Skip to main content

Documentation Index

Fetch the complete documentation index at: https://runcrate.ai/docs/llms.txt

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

When a new team member joins, your AI agent can walk them through the workspace: show available templates, explain the environment structure, create a dev environment, and deploy their first instance.

​
”I just joined the team. What resources and templates are available?”

The agent gives an overview:
  1. list_templates — shows available instance templates
  2. list_environments — shows the environment structure
  3. list_gpu_types — shows GPU options with pricing
  4. billing_balance — checks workspace credits
Welcome to the workspace. Here’s what’s available: Instance templates:
TemplateDescription
ubuntu-devboxUbuntu 22.04 with Python 3.11, CUDA 12.4, dev tools
ubuntu-trainUbuntu with PyTorch, DeepSpeed, NCCL pre-installed
ubuntu-inferenceUbuntu with vLLM, TGI, optimized for serving
Environments: main (default), staging, production GPU options:
GPUVRAMPrice/hr
RTX 409024 GB$0.35/hr
A100 80 GB80 GB$1.60/hr
H100 80 GB80 GB$2.50/hr
Workspace balance: $1,247.80 For development, I’d recommend starting with an RTX 4090 ($0.35/hr) using the ubuntu-devbox template.

​
”Create a dev environment for me and deploy my first instance.”

The agent sets up the new team member:
  1. create_environment — creates a dev-alex environment
  2. create_instance — deploys alex-devbox with RTX 4090 using the devbox template
  3. instance_status — waits until running
  4. get_instance — retrieves connection details
  • Created environment dev-alex
  • Deployed alex-devbox (RTX 4090, 24 GB VRAM, $0.35/hr)
Connect: ssh root@203.0.113.55 Pre-installed: Python 3.11, CUDA 12.4, pip, git, htop, tmux. Your instance runs in the dev-alex environment, separate from staging/production. All environments share the workspace credit balance. Terminate when done to stop billing.

​
”Show me what the team is running so I understand the setup.”

The agent maps out the workspace:
  1. list_environments — lists all environments
  2. list_instances — lists all instances across environments
  3. list_volumes — lists all storage
  4. list_ssh_keys — lists team SSH keys
Team infrastructure overview: production (do not modify without approval)
  • prod-inference (H100) — Llama 70B vLLM server, running 12 days
  • model-weights volume (500 GB) — production model checkpoints
staging
  • staging-test (A100) — testing new model versions before promotion
main (shared development)
  • dev-box (RTX 4090) — Sarah’s dev instance
dev-alex (your environment)
  • alex-devbox (RTX 4090) — your dev instance
Team SSH keys: 4 keys registered (macbook-pro, ci-deploy, workstation, alex-laptop) Ground rules: Production instances are managed by the ML ops lead. Staging is for pre-production testing. Use your own environment (dev-alex) for experiments. Terminate instances when not in use.

​
Tools used in this workflow

ToolPurpose
list_templatesShow available instance templates
list_environments / create_environmentView and create environments
list_gpu_typesShow GPU options and pricing
billing_balanceCheck workspace budget
create_instance / instance_status / get_instanceDeploy and connect
list_instances / list_volumes / list_ssh_keysMap team infrastructure