the top.
Each row shows:
- Sandbox ID: A unique UUID identifying the sandbox instance
- Resources: CPU cores, RAM, and disk allocated to that sandbox (e.g. 4 CPU / 8 GB / 10 GB)
- Created: The exact timestamp when the sandbox was provisioned
- Status: Current state of the sandbox
Status values
| Status | Meaning |
|---|---|
| Running | The sandbox is active and consuming compute |
| Completed | The sandbox has finished and its resources have been released |
Use the Show Running button in the top-right corner of the table to filter the list to only active sandboxes. Use Refresh to pull the latest state.
Sandbox Details
Clicking any sandbox ID in the list opens its detail view. This gives you a complete picture of a single sandbox instance.
Basic Info
| Field | Description |
|---|---|
| Sandbox ID | Full UUID of the sandbox |
| Created | Timestamp and relative time since creation |
| Status | Current state (Running or Completed) |
| Region | The physical datacenter region where the sandbox is running (e.g. US Council Bluffs, Iowa) |
Resources
| Field | Description |
|---|---|
| CPU | Cores allocated to this sandbox / total vCPUs on the node |
| Memory | RAM allocated to this sandbox / total RAM on the node |
| Disk | Storage allocated / total pool capacity (XFS) |
| GPU | GPU attached to the sandbox, if any (None for standard sandboxes) |
Each sandbox gets a dedicated slice of the node's resources. The fractions shown (e.g. 4 cores / 44 vCPU) reflect how the sandbox is carved out of its host node.
Operational Timeline
The timeline at the bottom of the detail view shows uptime history for this sandbox as a segmented bar. Each segment represents a time interval; green indicates the sandbox was running and healthy during that period. This is useful for confirming that a sandbox ran continuously or identifying any gaps in its operational history.
OLLM Integration with the Vercel AI SDK
Step-by-step guide to integrating OLLM with the Vercel AI SDK. Configure the OpenAI-compatible provider for secure, TEE-backed LLM inference in your Next.js application.
Generate Themes from a Website with OLLM
Build an automated workflow that scrapes a website and generates design themes using OLLM confidential inference. Includes step-by-step code examples for extraction and theme generation.