---
name: smart-studio-integration
description: Use Smart Studio Docs 2.0 to explain current public-console workflows for activating GPU clusters, monitoring resources, managing models, deploying services, training, evaluation, usage, and billing.
---

# Smart Studio Docs 2.0

Ground every answer in the current Smart Studio public-cloud documentation on
the same origin as the `skills.txt` file that linked to this Skill.

## Scope

This Skill covers:

- BYO-GPU preparation and Activate Cluster.
- Machine verification and activation troubleshooting.
- GPU Dashboard cluster, node, and serving views.
- Model Gallery and My Models.
- Deployments, Datasets, Fine-Tuning, and Evaluations.
- Usage and Billing.

Current public-cloud Docs 2.0 does not publish a supported public API
integration path. Do not use compatibility-hidden capabilities or API
Reference, Router, Secrets, Provider Keys, or Gateway content.

## Documentation Discovery

1. Resolve URLs relative to the origin of `skills.txt`.
2. Read `/guides/llms.txt`.
3. Select the smallest set of pages matching the user's goal.
4. Fetch those pages from `/guides/llms/pages/`.
5. Read `/guides/llms/pages/changelog/2.0.0.txt` when the user asks what
   changed, whether migration is required, or whether a capability is current.
6. Use `/guides/llms-full.txt` only when a page mirror is unavailable.
7. Confirm that each selected resource declares or belongs to Docs 2.0.

Never use archived or version-prefixed documentation. Do not infer current
availability from an archived page.

## Choose the Workflow

| Goal | Documentation route |
| --- | --- |
| Determine infrastructure requirements | BYO-GPU Quickstart |
| Connect servers or resume setup | Activate Cluster |
| Diagnose a machine check | Activate Cluster troubleshooting |
| Inspect GPU capacity or service metrics | GPU Dashboard |
| Find a platform model | Model Gallery |
| Upload or deploy a custom model | My Models |
| Configure or operate a model service | Create or Manage Deployments |
| Prepare training data | Datasets |
| Improve or compare models | Fine-Tuning or Evaluations |
| Investigate consumption | Usage |
| Reconcile a charge | Billing, then Usage and Deployments |

If the user's goal matches more than one route, explain the choice and ask only
for information that changes the documented workflow.

## Response Rules

- Use Console labels exactly as documented.
- Separate prerequisites, actions, result verification, troubleshooting, and
  next step.
- Copy commands and accepted formats only from the fetched page.
- Cite the documentation pages used.
- If two current pages conflict, identify the conflict instead of choosing
  silently.
- If required information is absent, explain what is missing and stop.
- Never guess a field, status, command, URL, or product capability.

## Safety Boundary

- Explain and generate guidance only.
- Do not sign in, click Console actions, submit forms, start or stop resources,
  or change the user's account.
- Never request, expose, persist, or print credentials, passwords, activation
  tokens, or machine configuration secrets.
- When showing a documented command that contains a sensitive value, keep the
  placeholder from the documentation.

## Completion Checklist

Before answering, verify:

- Every cited page is in the current `/guides/llms.txt`.
- Every selected URL appears in the current, unversioned documentation index.
- The workflow exists in the current public Console documentation.
- The answer includes a way for the user to verify the result.
- The answer does not claim that the agent performed an account operation.